Merge 1f9692cd04 into 9a08757a70

Implement facet search extraction

Cargo.lock

@@ -80,7 +80,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e01ed3140b2f8d422c68afa1ed2e85d996ea619c988ac834d255db32138655cb"
 dependencies = [
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -216,7 +216,7 @@ dependencies = [
  "actix-router",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -296,9 +296,9 @@ dependencies = [
 
 [[package]]
 name = "allocator-api2"
-version = "0.2.16"
+version = "0.2.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0942ffc6dcaadf03badf6e6a2d0228460359d5e34b57ccdc720b7382dfbd5ec5"
+checksum = "5c6cb57a04249c6480766f7f7cef5467412af1490f8d1e243141daddada3264f"
 
 [[package]]
 name = "anes"
@@ -441,7 +441,7 @@ checksum = "6e0c28dcc82d7c8ead5cb13beb15405b57b8546e93215673ff8ca0349a028107"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -494,11 +494,13 @@ name = "benchmarks"
 version = "1.11.0"
 dependencies = [
  "anyhow",
+ "bumpalo",
  "bytes",
  "convert_case 0.6.0",
  "criterion",
  "csv",
  "flate2",
+ "memmap2",
  "milli",
  "mimalloc",
  "rand",
@@ -506,6 +508,7 @@ dependencies = [
  "reqwest",
  "roaring",
  "serde_json",
+ "tempfile",
 ]
 
 [[package]]
@@ -549,7 +552,7 @@ dependencies = [
  "regex",
  "rustc-hash 1.1.0",
  "shlex",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -593,6 +596,15 @@ dependencies = [
  "serde",
 ]
 
+[[package]]
+name = "bitpacking"
+version = "0.9.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4c1d3e2bfd8d06048a179f7b17afc3188effa10385e7b00dc65af6aae732ea92"
+dependencies = [
+ "crunchy",
+]
+
 [[package]]
 name = "bitvec"
 version = "1.0.1"
@@ -634,7 +646,7 @@ dependencies = [
  "proc-macro-crate",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
  "syn_derive",
 ]
 
@@ -684,6 +696,10 @@ name = "bumpalo"
 version = "3.16.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "79296716171880943b8470b5f8d03aa55eb2e645a4874bdbb28adb49162e012c"
+dependencies = [
+ "allocator-api2",
+ "serde",
+]
 
 [[package]]
 name = "byte-unit"
@@ -741,7 +757,7 @@ checksum = "4da9a32f3fed317401fa3c862968128267c3106685286e15d5aaa3d7389c2f60"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -954,8 +970,7 @@ dependencies = [
 [[package]]
 name = "charabia"
 version = "0.9.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "55ff52497324e7d168505a16949ae836c14595606fab94687238d2f6c8d4c798"
+source = "git+https://github.com/meilisearch/charabia?branch=mutualize-char-normalizer#f8d8308cdb8db80819be7eeed5652cc4a995cc71"
 dependencies = [
  "aho-corasick",
  "csv",
@@ -1052,7 +1067,7 @@ dependencies = [
  "heck 0.5.0",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1378,7 +1393,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "strsim 0.11.1",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1400,7 +1415,7 @@ checksum = "733cabb43482b1a1b53eee8583c2b9e8684d592215ea83efd305dd31bc2f0178"
 dependencies = [
  "darling_core 0.20.9",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1454,7 +1469,7 @@ checksum = "67e77553c4162a157adbf834ebae5b415acbecbeafc7a74b0e886657506a7611"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1496,7 +1511,7 @@ dependencies = [
  "darling 0.20.9",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1516,7 +1531,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "206868b8242f27cecce124c19fd88157fbd0dd334df2587f36417bafbc85097b"
 dependencies = [
  "derive_builder_core 0.20.0",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1558,7 +1573,7 @@ dependencies = [
  "convert_case 0.6.0",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1622,7 +1637,7 @@ checksum = "97369cbbc041bc366949bc74d34658d6cda5621039731c6310521892a3a20ae0"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1780,7 +1795,7 @@ dependencies = [
  "heck 0.4.1",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1800,7 +1815,7 @@ checksum = "a1ab991c1362ac86c61ab6f556cff143daa22e5a15e4e189df818b2fd19fe65b"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1848,9 +1863,9 @@ dependencies = [
 
 [[package]]
 name = "fastrand"
-version = "2.1.0"
+version = "2.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9fc0510504f03c51ada170672ac806f1f105a88aa97a5281117e1ddc3368e51a"
+checksum = "486f806e73c5707928240ddc295403b1b93c96a02038563881c4a2fd84b81ac4"
 
 [[package]]
 name = "file-store"
@@ -1908,6 +1923,12 @@ version = "1.0.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
 
+[[package]]
+name = "foldhash"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f81ec6369c545a7d40e4589b5597581fa1c441fe1cce96dd1de43159910a36a2"
+
 [[package]]
 name = "form_urlencoded"
 version = "1.2.1"
@@ -1985,7 +2006,7 @@ checksum = "87750cf4b7a4c0625b1529e4c543c2182106e4dedc60a2a6455e00d212c489ac"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -2023,7 +2044,9 @@ name = "fuzzers"
 version = "1.11.0"
 dependencies = [
  "arbitrary",
+ "bumpalo",
  "clap",
+ "either",
  "fastrand",
  "milli",
  "serde",
@@ -2241,11 +2264,11 @@ checksum = "d2fabcfbdc87f4758337ca535fb41a6d701b65693ce38287d856d1674551ec9b"
 [[package]]
 name = "grenad"
 version = "0.4.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "350d89047298d3b1b40050acd11ab76e487b854a104b760ebc5a7f375093de77"
+source = "git+https://github.com/meilisearch/grenad?branch=various-improvements#58ac87d852413571102f44c5e55ca13509a3f1a0"
 dependencies = [
  "bytemuck",
  "byteorder",
+ "either",
  "rayon",
  "tempfile",
 ]
@@ -2336,6 +2359,18 @@ dependencies = [
  "allocator-api2",
 ]
 
+[[package]]
+name = "hashbrown"
+version = "0.15.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3a9bfc1af68b1726ea47d3d5109de126281def866b33970e10fbab11b5dafab3"
+dependencies = [
+ "allocator-api2",
+ "equivalent",
+ "foldhash",
+ "serde",
+]
+
 [[package]]
 name = "heapless"
 version = "0.8.0"
@@ -2578,6 +2613,7 @@ dependencies = [
  "arroy 0.5.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "big_s",
  "bincode",
+ "bumpalo",
  "crossbeam",
  "csv",
  "derive_builder 0.20.0",
@@ -2590,7 +2626,9 @@ dependencies = [
  "meili-snap",
  "meilisearch-auth",
  "meilisearch-types",
+ "memmap2",
  "page_size",
+ "raw-collections",
  "rayon",
  "roaring",
  "serde",
@@ -2670,8 +2708,7 @@ checksum = "28b29a3cd74f0f4598934efe3aeba42bae0eb4680554128851ebbecb02af14e6"
 [[package]]
 name = "irg-kvariants"
 version = "0.1.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ef2af7c331f2536964a32b78a7d2e0963d78b42f4a76323b16cc7d94b1ddce26"
+source = "git+https://github.com/meilisearch/charabia?branch=mutualize-char-normalizer#f8d8308cdb8db80819be7eeed5652cc4a995cc71"
 dependencies = [
  "csv",
  "once_cell",
@@ -2835,9 +2872,9 @@ dependencies = [
 
 [[package]]
 name = "libc"
-version = "0.2.155"
+version = "0.2.164"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97b3888a4aecf77e811145cadf6eef5901f4782c53886191b2f693f24761847c"
+checksum = "433bfe06b8c75da9b2e3fbea6e5329ff87748f0b144ef75306e674c3f6f7c13f"
 
 [[package]]
 name = "libgit2-sys"
@@ -3221,9 +3258,9 @@ checksum = "0717cef1bc8b636c6e1c1bbdefc09e6322da8a9321966e8928ef80d20f7f770f"
 
 [[package]]
 name = "linux-raw-sys"
-version = "0.4.12"
+version = "0.4.14"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c4cd1a83af159aa67994778be9070f0ae1bd732942279cabb14f86f986a21456"
+checksum = "78b3ae25bc7c8c38cec158d1f2757ee79e9b3740fbc7ccf0e59e4b08d793fa89"
 
 [[package]]
 name = "liquid"
@@ -3264,7 +3301,7 @@ checksum = "915f6d0a2963a27cd5205c1902f32ddfe3bc035816afd268cf88c0fc0f8d287e"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3368,7 +3405,7 @@ dependencies = [
  "once_cell",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3507,6 +3544,7 @@ version = "1.11.0"
 dependencies = [
  "actix-web",
  "anyhow",
+ "bumpalo",
  "convert_case 0.6.0",
  "csv",
  "deserr",
@@ -3519,6 +3557,7 @@ dependencies = [
  "meili-snap",
  "memmap2",
  "milli",
+ "raw-collections",
  "roaring",
  "serde",
  "serde-cs",
@@ -3555,9 +3594,9 @@ checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"
 
 [[package]]
 name = "memmap2"
-version = "0.9.4"
+version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fe751422e4a8caa417e13c3ea66452215d7d63e19e604f4980461212f3ae1322"
+checksum = "fd3f7eed9d3848f8b98834af67102b720745c4ec028fcd0aa0239277e7de374f"
 dependencies = [
  "libc",
  "stable_deref_trait",
@@ -3567,11 +3606,13 @@ dependencies = [
 name = "milli"
 version = "1.11.0"
 dependencies = [
+ "allocator-api2",
  "arroy 0.5.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "big_s",
  "bimap",
  "bincode",
  "bstr",
+ "bumpalo",
  "bytemuck",
  "byteorder",
  "candle-core",
@@ -3583,12 +3624,14 @@ dependencies = [
  "csv",
  "deserr",
  "either",
+ "enum-iterator",
  "filter-parser",
  "flatten-serde-json",
  "fst",
  "fxhash",
  "geoutils",
  "grenad",
+ "hashbrown 0.15.1",
  "heed",
  "hf-hub",
  "indexmap",
@@ -3607,11 +3650,13 @@ dependencies = [
  "once_cell",
  "ordered-float",
  "rand",
+ "raw-collections",
  "rayon",
  "rayon-par-bridge",
  "rhai",
  "roaring",
  "rstar",
+ "rustc-hash 2.0.0",
  "serde",
  "serde_json",
  "slice-group-by",
@@ -3620,10 +3665,12 @@ dependencies = [
  "smartstring",
  "tempfile",
  "thiserror",
+ "thread_local",
  "tiktoken-rs",
  "time",
  "tokenizers",
  "tracing",
+ "uell",
  "ureq",
  "url",
  "uuid",
@@ -3699,7 +3746,7 @@ checksum = "371717c0a5543d6a800cac822eac735aa7d2d2fbb41002e9856a4089532dbdce"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3835,7 +3882,7 @@ dependencies = [
  "proc-macro-crate",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3864,9 +3911,8 @@ dependencies = [
 
 [[package]]
 name = "obkv"
-version = "0.2.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a2e27bcfe835a379d32352112f6b8dbae2d99d16a5fff42abe6e5ba5386c1e5a"
+version = "0.3.0"
+source = "git+https://github.com/kerollmops/obkv?branch=unsized-kvreader#ce535874008ecac554f02e0c670e6caf62134d6b"
 
 [[package]]
 name = "once_cell"
@@ -4047,7 +4093,7 @@ dependencies = [
  "pest_meta",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -4101,7 +4147,7 @@ dependencies = [
  "phf_shared",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -4130,7 +4176,7 @@ checksum = "266c042b60c9c76b8d53061e52b2e0d1116abc57cefc8c5cd671619a56ac3690"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -4247,9 +4293,9 @@ dependencies = [
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.81"
+version = "1.0.89"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3d1597b0c024618f09a9c3b8655b7e430397a36d23fdafec26d6965e9eec3eba"
+checksum = "f139b0662de085916d1fb67d2b4169d1addddda1919e696f3252b740b629986e"
 dependencies = [
  "unicode-ident",
 ]
@@ -4434,6 +4480,19 @@ dependencies = [
  "rand",
 ]
 
+[[package]]
+name = "raw-collections"
+version = "0.1.0"
+source = "git+https://github.com/dureuill/raw-collections.git#15e5d7bdebc0c149b2a28b2454f307c717d07f8a"
+dependencies = [
+ "allocator-api2",
+ "bitpacking",
+ "bumpalo",
+ "hashbrown 0.15.1",
+ "serde",
+ "serde_json",
+]
+
 [[package]]
 name = "raw-cpuid"
 version = "10.7.0"
@@ -4631,7 +4690,7 @@ source = "git+https://github.com/rhaiscript/rhai?rev=ef3df63121d27aacd838f366f2b
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -4681,8 +4740,7 @@ dependencies = [
 [[package]]
 name = "roaring"
 version = "0.10.6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f4b84ba6e838ceb47b41de5194a60244fac43d9fe03b71dbe8c5a201081d6d1"
+source = "git+https://github.com/RoaringBitmap/roaring-rs?branch=clone-iter-slice#8ff028e484fb6192a0acf5a669eaf18c30cada6e"
 dependencies = [
  "bytemuck",
  "byteorder",
@@ -4746,9 +4804,9 @@ dependencies = [
 
 [[package]]
 name = "rustix"
-version = "0.38.31"
+version = "0.38.41"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ea3e1a662af26cd7a3ba09c0297a31af215563ecf42817c98df621387f4e949"
+checksum = "d7f649912bc1495e167a6edee79151c84b1bad49748cb4f1f1167f459f6224f6"
 dependencies = [
  "bitflags 2.6.0",
  "errno",
@@ -4873,9 +4931,9 @@ checksum = "a3f0bf26fd526d2a95683cd0f87bf103b8539e2ca1ef48ce002d67aad59aa0b4"
 
 [[package]]
 name = "serde"
-version = "1.0.209"
+version = "1.0.214"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "99fce0ffe7310761ca6bf9faf5115afbc19688edd00171d81b1bb1b116c63e09"
+checksum = "f55c3193aca71c12ad7890f1785d2b73e1b9f63a0bbc353c08ef26fe03fc56b5"
 dependencies = [
  "serde_derive",
 ]
@@ -4891,23 +4949,24 @@ dependencies = [
 
 [[package]]
 name = "serde_derive"
-version = "1.0.209"
+version = "1.0.214"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a5831b979fd7b5439637af1752d535ff49f4860c0f341d1baeb6faf0f4242170"
+checksum = "de523f781f095e28fa605cdce0f8307e451cc0fd14e2eb4cd2e98a355b147766"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
 name = "serde_json"
-version = "1.0.120"
+version = "1.0.132"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4e0d21c9a8cae1235ad58a00c11cb40d4b1e5c784f1ef2c537876ed6ffd8b7c5"
+checksum = "d726bfaff4b320266d395898905d0eba0345aae23b54aee3a737e260fd46db03"
 dependencies = [
  "indexmap",
  "itoa",
+ "memchr",
  "ryu",
  "serde",
 ]
@@ -5190,7 +5249,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "rustversion",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -5212,9 +5271,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.60"
+version = "2.0.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "909518bc7b1c9b779f1bbf07f2929d35af9f0f37e47c6e9ef7f9dddc1e1821f3"
+checksum = "25aa4ce346d03a6dcd68dd8b4010bcb74e54e62c90c573f394c46eae99aba32d"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -5230,7 +5289,7 @@ dependencies = [
  "proc-macro-error",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -5256,7 +5315,7 @@ checksum = "c8af7666ab7b6390ab78131fb5b0fce11d6b7a6951602017c35fa82800708971"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -5316,12 +5375,13 @@ dependencies = [
 
 [[package]]
 name = "tempfile"
-version = "3.10.1"
+version = "3.14.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "85b77fafb263dd9d05cbeac119526425676db3784113aa9295c88498cbf8bff1"
+checksum = "28cce251fcbc87fac86a866eeb0d6c2d536fc16d06f184bb61aeae11aa4cee0c"
 dependencies = [
  "cfg-if",
  "fastrand",
+ "once_cell",
  "rustix",
  "windows-sys 0.52.0",
 ]
@@ -5361,14 +5421,14 @@ checksum = "46c3384250002a6d5af4d114f2845d37b57521033f30d5c3f46c4d70e1197533"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
 name = "thread_local"
-version = "1.1.7"
+version = "1.1.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3fdd6f064ccff2d6567adcb3873ca630700f00b5ad3f060c25b5dcfd9a4ce152"
+checksum = "8b9ef9bad013ada3808854ceac7b46812a6465ba368859a37e2100283d2d719c"
 dependencies = [
  "cfg-if",
  "once_cell",
@@ -5513,7 +5573,7 @@ checksum = "5f5ae998a069d4b5aba8ee9dad856af7d520c3699e6159b185c2acd48155d39a"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -5645,7 +5705,7 @@ checksum = "34704c8d6ebcbc939824180af020566b01a7c01f80641264eba0999f6c2b6be7"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -5740,6 +5800,15 @@ version = "0.1.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ed646292ffc8188ef8ea4d1e0e0150fb15a5c2e12ad9b8fc191ae7a8a7f3c4b9"
 
+[[package]]
+name = "uell"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "40de5982e28612e20330e77d81f1559b74f66caf3c7fc10b19ada4843f4b4fd7"
+dependencies = [
+ "bumpalo",
+]
+
 [[package]]
 name = "unescaper"
 version = "0.1.5"
@@ -5991,7 +6060,7 @@ dependencies = [
  "once_cell",
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
  "wasm-bindgen-shared",
 ]
 
@@ -6025,7 +6094,7 @@ checksum = "e94f17b526d0a461a191c78ea52bbce64071ed5c04c9ffe424dcb38f74171bb7"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
  "wasm-bindgen-backend",
  "wasm-bindgen-shared",
 ]
@@ -6458,7 +6527,7 @@ checksum = "9e6936f0cce458098a201c245a11bef556c6a0181129c7034d10d76d1ec3a2b8"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
  "synstructure",
 ]
 
@@ -6479,7 +6548,7 @@ checksum = "9ce1b18ccd8e73a9321186f97e46f9f04b778851177567b1975109d26a08d2a6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -6499,7 +6568,7 @@ checksum = "e6a647510471d372f2e6c2e6b7219e44d8c574d24fdc11c610a61455782f18c3"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
  "synstructure",
 ]
 
@@ -6520,7 +6589,7 @@ checksum = "ce36e65b0d2999d2aafac989fb249189a141aee1f53c612c1f37d72631959f69"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.60",
+ "syn 2.0.87",
 ]
 
 [[package]]

Cargo.toml

@@ -44,23 +44,5 @@ opt-level = 3
 [profile.dev.package.roaring]
 opt-level = 3
 
-[profile.dev.package.lindera-ipadic-builder]
-opt-level = 3
-[profile.dev.package.encoding]
-opt-level = 3
-[profile.dev.package.yada]
-opt-level = 3
-
-[profile.release.package.lindera-ipadic-builder]
-opt-level = 3
-[profile.release.package.encoding]
-opt-level = 3
-[profile.release.package.yada]
-opt-level = 3
-
-[profile.bench.package.lindera-ipadic-builder]
-opt-level = 3
-[profile.bench.package.encoding]
-opt-level = 3
-[profile.bench.package.yada]
-opt-level = 3
+[patch.crates-io]
+roaring = { git = "https://github.com/RoaringBitmap/roaring-rs", branch = "clone-iter-slice" }

crates/benchmarks/Cargo.toml

@@ -12,10 +12,13 @@ license.workspace = true
 
 [dependencies]
 anyhow = "1.0.86"
+bumpalo = "3.16.0"
 csv = "1.3.0"
+memmap2 = "0.9.5"
 milli = { path = "../milli" }
 mimalloc = { version = "0.1.43", default-features = false }
 serde_json = { version = "1.0.120", features = ["preserve_order"] }
+tempfile = "3.14.0"
 
 [dev-dependencies]
 criterion = { version = "0.5.1", features = ["html_reports"] }

crates/benchmarks/benches/utils.rs

@@ -1,17 +1,18 @@
 #![allow(dead_code)]
 
 use std::fs::{create_dir_all, remove_dir_all, File};
-use std::io::{self, BufRead, BufReader, Cursor, Read, Seek};
-use std::num::ParseFloatError;
+use std::io::{self, BufReader, BufWriter, Read};
 use std::path::Path;
-use std::str::FromStr;
+use std::str::FromStr as _;
 
+use anyhow::Context;
+use bumpalo::Bump;
 use criterion::BenchmarkId;
-use milli::documents::{DocumentsBatchBuilder, DocumentsBatchReader};
+use memmap2::Mmap;
 use milli::heed::EnvOpenOptions;
-use milli::update::{
-    IndexDocuments, IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Settings,
-};
+use milli::update::new::indexer;
+use milli::update::{IndexDocumentsMethod, IndexerConfig, Settings};
+use milli::vector::EmbeddingConfigs;
 use milli::{Criterion, Filter, Index, Object, TermsMatchingStrategy};
 use serde_json::Value;
 
@@ -65,7 +66,7 @@ pub fn base_setup(conf: &Conf) -> Index {
 
     let mut options = EnvOpenOptions::new();
     options.map_size(100 * 1024 * 1024 * 1024); // 100 GB
-    options.max_readers(10);
+    options.max_readers(100);
     let index = Index::new(options, conf.database_name).unwrap();
 
     let config = IndexerConfig::default();
@@ -92,18 +93,43 @@ pub fn base_setup(conf: &Conf) -> Index {
 
     let config = IndexerConfig::default();
     let mut wtxn = index.write_txn().unwrap();
-    let indexing_config = IndexDocumentsConfig {
-        autogenerate_docids: conf.primary_key.is_none(),
-        update_method: IndexDocumentsMethod::ReplaceDocuments,
-        ..Default::default()
-    };
-    let builder =
-        IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| (), || false).unwrap();
+    let rtxn = index.read_txn().unwrap();
+    let db_fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
+    let mut new_fields_ids_map = db_fields_ids_map.clone();
+
     let documents = documents_from(conf.dataset, conf.dataset_format);
-    let (builder, user_error) = builder.add_documents(documents).unwrap();
-    user_error.unwrap();
-    builder.execute().unwrap();
+    let mut indexer = indexer::DocumentOperation::new(IndexDocumentsMethod::ReplaceDocuments);
+    indexer.add_documents(&documents).unwrap();
+
+    let indexer_alloc = Bump::new();
+    let (document_changes, _operation_stats, primary_key) = indexer
+        .into_changes(
+            &indexer_alloc,
+            &index,
+            &rtxn,
+            None,
+            &mut new_fields_ids_map,
+            &|| false,
+            &|_progress| (),
+        )
+        .unwrap();
+
+    indexer::index(
+        &mut wtxn,
+        &index,
+        config.grenad_parameters(),
+        &db_fields_ids_map,
+        new_fields_ids_map,
+        primary_key,
+        &document_changes,
+        EmbeddingConfigs::default(),
+        &|| false,
+        &|_| (),
+    )
+    .unwrap();
+
     wtxn.commit().unwrap();
+    drop(rtxn);
 
     index
 }
@@ -140,49 +166,96 @@ pub fn run_benches(c: &mut criterion::Criterion, confs: &[Conf]) {
     }
 }
 
-pub fn documents_from(filename: &str, filetype: &str) -> DocumentsBatchReader<impl BufRead + Seek> {
-    let reader = File::open(filename)
-        .unwrap_or_else(|_| panic!("could not find the dataset in: {}", filename));
-    let reader = BufReader::new(reader);
-    let documents = match filetype {
-        "csv" => documents_from_csv(reader).unwrap(),
-        "json" => documents_from_json(reader).unwrap(),
-        "jsonl" => documents_from_jsonl(reader).unwrap(),
-        otherwise => panic!("invalid update format {:?}", otherwise),
-    };
-    DocumentsBatchReader::from_reader(Cursor::new(documents)).unwrap()
+pub fn documents_from(filename: &str, filetype: &str) -> Mmap {
+    let file = File::open(filename)
+        .unwrap_or_else(|_| panic!("could not find the dataset in: {filename}"));
+    match filetype {
+        "csv" => documents_from_csv(file).unwrap(),
+        "json" => documents_from_json(file).unwrap(),
+        "jsonl" => documents_from_jsonl(file).unwrap(),
+        otherwise => panic!("invalid update format {otherwise:?}"),
+    }
 }
 
-fn documents_from_jsonl(reader: impl BufRead) -> anyhow::Result<Vec<u8>> {
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
+fn documents_from_jsonl(file: File) -> anyhow::Result<Mmap> {
+    unsafe { Mmap::map(&file).map_err(Into::into) }
+}
 
-    for result in serde_json::Deserializer::from_reader(reader).into_iter::<Object>() {
-        let object = result?;
-        documents.append_json_object(&object)?;
+fn documents_from_json(file: File) -> anyhow::Result<Mmap> {
+    let reader = BufReader::new(file);
+    let documents: Vec<milli::Object> = serde_json::from_reader(reader)?;
+    let mut output = tempfile::tempfile().map(BufWriter::new)?;
+
+    for document in documents {
+        serde_json::to_writer(&mut output, &document)?;
     }
 
-    documents.into_inner().map_err(Into::into)
+    let file = output.into_inner()?;
+    unsafe { Mmap::map(&file).map_err(Into::into) }
 }
 
-fn documents_from_json(reader: impl BufRead) -> anyhow::Result<Vec<u8>> {
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
+fn documents_from_csv(file: File) -> anyhow::Result<Mmap> {
+    let output = tempfile::tempfile()?;
+    let mut output = BufWriter::new(output);
+    let mut reader = csv::ReaderBuilder::new().from_reader(file);
 
-    documents.append_json_array(reader)?;
+    let headers = reader.headers().context("while retrieving headers")?.clone();
+    let typed_fields: Vec<_> = headers.iter().map(parse_csv_header).collect();
+    let mut object: serde_json::Map<_, _> =
+        typed_fields.iter().map(|(k, _)| (k.to_string(), Value::Null)).collect();
 
-    documents.into_inner().map_err(Into::into)
-}
+    let mut line = 0;
+    let mut record = csv::StringRecord::new();
+    while reader.read_record(&mut record).context("while reading a record")? {
+        // We increment here and not at the end of the loop
+        // to take the header offset into account.
+        line += 1;
 
-fn documents_from_csv(reader: impl BufRead) -> anyhow::Result<Vec<u8>> {
-    let csv = csv::Reader::from_reader(reader);
+        // Reset the document values
+        object.iter_mut().for_each(|(_, v)| *v = Value::Null);
 
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
-    documents.append_csv(csv)?;
+        for (i, (name, atype)) in typed_fields.iter().enumerate() {
+            let value = &record[i];
+            let trimmed_value = value.trim();
+            let value = match atype {
+                AllowedType::Number if trimmed_value.is_empty() => Value::Null,
+                AllowedType::Number => {
+                    match trimmed_value.parse::<i64>() {
+                        Ok(integer) => Value::from(integer),
+                        Err(_) => match trimmed_value.parse::<f64>() {
+                            Ok(float) => Value::from(float),
+                            Err(error) => {
+                                anyhow::bail!("document format error on line {line}: {error}. For value: {value}")
+                            }
+                        },
+                    }
+                }
+                AllowedType::Boolean if trimmed_value.is_empty() => Value::Null,
+                AllowedType::Boolean => match trimmed_value.parse::<bool>() {
+                    Ok(bool) => Value::from(bool),
+                    Err(error) => {
+                        anyhow::bail!(
+                            "document format error on line {line}: {error}. For value: {value}"
+                        )
+                    }
+                },
+                AllowedType::String if value.is_empty() => Value::Null,
+                AllowedType::String => Value::from(value),
+            };
 
-    documents.into_inner().map_err(Into::into)
+            *object.get_mut(name).expect("encountered an unknown field") = value;
+        }
+
+        serde_json::to_writer(&mut output, &object).context("while writing to disk")?;
+    }
+
+    let output = output.into_inner()?;
+    unsafe { Mmap::map(&output).map_err(Into::into) }
 }
 
 enum AllowedType {
     String,
+    Boolean,
     Number,
 }
 
@@ -191,8 +264,9 @@ fn parse_csv_header(header: &str) -> (String, AllowedType) {
     match header.rsplit_once(':') {
         Some((field_name, field_type)) => match field_type {
             "string" => (field_name.to_string(), AllowedType::String),
+            "boolean" => (field_name.to_string(), AllowedType::Boolean),
             "number" => (field_name.to_string(), AllowedType::Number),
-            // we may return an error in this case.
+            // if the pattern isn't recognized, we keep the whole field.
             _otherwise => (header.to_string(), AllowedType::String),
         },
         None => (header.to_string(), AllowedType::String),
@@ -230,10 +304,13 @@ impl<R: Read> Iterator for CSVDocumentDeserializer<R> {
                 for ((field_name, field_type), value) in
                     self.headers.iter().zip(csv_document.into_iter())
                 {
-                    let parsed_value: Result<Value, ParseFloatError> = match field_type {
+                    let parsed_value: anyhow::Result<Value> = match field_type {
                         AllowedType::Number => {
                             value.parse::<f64>().map(Value::from).map_err(Into::into)
                         }
+                        AllowedType::Boolean => {
+                            value.parse::<bool>().map(Value::from).map_err(Into::into)
+                        }
                         AllowedType::String => Ok(Value::String(value.to_string())),
                     };
 

crates/fuzzers/Cargo.toml

@@ -12,7 +12,9 @@ license.workspace = true
 
 [dependencies]
 arbitrary = { version = "1.3.2", features = ["derive"] }
+bumpalo = "3.16.0"
 clap = { version = "4.5.9", features = ["derive"] }
+either = "1.13.0"
 fastrand = "2.1.0"
 milli = { path = "../milli" }
 serde = { version = "1.0.204", features = ["derive"] }

crates/fuzzers/src/bin/fuzz-indexing.rs

@@ -4,11 +4,17 @@ use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
 use std::time::Duration;
 
 use arbitrary::{Arbitrary, Unstructured};
+use bumpalo::Bump;
 use clap::Parser;
+use either::Either;
 use fuzzers::Operation;
+use milli::documents::mmap_from_objects;
 use milli::heed::EnvOpenOptions;
-use milli::update::{IndexDocuments, IndexDocumentsConfig, IndexerConfig};
+use milli::update::new::indexer;
+use milli::update::{IndexDocumentsMethod, IndexerConfig};
+use milli::vector::EmbeddingConfigs;
 use milli::Index;
+use serde_json::Value;
 use tempfile::TempDir;
 
 #[derive(Debug, Arbitrary)]
@@ -58,7 +64,6 @@ fn main() {
             };
             let index = Index::new(options, tempdir.path()).unwrap();
             let indexer_config = IndexerConfig::default();
-            let index_documents_config = IndexDocumentsConfig::default();
 
             std::thread::scope(|s| {
                 loop {
@@ -75,38 +80,71 @@ fn main() {
 
                     let handle = s.spawn(|| {
                         let mut wtxn = index.write_txn().unwrap();
+                        let rtxn = index.read_txn().unwrap();
 
                         for batch in batches {
-                            let mut builder = IndexDocuments::new(
-                                &mut wtxn,
-                                &index,
-                                &indexer_config,
-                                index_documents_config.clone(),
-                                |_| (),
-                                || false,
-                            )
-                            .unwrap();
+                            let db_fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
+                            let mut new_fields_ids_map = db_fields_ids_map.clone();
 
+                            let indexer_alloc = Bump::new();
+                            let embedders = EmbeddingConfigs::default();
+                            let mut indexer = indexer::DocumentOperation::new(
+                                IndexDocumentsMethod::ReplaceDocuments,
+                            );
+
+                            let mut operations = Vec::new();
                             for op in batch.0 {
                                 match op {
                                     Operation::AddDoc(doc) => {
-                                        let documents =
-                                            milli::documents::objects_from_json_value(doc.to_d());
-                                        let documents =
-                                            milli::documents::documents_batch_reader_from_objects(
-                                                documents,
-                                            );
-                                        let (b, _added) = builder.add_documents(documents).unwrap();
-                                        builder = b;
+                                        let object = match doc.to_d() {
+                                            Value::Object(object) => object,
+                                            _ => unreachable!(),
+                                        };
+                                        let documents = mmap_from_objects(vec![object]);
+                                        operations.push(Either::Left(documents));
                                     }
                                     Operation::DeleteDoc(id) => {
-                                        let (b, _removed) =
-                                            builder.remove_documents(vec![id.to_s()]).unwrap();
-                                        builder = b;
+                                        let id = indexer_alloc.alloc_str(&id.to_s());
+                                        let ids = indexer_alloc.alloc_slice_copy(&[&*id]);
+                                        operations.push(Either::Right(ids));
                                     }
                                 }
                             }
-                            builder.execute().unwrap();
+
+                            for op in &operations {
+                                match op {
+                                    Either::Left(documents) => {
+                                        indexer.add_documents(documents).unwrap()
+                                    }
+                                    Either::Right(ids) => indexer.delete_documents(ids),
+                                }
+                            }
+
+                            let (document_changes, _operation_stats, primary_key) = indexer
+                                .into_changes(
+                                    &indexer_alloc,
+                                    &index,
+                                    &rtxn,
+                                    None,
+                                    &mut new_fields_ids_map,
+                                    &|| false,
+                                    &|_progress| (),
+                                )
+                                .unwrap();
+
+                            indexer::index(
+                                &mut wtxn,
+                                &index,
+                                indexer_config.grenad_parameters(),
+                                &db_fields_ids_map,
+                                new_fields_ids_map,
+                                primary_key,
+                                &document_changes,
+                                embedders,
+                                &|| false,
+                                &|_| (),
+                            )
+                            .unwrap();
 
                             // after executing a batch we check if the database is corrupted
                             let res = index.search(&wtxn).execute().unwrap();

crates/index-scheduler/Cargo.toml

@@ -22,6 +22,7 @@ flate2 = "1.0.30"
 meilisearch-auth = { path = "../meilisearch-auth" }
 meilisearch-types = { path = "../meilisearch-types" }
 page_size = "0.6.0"
+raw-collections = { git = "https://github.com/dureuill/raw-collections.git", version = "0.1.0" }
 rayon = "1.10.0"
 roaring = { version = "0.10.6", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
@@ -29,6 +30,7 @@ serde_json = { version = "1.0.120", features = ["preserve_order"] }
 synchronoise = "1.0.1"
 tempfile = "3.10.1"
 thiserror = "1.0.61"
+memmap2 = "0.9.4"
 time = { version = "0.3.36", features = [
     "serde-well-known",
     "formatting",
@@ -38,6 +40,7 @@ time = { version = "0.3.36", features = [
 tracing = "0.1.40"
 ureq = "2.10.0"
 uuid = { version = "1.10.0", features = ["serde", "v4"] }
+bumpalo = "3.16.0"
 
 [dev-dependencies]
 arroy = "0.5.0"

crates/index-scheduler/src/batch.rs

@@ -22,22 +22,27 @@ use std::ffi::OsStr;
 use std::fmt;
 use std::fs::{self, File};
 use std::io::BufWriter;
+use std::sync::atomic::{self, AtomicU64};
+use std::time::Duration;
 
+use bumpalo::collections::CollectIn;
+use bumpalo::Bump;
 use dump::IndexMetadata;
 use meilisearch_types::batches::BatchId;
 use meilisearch_types::error::Code;
 use meilisearch_types::heed::{RoTxn, RwTxn};
-use meilisearch_types::milli::documents::{obkv_to_object, DocumentsBatchReader};
+use meilisearch_types::milli::documents::{obkv_to_object, DocumentsBatchReader, PrimaryKey};
 use meilisearch_types::milli::heed::CompactionOption;
-use meilisearch_types::milli::update::{
-    IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Settings as MilliSettings,
-};
+use meilisearch_types::milli::update::new::indexer::{self, UpdateByFunction};
+use meilisearch_types::milli::update::{IndexDocumentsMethod, Settings as MilliSettings};
 use meilisearch_types::milli::vector::parsed_vectors::{
     ExplicitVectors, VectorOrArrayOfVectors, RESERVED_VECTORS_FIELD_NAME,
 };
-use meilisearch_types::milli::{self, Filter, Object};
+use meilisearch_types::milli::{self, Filter, ThreadPoolNoAbortBuilder};
 use meilisearch_types::settings::{apply_settings_to_builder, Settings, Unchecked};
-use meilisearch_types::tasks::{Details, IndexSwap, Kind, KindWithContent, Status, Task};
+use meilisearch_types::tasks::{
+    Details, IndexSwap, Kind, KindWithContent, Status, Task, TaskProgress,
+};
 use meilisearch_types::{compression, Index, VERSION_FILE_NAME};
 use roaring::RoaringBitmap;
 use time::macros::format_description;
@@ -46,7 +51,7 @@ use uuid::Uuid;
 
 use crate::autobatcher::{self, BatchKind};
 use crate::utils::{self, swap_index_uid_in_task, ProcessingBatch};
-use crate::{Error, IndexScheduler, MustStopProcessing, Result, TaskId};
+use crate::{Error, IndexScheduler, Result, TaskId};
 
 /// Represents a combination of tasks that can all be processed at the same time.
 ///
@@ -886,10 +891,8 @@ impl IndexScheduler {
                             while let Some(doc) =
                                 cursor.next_document().map_err(milli::Error::from)?
                             {
-                                dump_content_file.push_document(&obkv_to_object(
-                                    &doc,
-                                    &documents_batch_index,
-                                )?)?;
+                                dump_content_file
+                                    .push_document(&obkv_to_object(doc, &documents_batch_index)?)?;
                             }
                             dump_content_file.flush()?;
                         }
@@ -1232,6 +1235,44 @@ impl IndexScheduler {
         index: &'i Index,
         operation: IndexOperation,
     ) -> Result<Vec<Task>> {
+        let indexer_alloc = Bump::new();
+
+        let started_processing_at = std::time::Instant::now();
+        let secs_since_started_processing_at = AtomicU64::new(0);
+        const PRINT_SECS_DELTA: u64 = 1;
+
+        let processing_tasks = self.processing_tasks.clone();
+        let must_stop_processing = self.must_stop_processing.clone();
+        let send_progress = |progress| {
+            let now = std::time::Instant::now();
+            let elapsed = secs_since_started_processing_at.load(atomic::Ordering::Relaxed);
+            let previous = started_processing_at + Duration::from_secs(elapsed);
+            let elapsed = now - previous;
+
+            if elapsed.as_secs() < PRINT_SECS_DELTA {
+                return;
+            }
+
+            secs_since_started_processing_at
+                .store((now - started_processing_at).as_secs(), atomic::Ordering::Relaxed);
+
+            let TaskProgress {
+                current_step,
+                finished_steps,
+                total_steps,
+                finished_substeps,
+                total_substeps,
+            } = processing_tasks.write().unwrap().update_progress(progress);
+
+            tracing::info!(
+                current_step,
+                finished_steps,
+                total_steps,
+                finished_substeps,
+                total_substeps
+            );
+        };
+
         match operation {
             IndexOperation::DocumentClear { mut tasks, .. } => {
                 let count = milli::update::ClearDocuments::new(index_wtxn, index).execute()?;
@@ -1261,155 +1302,156 @@ impl IndexScheduler {
                 operations,
                 mut tasks,
             } => {
-                let started_processing_at = std::time::Instant::now();
-                let mut primary_key_has_been_set = false;
-                let must_stop_processing = self.must_stop_processing.clone();
-                let indexer_config = self.index_mapper.indexer_config();
-
-                if let Some(primary_key) = primary_key {
-                    match index.primary_key(index_wtxn)? {
-                        // if a primary key was set AND had already been defined in the index
-                        // but to a different value, we can make the whole batch fail.
-                        Some(pk) => {
-                            if primary_key != pk {
-                                return Err(milli::Error::from(
-                                    milli::UserError::PrimaryKeyCannotBeChanged(pk.to_string()),
-                                )
-                                .into());
-                            }
-                        }
-                        // if the primary key was set and there was no primary key set for this index
-                        // we set it to the received value before starting the indexing process.
-                        None => {
-                            let mut builder =
-                                milli::update::Settings::new(index_wtxn, index, indexer_config);
-                            builder.set_primary_key(primary_key);
-                            builder.execute(
-                                |indexing_step| tracing::debug!(update = ?indexing_step),
-                                || must_stop_processing.clone().get(),
-                            )?;
-                            primary_key_has_been_set = true;
+                // TODO: at some point, for better efficiency we might want to reuse the bumpalo for successive batches.
+                // this is made difficult by the fact we're doing private clones of the index scheduler and sending it
+                // to a fresh thread.
+                let mut content_files = Vec::new();
+                for operation in &operations {
+                    if let DocumentOperation::Add(content_uuid) = operation {
+                        let content_file = self.file_store.get_update(*content_uuid)?;
+                        let mmap = unsafe { memmap2::Mmap::map(&content_file)? };
+                        if !mmap.is_empty() {
+                            content_files.push(mmap);
                         }
                     }
                 }
 
-                let config = IndexDocumentsConfig { update_method: method, ..Default::default() };
+                let rtxn = index.read_txn()?;
+                let db_fields_ids_map = index.fields_ids_map(&rtxn)?;
+                let mut new_fields_ids_map = db_fields_ids_map.clone();
 
-                let embedder_configs = index.embedding_configs(index_wtxn)?;
-                // TODO: consider Arc'ing the map too (we only need read access + we'll be cloning it multiple times, so really makes sense)
-                let embedders = self.embedders(embedder_configs)?;
-
-                let mut builder = milli::update::IndexDocuments::new(
-                    index_wtxn,
-                    index,
-                    indexer_config,
-                    config,
-                    |indexing_step| tracing::trace!(?indexing_step, "Update"),
-                    || must_stop_processing.get(),
-                )?;
-
-                for (operation, task) in operations.into_iter().zip(tasks.iter_mut()) {
+                let mut content_files_iter = content_files.iter();
+                let mut indexer = indexer::DocumentOperation::new(method);
+                let embedders = index.embedding_configs(index_wtxn)?;
+                let embedders = self.embedders(embedders)?;
+                for operation in operations {
                     match operation {
-                        DocumentOperation::Add(content_uuid) => {
-                            let content_file = self.file_store.get_update(content_uuid)?;
-                            let reader = DocumentsBatchReader::from_reader(content_file)
-                                .map_err(milli::Error::from)?;
-                            let (new_builder, user_result) = builder.add_documents(reader)?;
-                            builder = new_builder;
-
-                            builder = builder.with_embedders(embedders.clone());
-
-                            let received_documents =
-                                if let Some(Details::DocumentAdditionOrUpdate {
-                                    received_documents,
-                                    ..
-                                }) = task.details
-                                {
-                                    received_documents
-                                } else {
-                                    // In the case of a `documentAdditionOrUpdate` the details MUST be set
-                                    unreachable!();
-                                };
-
-                            match user_result {
-                                Ok(count) => {
-                                    task.status = Status::Succeeded;
-                                    task.details = Some(Details::DocumentAdditionOrUpdate {
-                                        received_documents,
-                                        indexed_documents: Some(count),
-                                    })
-                                }
-                                Err(e) => {
-                                    task.status = Status::Failed;
-                                    task.details = Some(Details::DocumentAdditionOrUpdate {
-                                        received_documents,
-                                        indexed_documents: Some(0),
-                                    });
-                                    task.error = Some(milli::Error::from(e).into());
-                                }
-                            }
+                        DocumentOperation::Add(_content_uuid) => {
+                            let mmap = content_files_iter.next().unwrap();
+                            indexer.add_documents(mmap)?;
+                            // builder = builder.with_embedders(embedders.clone());
                         }
                         DocumentOperation::Delete(document_ids) => {
-                            let (new_builder, user_result) =
-                                builder.remove_documents(document_ids)?;
-                            builder = new_builder;
-                            // Uses Invariant: remove documents actually always returns Ok for the inner result
-                            let count = user_result.unwrap();
-                            let provided_ids =
-                                if let Some(Details::DocumentDeletion { provided_ids, .. }) =
-                                    task.details
-                                {
-                                    provided_ids
-                                } else {
-                                    // In the case of a `documentAdditionOrUpdate` the details MUST be set
-                                    unreachable!();
-                                };
-
-                            task.status = Status::Succeeded;
-                            task.details = Some(Details::DocumentDeletion {
-                                provided_ids,
-                                deleted_documents: Some(count),
-                            });
+                            let document_ids: bumpalo::collections::vec::Vec<_> = document_ids
+                                .iter()
+                                .map(|s| &*indexer_alloc.alloc_str(s))
+                                .collect_in(&indexer_alloc);
+                            indexer.delete_documents(document_ids.into_bump_slice());
                         }
                     }
                 }
 
-                if !tasks.iter().all(|res| res.error.is_some()) {
-                    let addition = builder.execute()?;
-                    tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
-                } else if primary_key_has_been_set {
-                    // Everything failed but we've set a primary key.
-                    // We need to remove it.
-                    let mut builder =
-                        milli::update::Settings::new(index_wtxn, index, indexer_config);
-                    builder.reset_primary_key();
-                    builder.execute(
-                        |indexing_step| tracing::trace!(update = ?indexing_step),
-                        || must_stop_processing.clone().get(),
-                    )?;
+                let local_pool;
+                let indexer_config = self.index_mapper.indexer_config();
+                let pool = match &indexer_config.thread_pool {
+                    Some(pool) => pool,
+                    None => {
+                        local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                        &local_pool
+                    }
+                };
+
+                let (document_changes, operation_stats, primary_key) = indexer.into_changes(
+                    &indexer_alloc,
+                    index,
+                    &rtxn,
+                    primary_key.as_deref(),
+                    &mut new_fields_ids_map,
+                    &|| must_stop_processing.get(),
+                    &send_progress,
+                )?;
+
+                let mut addition = 0;
+                for (stats, task) in operation_stats.into_iter().zip(&mut tasks) {
+                    addition += stats.document_count;
+                    match stats.error {
+                        Some(error) => {
+                            task.status = Status::Failed;
+                            task.error = Some(milli::Error::UserError(error).into());
+                        }
+                        None => task.status = Status::Succeeded,
+                    }
+
+                    task.details = match task.details {
+                        Some(Details::DocumentAdditionOrUpdate { received_documents, .. }) => {
+                            Some(Details::DocumentAdditionOrUpdate {
+                                received_documents,
+                                indexed_documents: Some(stats.document_count),
+                            })
+                        }
+                        Some(Details::DocumentDeletion { provided_ids, .. }) => {
+                            Some(Details::DocumentDeletion {
+                                provided_ids,
+                                deleted_documents: Some(stats.document_count),
+                            })
+                        }
+                        _ => {
+                            // In the case of a `documentAdditionOrUpdate` or `DocumentDeletion`
+                            // the details MUST be set to either addition or deletion
+                            unreachable!();
+                        }
+                    }
                 }
 
+                if tasks.iter().any(|res| res.error.is_none()) {
+                    pool.install(|| {
+                        indexer::index(
+                            index_wtxn,
+                            index,
+                            indexer_config.grenad_parameters(),
+                            &db_fields_ids_map,
+                            new_fields_ids_map,
+                            primary_key,
+                            &document_changes,
+                            embedders,
+                            &|| must_stop_processing.get(),
+                            &send_progress,
+                        )
+                    })
+                    .unwrap()?;
+
+                    tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
+                }
+                // else if primary_key_has_been_set {
+                //     // Everything failed but we've set a primary key.
+                //     // We need to remove it.
+                //     let mut builder =
+                //         milli::update::Settings::new(index_wtxn, index, indexer_config);
+                //     builder.reset_primary_key();
+                //     builder.execute(
+                //         |indexing_step| tracing::trace!(update = ?indexing_step),
+                //         || must_stop_processing.clone().get(),
+                //     )?;
+                // }
+
                 Ok(tasks)
             }
             IndexOperation::DocumentEdition { mut task, .. } => {
-                let (filter, context, function) =
-                    if let KindWithContent::DocumentEdition {
-                        filter_expr, context, function, ..
-                    } = &task.kind
-                    {
-                        (filter_expr, context, function)
-                    } else {
-                        unreachable!()
-                    };
-                let result_count = edit_documents_by_function(
-                    index_wtxn,
-                    filter,
-                    context.clone(),
+                let (filter, code) = if let KindWithContent::DocumentEdition {
+                    filter_expr,
+                    context: _,
                     function,
-                    self.index_mapper.indexer_config(),
-                    self.must_stop_processing.clone(),
-                    index,
-                );
+                    ..
+                } = &task.kind
+                {
+                    (filter_expr, function)
+                } else {
+                    unreachable!()
+                };
+
+                let candidates = match filter.as_ref().map(Filter::from_json) {
+                    Some(Ok(Some(filter))) => {
+                        filter.evaluate(index_wtxn, index).map_err(|err| match err {
+                            milli::Error::UserError(milli::UserError::InvalidFilter(_)) => {
+                                Error::from(err).with_custom_error_code(Code::InvalidDocumentFilter)
+                            }
+                            e => e.into(),
+                        })?
+                    }
+                    None | Some(Ok(None)) => index.documents_ids(index_wtxn)?,
+                    Some(Err(e)) => return Err(e.into()),
+                };
+
                 let (original_filter, context, function) = if let Some(Details::DocumentEdition {
                     original_filter,
                     context,
@@ -1423,6 +1465,68 @@ impl IndexScheduler {
                     unreachable!();
                 };
 
+                if candidates.is_empty() {
+                    task.status = Status::Succeeded;
+                    task.details = Some(Details::DocumentEdition {
+                        original_filter,
+                        context,
+                        function,
+                        deleted_documents: Some(0),
+                        edited_documents: Some(0),
+                    });
+
+                    return Ok(vec![task]);
+                }
+
+                let rtxn = index.read_txn()?;
+                let db_fields_ids_map = index.fields_ids_map(&rtxn)?;
+                let mut new_fields_ids_map = db_fields_ids_map.clone();
+                // candidates not empty => index not empty => a primary key is set
+                let primary_key = index.primary_key(&rtxn)?.unwrap();
+
+                let primary_key = PrimaryKey::new_or_insert(primary_key, &mut new_fields_ids_map)
+                    .map_err(milli::Error::from)?;
+
+                let result_count = Ok((candidates.len(), candidates.len())) as Result<_>;
+
+                if task.error.is_none() {
+                    let local_pool;
+                    let indexer_config = self.index_mapper.indexer_config();
+                    let pool = match &indexer_config.thread_pool {
+                        Some(pool) => pool,
+                        None => {
+                            local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                            &local_pool
+                        }
+                    };
+
+                    pool.install(|| {
+                        let indexer =
+                            UpdateByFunction::new(candidates, context.clone(), code.clone());
+                        let document_changes = indexer.into_changes(&primary_key)?;
+                        let embedders = index.embedding_configs(index_wtxn)?;
+                        let embedders = self.embedders(embedders)?;
+
+                        indexer::index(
+                            index_wtxn,
+                            index,
+                            indexer_config.grenad_parameters(),
+                            &db_fields_ids_map,
+                            new_fields_ids_map,
+                            None, // cannot change primary key in DocumentEdition
+                            &document_changes,
+                            embedders,
+                            &|| must_stop_processing.get(),
+                            &send_progress,
+                        )?;
+
+                        Result::Ok(())
+                    })
+                    .unwrap()?;
+
+                    // tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
+                }
+
                 match result_count {
                     Ok((deleted_documents, edited_documents)) => {
                         task.status = Status::Succeeded;
@@ -1523,26 +1627,55 @@ impl IndexScheduler {
                     }
                 }
 
-                let config = IndexDocumentsConfig {
-                    update_method: IndexDocumentsMethod::ReplaceDocuments,
-                    ..Default::default()
-                };
+                if to_delete.is_empty() {
+                    return Ok(tasks);
+                }
 
-                let must_stop_processing = self.must_stop_processing.clone();
-                let mut builder = milli::update::IndexDocuments::new(
-                    index_wtxn,
-                    index,
-                    self.index_mapper.indexer_config(),
-                    config,
-                    |indexing_step| tracing::debug!(update = ?indexing_step),
-                    || must_stop_processing.get(),
-                )?;
+                let rtxn = index.read_txn()?;
+                let db_fields_ids_map = index.fields_ids_map(&rtxn)?;
+                let mut new_fields_ids_map = db_fields_ids_map.clone();
 
-                let (new_builder, _count) =
-                    builder.remove_documents_from_db_no_batch(&to_delete)?;
-                builder = new_builder;
+                // to_delete not empty => index not empty => primary key set
+                let primary_key = index.primary_key(&rtxn)?.unwrap();
 
-                let _ = builder.execute()?;
+                let primary_key = PrimaryKey::new_or_insert(primary_key, &mut new_fields_ids_map)
+                    .map_err(milli::Error::from)?;
+
+                if !tasks.iter().all(|res| res.error.is_some()) {
+                    let local_pool;
+                    let indexer_config = self.index_mapper.indexer_config();
+                    let pool = match &indexer_config.thread_pool {
+                        Some(pool) => pool,
+                        None => {
+                            local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                            &local_pool
+                        }
+                    };
+
+                    let mut indexer = indexer::DocumentDeletion::new();
+                    indexer.delete_documents_by_docids(to_delete);
+                    let document_changes = indexer.into_changes(&indexer_alloc, primary_key);
+                    let embedders = index.embedding_configs(index_wtxn)?;
+                    let embedders = self.embedders(embedders)?;
+
+                    pool.install(|| {
+                        indexer::index(
+                            index_wtxn,
+                            index,
+                            indexer_config.grenad_parameters(),
+                            &db_fields_ids_map,
+                            new_fields_ids_map,
+                            None, // document deletion never changes primary key
+                            &document_changes,
+                            embedders,
+                            &|| must_stop_processing.get(),
+                            &send_progress,
+                        )
+                    })
+                    .unwrap()?;
+
+                    // tracing::info!(indexing_result = ?addition, processed_in = ?started_processing_at.elapsed(), "document indexing done");
+                }
 
                 Ok(tasks)
             }
@@ -1560,7 +1693,6 @@ impl IndexScheduler {
                     task.status = Status::Succeeded;
                 }
 
-                let must_stop_processing = self.must_stop_processing.clone();
                 builder.execute(
                     |indexing_step| tracing::debug!(update = ?indexing_step),
                     || must_stop_processing.get(),
@@ -1784,44 +1916,3 @@ impl IndexScheduler {
         Ok(tasks)
     }
 }
-
-fn edit_documents_by_function<'a>(
-    wtxn: &mut RwTxn<'a>,
-    filter: &Option<serde_json::Value>,
-    context: Option<Object>,
-    code: &str,
-    indexer_config: &IndexerConfig,
-    must_stop_processing: MustStopProcessing,
-    index: &'a Index,
-) -> Result<(u64, u64)> {
-    let candidates = match filter.as_ref().map(Filter::from_json) {
-        Some(Ok(Some(filter))) => filter.evaluate(wtxn, index).map_err(|err| match err {
-            milli::Error::UserError(milli::UserError::InvalidFilter(_)) => {
-                Error::from(err).with_custom_error_code(Code::InvalidDocumentFilter)
-            }
-            e => e.into(),
-        })?,
-        None | Some(Ok(None)) => index.documents_ids(wtxn)?,
-        Some(Err(e)) => return Err(e.into()),
-    };
-
-    let config = IndexDocumentsConfig {
-        update_method: IndexDocumentsMethod::ReplaceDocuments,
-        ..Default::default()
-    };
-
-    let mut builder = milli::update::IndexDocuments::new(
-        wtxn,
-        index,
-        indexer_config,
-        config,
-        |indexing_step| tracing::debug!(update = ?indexing_step),
-        || must_stop_processing.get(),
-    )?;
-
-    let (new_builder, count) = builder.edit_documents(&candidates, context, code)?;
-    builder = new_builder;
-
-    let _ = builder.execute()?;
-    Ok(count.unwrap())
-}

crates/index-scheduler/src/lib.rs

@@ -56,11 +56,12 @@ use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str, I128};
 use meilisearch_types::heed::{self, Database, Env, PutFlags, RoTxn, RwTxn};
 use meilisearch_types::milli::documents::DocumentsBatchBuilder;
 use meilisearch_types::milli::index::IndexEmbeddingConfig;
+use meilisearch_types::milli::update::new::indexer::document_changes::Progress;
 use meilisearch_types::milli::update::IndexerConfig;
 use meilisearch_types::milli::vector::{Embedder, EmbedderOptions, EmbeddingConfigs};
 use meilisearch_types::milli::{self, CboRoaringBitmapCodec, Index, RoaringBitmapCodec, BEU32};
 use meilisearch_types::task_view::TaskView;
-use meilisearch_types::tasks::{Kind, KindWithContent, Status, Task};
+use meilisearch_types::tasks::{Kind, KindWithContent, Status, Task, TaskProgress};
 use rayon::current_num_threads;
 use rayon::prelude::{IntoParallelIterator, ParallelIterator};
 use roaring::RoaringBitmap;
@@ -167,12 +168,14 @@ pub struct ProcessingTasks {
     batch: Option<ProcessingBatch>,
     /// The list of tasks ids that are currently running.
     processing: RoaringBitmap,
+    /// The progress on processing tasks
+    progress: Option<TaskProgress>,
 }
 
 impl ProcessingTasks {
     /// Creates an empty `ProcessingAt` struct.
     fn new() -> ProcessingTasks {
-        ProcessingTasks { batch: None, processing: RoaringBitmap::new() }
+        ProcessingTasks { batch: None, processing: RoaringBitmap::new(), progress: None }
     }
 
     /// Stores the currently processing tasks, and the date time at which it started.
@@ -181,11 +184,18 @@ impl ProcessingTasks {
         self.processing = processing;
     }
 
+    fn update_progress(&mut self, progress: Progress) -> TaskProgress {
+        self.progress.get_or_insert_with(TaskProgress::default).update(progress)
+    }
+
     /// Set the processing tasks to an empty list
     fn stop_processing(&mut self) -> Self {
+        self.progress = None;
+
         Self {
             batch: std::mem::take(&mut self.batch),
             processing: std::mem::take(&mut self.processing),
+            progress: None,
         }
     }
 
@@ -768,7 +778,7 @@ impl IndexScheduler {
 
     /// Return the task ids matched by the given query from the index scheduler's point of view.
     pub(crate) fn get_task_ids(&self, rtxn: &RoTxn, query: &Query) -> Result<RoaringBitmap> {
-        let ProcessingTasks { batch: processing_batch, processing: processing_tasks } =
+        let ProcessingTasks { batch: processing_batch, processing: processing_tasks, progress: _ } =
             self.processing_tasks.read().unwrap().clone();
         let Query {
             limit,
@@ -1352,9 +1362,11 @@ impl IndexScheduler {
         let tasks =
             self.get_existing_tasks(&rtxn, tasks.take(query.limit.unwrap_or(u32::MAX) as usize))?;
 
-        let ProcessingTasks { batch, processing } =
+        let ProcessingTasks { batch, processing, progress } =
             self.processing_tasks.read().map_err(|_| Error::CorruptedTaskQueue)?.clone();
 
+        let _ = progress;
+
         let ret = tasks.into_iter();
         if processing.is_empty() || batch.is_none() {
             Ok((ret.collect(), total))
@@ -2290,7 +2302,7 @@ mod tests {
                 dumps_path: tempdir.path().join("dumps"),
                 webhook_url: None,
                 webhook_authorization_header: None,
-                task_db_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
+                task_db_size: 1000 * 1000 * 10, // 10 MB, we don't use MiB on purpose.
                 index_base_map_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
                 enable_mdb_writemap: false,
                 index_growth_amount: 1000 * 1000 * 1000 * 1000, // 1 TB
@@ -5187,11 +5199,11 @@ mod tests {
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "only_first_task_succeed");
 
         // The second batch should fail.
-        handle.advance_one_failed_batch();
+        handle.advance_one_successful_batch();
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "second_task_fails");
 
         // The second batch should fail.
-        handle.advance_one_failed_batch();
+        handle.advance_one_successful_batch();
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "third_task_fails");
 
         // Is the primary key still what we expect?
@@ -5252,7 +5264,7 @@ mod tests {
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "only_first_task_succeed");
 
         // The second batch should fail and contains two tasks.
-        handle.advance_one_failed_batch();
+        handle.advance_one_successful_batch();
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "second_and_third_tasks_fails");
 
         // Is the primary key still what we expect?
@@ -5331,7 +5343,8 @@ mod tests {
         snapshot!(primary_key, @"id");
 
         // We're trying to `bork` again, but now there is already a primary key set for this index.
-        handle.advance_one_failed_batch();
+        // NOTE: it's marked as successful because the batch didn't fails, it's the individual tasks that failed.
+        handle.advance_one_successful_batch();
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "fourth_task_fails");
 
         // Finally the last task should succeed since its primary key is the same as the valid one.
@@ -5491,7 +5504,7 @@ mod tests {
         snapshot!(primary_key.is_none(), @"false");
 
         // The second batch should contains only one task that fails because it tries to update the primary key to `bork`.
-        handle.advance_one_failed_batch();
+        handle.advance_one_successful_batch();
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "second_task_fails");
 
         // The third batch should succeed and only contains one task.
@@ -6136,9 +6149,10 @@ mod tests {
 
             let configs = index_scheduler.embedders(configs).unwrap();
             let (hf_embedder, _, _) = configs.get(&simple_hf_name).unwrap();
-            let beagle_embed = hf_embedder.embed_one(S("Intel the beagle best doggo")).unwrap();
-            let lab_embed = hf_embedder.embed_one(S("Max the lab best doggo")).unwrap();
-            let patou_embed = hf_embedder.embed_one(S("kefir the patou best doggo")).unwrap();
+            let beagle_embed =
+                hf_embedder.embed_one(S("Intel the beagle best doggo"), None).unwrap();
+            let lab_embed = hf_embedder.embed_one(S("Max the lab best doggo"), None).unwrap();
+            let patou_embed = hf_embedder.embed_one(S("kefir the patou best doggo"), None).unwrap();
             (fakerest_name, simple_hf_name, beagle_embed, lab_embed, patou_embed)
         };
 

crates/index-scheduler/src/snapshots/lib.rs/import_vectors/Intel to kefir succeeds.snap

@@ -1,5 +1,5 @@
 ---
-source: crates/index-scheduler/src/lib.rs
+source: crates/crates/index-scheduler/src/lib.rs
 snapshot_kind: text
 ---
 ### Autobatching Enabled = true
@@ -23,7 +23,7 @@ succeeded [0,1,2,]
 doggos [0,1,2,]
 ----------------------------------------------------------------------
 ### Index Mapper:
-doggos: { number_of_documents: 1, field_distribution: {"_vectors": 1, "breed": 1, "doggo": 1, "id": 1} }
+doggos: { number_of_documents: 1, field_distribution: {"breed": 1, "doggo": 1, "id": 1} }
 
 ----------------------------------------------------------------------
 ### Canceled By:

crates/index-scheduler/src/snapshots/lib.rs/import_vectors/Intel to kefir.snap

@@ -1,5 +1,5 @@
 ---
-source: crates/index-scheduler/src/lib.rs
+source: crates/crates/index-scheduler/src/lib.rs
 snapshot_kind: text
 ---
 ### Autobatching Enabled = true
@@ -23,7 +23,7 @@ succeeded [0,1,]
 doggos [0,1,2,]
 ----------------------------------------------------------------------
 ### Index Mapper:
-doggos: { number_of_documents: 1, field_distribution: {"_vectors": 1, "breed": 1, "doggo": 1, "id": 1} }
+doggos: { number_of_documents: 1, field_distribution: {"breed": 1, "doggo": 1, "id": 1} }
 
 ----------------------------------------------------------------------
 ### Canceled By:

crates/index-scheduler/src/snapshots/lib.rs/import_vectors/adding Intel succeeds.snap

@@ -1,5 +1,5 @@
 ---
-source: crates/index-scheduler/src/lib.rs
+source: crates/crates/index-scheduler/src/lib.rs
 snapshot_kind: text
 ---
 ### Autobatching Enabled = true
@@ -22,7 +22,7 @@ succeeded [0,1,]
 doggos [0,1,]
 ----------------------------------------------------------------------
 ### Index Mapper:
-doggos: { number_of_documents: 1, field_distribution: {"_vectors": 1, "breed": 1, "doggo": 1, "id": 1} }
+doggos: { number_of_documents: 1, field_distribution: {"breed": 1, "doggo": 1, "id": 1} }
 
 ----------------------------------------------------------------------
 ### Canceled By:

crates/meili-snap/Cargo.toml

@@ -11,6 +11,6 @@ edition.workspace = true
 license.workspace = true
 
 [dependencies]
-insta = { version = "^1.39.0", features = ["json", "redactions"] }
+insta = { version = "=1.39.0", features = ["json", "redactions"] }
 md5 = "0.7.0"
 once_cell = "1.19"

crates/meilisearch-types/Cargo.toml

@@ -13,6 +13,7 @@ license.workspace = true
 [dependencies]
 actix-web = { version = "4.8.0", default-features = false }
 anyhow = "1.0.86"
+bumpalo = "3.16.0"
 convert_case = "0.6.0"
 csv = "1.3.0"
 deserr = { version = "0.6.2", features = ["actix-web"] }
@@ -23,6 +24,7 @@ flate2 = "1.0.30"
 fst = "0.4.7"
 memmap2 = "0.9.4"
 milli = { path = "../milli" }
+raw-collections = { git = "https://github.com/dureuill/raw-collections.git", version = "0.1.0" }
 roaring = { version = "0.10.6", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
 serde-cs = "0.2.4"
@@ -70,4 +72,3 @@ swedish-recomposition = ["milli/swedish-recomposition"]
 german = ["milli/german"]
 # allow turkish normalization
 turkish = ["milli/turkish"]
-

crates/meilisearch-types/src/document_formats.rs

@@ -1,20 +1,25 @@
 use std::fmt::{self, Debug, Display};
 use std::fs::File;
-use std::io::{self, BufWriter, Write};
+use std::io::{self, BufWriter};
 use std::marker::PhantomData;
 
-use memmap2::MmapOptions;
-use milli::documents::{DocumentsBatchBuilder, Error};
+use bumpalo::Bump;
+use memmap2::Mmap;
+use milli::documents::Error;
+use milli::update::new::TopLevelMap;
 use milli::Object;
+use raw_collections::RawMap;
 use serde::de::{SeqAccess, Visitor};
 use serde::{Deserialize, Deserializer};
 use serde_json::error::Category;
+use serde_json::value::RawValue;
+use serde_json::{to_writer, Map, Value};
 
 use crate::error::{Code, ErrorCode};
 
 type Result<T> = std::result::Result<T, DocumentFormatError>;
 
-#[derive(Debug)]
+#[derive(Debug, Clone, Copy)]
 pub enum PayloadType {
     Ndjson,
     Json,
@@ -88,6 +93,26 @@ impl From<(PayloadType, Error)> for DocumentFormatError {
     }
 }
 
+impl From<(PayloadType, serde_json::Error)> for DocumentFormatError {
+    fn from((ty, error): (PayloadType, serde_json::Error)) -> Self {
+        if error.classify() == Category::Data {
+            Self::Io(error.into())
+        } else {
+            Self::MalformedPayload(Error::Json(error), ty)
+        }
+    }
+}
+
+impl From<(PayloadType, csv::Error)> for DocumentFormatError {
+    fn from((ty, error): (PayloadType, csv::Error)) -> Self {
+        if error.is_io_error() {
+            Self::Io(error.into())
+        } else {
+            Self::MalformedPayload(Error::Csv(error), ty)
+        }
+    }
+}
+
 impl From<io::Error> for DocumentFormatError {
     fn from(error: io::Error) -> Self {
         Self::Io(error)
@@ -103,67 +128,146 @@ impl ErrorCode for DocumentFormatError {
     }
 }
 
-/// Reads CSV from input and write an obkv batch to writer.
-pub fn read_csv(file: &File, writer: impl Write, delimiter: u8) -> Result<u64> {
-    let mut builder = DocumentsBatchBuilder::new(BufWriter::new(writer));
-    let mmap = unsafe { MmapOptions::new().map(file)? };
-    let csv = csv::ReaderBuilder::new().delimiter(delimiter).from_reader(mmap.as_ref());
-    builder.append_csv(csv).map_err(|e| (PayloadType::Csv { delimiter }, e))?;
-
-    let count = builder.documents_count();
-    let _ = builder.into_inner().map_err(DocumentFormatError::Io)?;
-
-    Ok(count as u64)
+// TODO remove that from the place I've borrowed it
+#[derive(Debug)]
+enum AllowedType {
+    String,
+    Boolean,
+    Number,
 }
 
-/// Reads JSON from temporary file and write an obkv batch to writer.
-pub fn read_json(file: &File, writer: impl Write) -> Result<u64> {
-    let mut builder = DocumentsBatchBuilder::new(BufWriter::new(writer));
-    let mmap = unsafe { MmapOptions::new().map(file)? };
-    let mut deserializer = serde_json::Deserializer::from_slice(&mmap);
+fn parse_csv_header(header: &str) -> (&str, AllowedType) {
+    // if there are several separators we only split on the last one.
+    match header.rsplit_once(':') {
+        Some((field_name, field_type)) => match field_type {
+            "string" => (field_name, AllowedType::String),
+            "boolean" => (field_name, AllowedType::Boolean),
+            "number" => (field_name, AllowedType::Number),
+            // if the pattern isn't recognized, we keep the whole field.
+            _otherwise => (header, AllowedType::String),
+        },
+        None => (header, AllowedType::String),
+    }
+}
 
-    match array_each(&mut deserializer, |obj| builder.append_json_object(&obj)) {
+/// Reads CSV from file and write it in NDJSON in a file checking it along the way.
+pub fn read_csv(input: &File, output: impl io::Write, delimiter: u8) -> Result<u64> {
+    let ptype = PayloadType::Csv { delimiter };
+    let mut output = BufWriter::new(output);
+    let mut reader = csv::ReaderBuilder::new().delimiter(delimiter).from_reader(input);
+
+    let headers = reader.headers().map_err(|e| DocumentFormatError::from((ptype, e)))?.clone();
+    let typed_fields: Vec<_> = headers.iter().map(parse_csv_header).collect();
+    let mut object: Map<_, _> =
+        typed_fields.iter().map(|(k, _)| (k.to_string(), Value::Null)).collect();
+
+    let mut line = 0;
+    let mut record = csv::StringRecord::new();
+    while reader.read_record(&mut record).map_err(|e| DocumentFormatError::from((ptype, e)))? {
+        // We increment here and not at the end of the loop
+        // to take the header offset into account.
+        line += 1;
+
+        // Reset the document values
+        object.iter_mut().for_each(|(_, v)| *v = Value::Null);
+
+        for (i, (name, atype)) in typed_fields.iter().enumerate() {
+            let value = &record[i];
+            let trimmed_value = value.trim();
+            let value = match atype {
+                AllowedType::Number if trimmed_value.is_empty() => Value::Null,
+                AllowedType::Number => match trimmed_value.parse::<i64>() {
+                    Ok(integer) => Value::from(integer),
+                    Err(_) => match trimmed_value.parse::<f64>() {
+                        Ok(float) => Value::from(float),
+                        Err(error) => {
+                            return Err(DocumentFormatError::MalformedPayload(
+                                Error::ParseFloat { error, line, value: value.to_string() },
+                                ptype,
+                            ))
+                        }
+                    },
+                },
+                AllowedType::Boolean if trimmed_value.is_empty() => Value::Null,
+                AllowedType::Boolean => match trimmed_value.parse::<bool>() {
+                    Ok(bool) => Value::from(bool),
+                    Err(error) => {
+                        return Err(DocumentFormatError::MalformedPayload(
+                            Error::ParseBool { error, line, value: value.to_string() },
+                            ptype,
+                        ))
+                    }
+                },
+                AllowedType::String if value.is_empty() => Value::Null,
+                AllowedType::String => Value::from(value),
+            };
+
+            *object.get_mut(*name).expect("encountered an unknown field") = value;
+        }
+
+        to_writer(&mut output, &object).map_err(|e| DocumentFormatError::from((ptype, e)))?;
+    }
+
+    Ok(line as u64)
+}
+
+/// Reads JSON from file and write it in NDJSON in a file checking it along the way.
+pub fn read_json(input: &File, output: impl io::Write) -> Result<u64> {
+    // We memory map to be able to deserailize into a TopLevelMap<'pl> that
+    // does not allocate when possible and only materialize the first/top level.
+    let input = unsafe { Mmap::map(input).map_err(DocumentFormatError::Io)? };
+    let mut doc_alloc = Bump::with_capacity(1024 * 1024 * 1024); // 1MiB
+
+    let mut out = BufWriter::new(output);
+    let mut deserializer = serde_json::Deserializer::from_slice(&input);
+    let res = array_each(&mut deserializer, |obj: &RawValue| {
+        doc_alloc.reset();
+        let map = RawMap::from_raw_value(obj, &doc_alloc)?;
+        to_writer(&mut out, &map)
+    });
+    let count = match res {
         // The json data has been deserialized and does not need to be processed again.
         // The data has been transferred to the writer during the deserialization process.
-        Ok(Ok(_)) => (),
-        Ok(Err(e)) => return Err(DocumentFormatError::Io(e)),
+        Ok(Ok(count)) => count,
+        Ok(Err(e)) => return Err(DocumentFormatError::from((PayloadType::Json, e))),
         Err(e) => {
             // Attempt to deserialize a single json string when the cause of the exception is not Category.data
             // Other types of deserialisation exceptions are returned directly to the front-end
-            if e.classify() != serde_json::error::Category::Data {
-                return Err(DocumentFormatError::MalformedPayload(
-                    Error::Json(e),
-                    PayloadType::Json,
-                ));
+            if e.classify() != Category::Data {
+                return Err(DocumentFormatError::from((PayloadType::Json, e)));
             }
 
-            let content: Object = serde_json::from_slice(&mmap)
+            let content: Object = serde_json::from_slice(&input)
                 .map_err(Error::Json)
                 .map_err(|e| (PayloadType::Json, e))?;
-            builder.append_json_object(&content).map_err(DocumentFormatError::Io)?;
+            to_writer(&mut out, &content)
+                .map(|_| 1)
+                .map_err(|e| DocumentFormatError::from((PayloadType::Json, e)))?
         }
+    };
+
+    match out.into_inner() {
+        Ok(_) => Ok(count),
+        Err(ie) => Err(DocumentFormatError::Io(ie.into_error())),
     }
-
-    let count = builder.documents_count();
-    let _ = builder.into_inner().map_err(DocumentFormatError::Io)?;
-
-    Ok(count as u64)
 }
 
-/// Reads JSON from temporary file  and write an obkv batch to writer.
-pub fn read_ndjson(file: &File, writer: impl Write) -> Result<u64> {
-    let mut builder = DocumentsBatchBuilder::new(BufWriter::new(writer));
-    let mmap = unsafe { MmapOptions::new().map(file)? };
+/// Reads NDJSON from file and write it in NDJSON in a file checking it along the way.
+pub fn read_ndjson(input: &File, output: impl io::Write) -> Result<u64> {
+    // We memory map to be able to deserailize into a TopLevelMap<'pl> that
+    // does not allocate when possible and only materialize the first/top level.
+    let input = unsafe { Mmap::map(input).map_err(DocumentFormatError::Io)? };
+    let mut output = BufWriter::new(output);
 
-    for result in serde_json::Deserializer::from_slice(&mmap).into_iter() {
-        let object = result.map_err(Error::Json).map_err(|e| (PayloadType::Ndjson, e))?;
-        builder.append_json_object(&object).map_err(Into::into).map_err(DocumentFormatError::Io)?;
+    let mut count = 0;
+    for result in serde_json::Deserializer::from_slice(&input).into_iter() {
+        count += 1;
+        result
+            .and_then(|map: TopLevelMap| to_writer(&mut output, &map))
+            .map_err(|e| DocumentFormatError::from((PayloadType::Ndjson, e)))?;
     }
 
-    let count = builder.documents_count();
-    let _ = builder.into_inner().map_err(Into::into).map_err(DocumentFormatError::Io)?;
-
-    Ok(count as u64)
+    Ok(count)
 }
 
 /// The actual handling of the deserialization process in serde
@@ -172,20 +276,23 @@ pub fn read_ndjson(file: &File, writer: impl Write) -> Result<u64> {
 /// ## References
 /// <https://serde.rs/stream-array.html>
 /// <https://github.com/serde-rs/json/issues/160>
-fn array_each<'de, D, T, F>(deserializer: D, f: F) -> std::result::Result<io::Result<u64>, D::Error>
+fn array_each<'de, D, T, F>(
+    deserializer: D,
+    f: F,
+) -> std::result::Result<serde_json::Result<u64>, D::Error>
 where
     D: Deserializer<'de>,
     T: Deserialize<'de>,
-    F: FnMut(T) -> io::Result<()>,
+    F: FnMut(T) -> serde_json::Result<()>,
 {
     struct SeqVisitor<T, F>(F, PhantomData<T>);
 
     impl<'de, T, F> Visitor<'de> for SeqVisitor<T, F>
     where
         T: Deserialize<'de>,
-        F: FnMut(T) -> io::Result<()>,
+        F: FnMut(T) -> serde_json::Result<()>,
     {
-        type Value = io::Result<u64>;
+        type Value = serde_json::Result<u64>;
 
         fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
             formatter.write_str("a nonempty sequence")
@@ -194,7 +301,7 @@ where
         fn visit_seq<A>(
             mut self,
             mut seq: A,
-        ) -> std::result::Result<io::Result<u64>, <A as SeqAccess<'de>>::Error>
+        ) -> std::result::Result<serde_json::Result<u64>, <A as SeqAccess<'de>>::Error>
         where
             A: SeqAccess<'de>,
         {
@@ -203,7 +310,7 @@ where
                 match self.0(value) {
                     Ok(()) => max += 1,
                     Err(e) => return Ok(Err(e)),
-                };
+                }
             }
             Ok(Ok(max))
         }

crates/meilisearch-types/src/tasks.rs

@@ -4,6 +4,7 @@ use std::fmt::{Display, Write};
 use std::str::FromStr;
 
 use enum_iterator::Sequence;
+use milli::update::new::indexer::document_changes::Progress;
 use milli::update::IndexDocumentsMethod;
 use milli::Object;
 use roaring::RoaringBitmap;
@@ -40,6 +41,62 @@ pub struct Task {
     pub kind: KindWithContent,
 }
 
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct TaskProgress {
+    pub current_step: &'static str,
+    pub finished_steps: u16,
+    pub total_steps: u16,
+    pub finished_substeps: Option<u32>,
+    pub total_substeps: Option<u32>,
+}
+
+impl Default for TaskProgress {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl TaskProgress {
+    pub fn new() -> Self {
+        Self {
+            current_step: "start",
+            finished_steps: 0,
+            total_steps: 1,
+            finished_substeps: None,
+            total_substeps: None,
+        }
+    }
+
+    pub fn update(&mut self, progress: Progress) -> TaskProgress {
+        if self.finished_steps > progress.finished_steps {
+            return *self;
+        }
+
+        if self.current_step != progress.step_name {
+            self.current_step = progress.step_name
+        }
+
+        self.total_steps = progress.total_steps;
+
+        if self.finished_steps < progress.finished_steps {
+            self.finished_substeps = None;
+            self.total_substeps = None;
+        }
+        self.finished_steps = progress.finished_steps;
+        if let Some((finished_substeps, total_substeps)) = progress.finished_total_substep {
+            if let Some(task_finished_substeps) = self.finished_substeps {
+                if task_finished_substeps > finished_substeps {
+                    return *self;
+                }
+            }
+            self.finished_substeps = Some(finished_substeps);
+            self.total_substeps = Some(total_substeps);
+        }
+        *self
+    }
+}
+
 impl Task {
     pub fn index_uid(&self) -> Option<&str> {
         use KindWithContent::*;

crates/meilisearch/Cargo.toml

@@ -57,7 +57,7 @@ meilisearch-types = { path = "../meilisearch-types" }
 mimalloc = { version = "0.1.43", default-features = false }
 mime = "0.3.17"
 num_cpus = "1.16.0"
-obkv = "0.2.2"
+obkv = { git = "https://github.com/kerollmops/obkv", branch = "unsized-kvreader" }
 once_cell = "1.19.0"
 ordered-float = "4.2.1"
 parking_lot = "0.12.3"

crates/meilisearch/src/search/mod.rs

@@ -796,8 +796,10 @@ fn prepare_search<'t>(
                     let span = tracing::trace_span!(target: "search::vector", "embed_one");
                     let _entered = span.enter();
 
+                    let deadline = std::time::Instant::now() + std::time::Duration::from_secs(10);
+
                     embedder
-                        .embed_one(query.q.clone().unwrap())
+                        .embed_one(query.q.clone().unwrap(), Some(deadline))
                         .map_err(milli::vector::Error::from)
                         .map_err(milli::Error::from)?
                 }
@@ -1687,7 +1689,7 @@ fn add_non_formatted_ids_to_formatted_options(
 fn make_document(
     displayed_attributes: &BTreeSet<FieldId>,
     field_ids_map: &FieldsIdsMap,
-    obkv: obkv::KvReaderU16,
+    obkv: &obkv::KvReaderU16,
 ) -> Result<Document, MeilisearchHttpError> {
     let mut document = serde_json::Map::new();
 

crates/meilisearch/tests/documents/add_documents.rs

@@ -1344,7 +1344,6 @@ async fn error_add_documents_missing_document_id() {
 }
 
 #[actix_rt::test]
-#[should_panic]
 async fn error_document_field_limit_reached_in_one_document() {
     let server = Server::new().await;
     let index = server.index("test");
@@ -1361,7 +1360,7 @@ async fn error_document_field_limit_reached_in_one_document() {
     let documents = json!([big_object]);
 
     let (response, code) = index.update_documents(documents, Some("id")).await;
-    snapshot!(code, @"500 Internal Server Error");
+    snapshot!(code, @"202 Accepted");
 
     let response = index.wait_task(response.uid()).await;
     snapshot!(code, @"202 Accepted");
@@ -1369,16 +1368,22 @@ async fn error_document_field_limit_reached_in_one_document() {
     snapshot!(response,
         @r###"
     {
-      "uid": 1,
+      "uid": "[uid]",
+      "batchUid": "[batch_uid]",
       "indexUid": "test",
-      "status": "succeeded",
+      "status": "failed",
       "type": "documentAdditionOrUpdate",
       "canceledBy": null,
       "details": {
         "receivedDocuments": 1,
-        "indexedDocuments": 1
+        "indexedDocuments": 0
+      },
+      "error": {
+        "message": "A document cannot contain more than 65,535 fields.",
+        "code": "max_fields_limit_exceeded",
+        "type": "invalid_request",
+        "link": "https://docs.meilisearch.com/errors#max_fields_limit_exceeded"
       },
-      "error": null,
       "duration": "[duration]",
       "enqueuedAt": "[date]",
       "startedAt": "[date]",
@@ -1676,7 +1681,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "The `_geo` field in the document with the id: `11` is not an object. Was expecting an object with the `_geo.lat` and `_geo.lng` fields but instead got `\"foobar\"`.",
+        "message": "The `_geo` field in the document with the id: `\"11\"` is not an object. Was expecting an object with the `_geo.lat` and `_geo.lng` fields but instead got `\"foobar\"`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1714,7 +1719,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find latitude nor longitude in the document with the id: `11`. Was expecting `_geo.lat` and `_geo.lng` fields.",
+        "message": "Could not find latitude nor longitude in the document with the id: `\"11\"`. Was expecting `_geo.lat` and `_geo.lng` fields.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1752,7 +1757,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find latitude nor longitude in the document with the id: `11`. Was expecting `_geo.lat` and `_geo.lng` fields.",
+        "message": "Could not find latitude nor longitude in the document with the id: `\"11\"`. Was expecting `_geo.lat` and `_geo.lng` fields.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1790,7 +1795,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find longitude in the document with the id: `11`. Was expecting a `_geo.lng` field.",
+        "message": "Could not find longitude in the document with the id: `\"11\"`. Was expecting a `_geo.lng` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1828,7 +1833,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find latitude in the document with the id: `11`. Was expecting a `_geo.lat` field.",
+        "message": "Could not find latitude in the document with the id: `\"11\"`. Was expecting a `_geo.lat` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1866,7 +1871,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find longitude in the document with the id: `11`. Was expecting a `_geo.lng` field.",
+        "message": "Could not find longitude in the document with the id: `\"11\"`. Was expecting a `_geo.lng` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1904,7 +1909,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find latitude in the document with the id: `11`. Was expecting a `_geo.lat` field.",
+        "message": "Could not find latitude in the document with the id: `\"11\"`. Was expecting a `_geo.lat` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1942,7 +1947,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not parse latitude nor longitude in the document with the id: `11`. Was expecting finite numbers but instead got `false` and `true`.",
+        "message": "Could not parse latitude nor longitude in the document with the id: `\"11\"`. Was expecting finite numbers but instead got `false` and `true`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -1980,7 +1985,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find longitude in the document with the id: `11`. Was expecting a `_geo.lng` field.",
+        "message": "Could not find longitude in the document with the id: `\"11\"`. Was expecting a `_geo.lng` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2018,7 +2023,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not find latitude in the document with the id: `11`. Was expecting a `_geo.lat` field.",
+        "message": "Could not find latitude in the document with the id: `\"11\"`. Was expecting a `_geo.lat` field.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2056,7 +2061,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not parse latitude nor longitude in the document with the id: `11`. Was expecting finite numbers but instead got `\"doggo\"` and `\"doggo\"`.",
+        "message": "Could not parse latitude nor longitude in the document with the id: `\"11\"`. Was expecting finite numbers but instead got `\"doggo\"` and `\"doggo\"`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2094,7 +2099,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "The `_geo` field in the document with the id: `11` contains the following unexpected fields: `{\"doggo\":\"are the best\"}`.",
+        "message": "The `_geo` field in the document with the id: `\"11\"` contains the following unexpected fields: `{\"doggo\":\"are the best\"}`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2133,7 +2138,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not parse longitude in the document with the id: `12`. Was expecting a finite number but instead got `null`.",
+        "message": "Could not parse longitude in the document with the id: `\"12\"`. Was expecting a finite number but instead got `null`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2170,7 +2175,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not parse latitude in the document with the id: `12`. Was expecting a finite number but instead got `null`.",
+        "message": "Could not parse latitude in the document with the id: `\"12\"`. Was expecting a finite number but instead got `null`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2207,7 +2212,7 @@ async fn add_documents_invalid_geo_field() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Could not parse latitude nor longitude in the document with the id: `13`. Was expecting finite numbers but instead got `null` and `null`.",
+        "message": "Could not parse latitude nor longitude in the document with the id: `\"13\"`. Was expecting finite numbers but instead got `null` and `null`.",
         "code": "invalid_document_geo_field",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_document_geo_field"
@@ -2227,7 +2232,7 @@ async fn add_invalid_geo_and_then_settings() {
     let index = server.index("test");
     index.create(Some("id")).await;
 
-    // _geo is not an object
+    // _geo is not a correct object
     let documents = json!([
         {
             "id": "11",
@@ -2257,7 +2262,7 @@ async fn add_invalid_geo_and_then_settings() {
     }
     "###);
 
-    let (ret, code) = index.update_settings(json!({"sortableAttributes": ["_geo"]})).await;
+    let (ret, code) = index.update_settings(json!({ "sortableAttributes": ["_geo"] })).await;
     snapshot!(code, @"202 Accepted");
     let ret = index.wait_task(ret.uid()).await;
     snapshot!(ret, @r###"

crates/meilisearch/tests/search/geo.rs

@@ -70,8 +70,8 @@ async fn geo_bounding_box_with_string_and_number() {
     let documents = DOCUMENTS.clone();
     index.update_settings_filterable_attributes(json!(["_geo"])).await;
     index.update_settings_sortable_attributes(json!(["_geo"])).await;
-    index.add_documents(documents, None).await;
-    index.wait_task(2).await;
+    let (ret, _code) = index.add_documents(documents, None).await;
+    index.wait_task(ret.uid()).await.succeeded();
 
     index
         .search(

crates/meilisearch/tests/search/mod.rs

@@ -750,9 +750,9 @@ async fn test_score_details() {
                     ],
                     "_vectors": {
                       "manual": [
-                        -100.0,
-                        231.0,
-                        32.0
+                        -100,
+                        231,
+                        32
                       ]
                     },
                     "_rankingScoreDetails": {
@@ -1543,9 +1543,9 @@ async fn simple_search_with_strange_synonyms() {
                 ],
                 "_vectors": {
                   "manual": [
-                    -100.0,
-                    231.0,
-                    32.0
+                    -100,
+                    231,
+                    32
                   ]
                 }
               }
@@ -1568,9 +1568,9 @@ async fn simple_search_with_strange_synonyms() {
                 ],
                 "_vectors": {
                   "manual": [
-                    -100.0,
-                    231.0,
-                    32.0
+                    -100,
+                    231,
+                    32
                   ]
                 }
               }
@@ -1593,9 +1593,9 @@ async fn simple_search_with_strange_synonyms() {
                 ],
                 "_vectors": {
                   "manual": [
-                    -100.0,
-                    231.0,
-                    32.0
+                    -100,
+                    231,
+                    32
                   ]
                 }
               }
@@ -1604,3 +1604,79 @@ async fn simple_search_with_strange_synonyms() {
         })
         .await;
 }
+
+#[actix_rt::test]
+async fn change_attributes_settings() {
+    let server = Server::new().await;
+    let index = server.index("test");
+
+    index.update_settings(json!({ "searchableAttributes": ["father", "mother"] })).await;
+
+    let documents = NESTED_DOCUMENTS.clone();
+    index.add_documents(json!(documents), None).await;
+    index.wait_task(1).await;
+
+    index.update_settings(json!({ "searchableAttributes": ["father", "mother", "doggos"], "filterableAttributes": ["doggos"] })).await;
+    index.wait_task(2).await;
+
+    // search
+    index
+        .search(
+            json!({
+                "q": "bobby",
+                "attributesToRetrieve": ["id", "doggos"]
+            }),
+            |response, code| {
+                assert_eq!(code, 200, "{}", response);
+                meili_snap::snapshot!(meili_snap::json_string!(response["hits"]), @r###"
+                [
+                  {
+                    "id": 852,
+                    "doggos": [
+                      {
+                        "name": "bobby",
+                        "age": 2
+                      },
+                      {
+                        "name": "buddy",
+                        "age": 4
+                      }
+                    ]
+                  }
+                ]
+                "###);
+            },
+        )
+        .await;
+
+    // filter
+    index
+        .search(
+            json!({
+                "q": "",
+                "filter": "doggos.age < 5",
+                "attributesToRetrieve": ["id", "doggos"]
+            }),
+            |response, code| {
+                assert_eq!(code, 200, "{}", response);
+                meili_snap::snapshot!(meili_snap::json_string!(response["hits"]), @r###"
+                [
+                  {
+                    "id": 852,
+                    "doggos": [
+                      {
+                        "name": "bobby",
+                        "age": 2
+                      },
+                      {
+                        "name": "buddy",
+                        "age": 4
+                      }
+                    ]
+                  }
+                ]
+                "###);
+            },
+        )
+        .await;
+}

crates/meilisearch/tests/search/multi.rs

@@ -113,9 +113,9 @@ async fn simple_search_single_index() {
             ],
             "_vectors": {
               "manual": [
-                -100.0,
-                340.0,
-                90.0
+                -100,
+                340,
+                90
               ]
             }
           }
@@ -138,9 +138,9 @@ async fn simple_search_single_index() {
             ],
             "_vectors": {
               "manual": [
-                1.0,
-                2.0,
-                54.0
+                1,
+                2,
+                54
               ]
             }
           }
@@ -182,9 +182,9 @@ async fn federation_single_search_single_index() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -305,9 +305,9 @@ async fn federation_two_search_single_index() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -325,9 +325,9 @@ async fn federation_two_search_single_index() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -480,9 +480,9 @@ async fn simple_search_two_indexes() {
             ],
             "_vectors": {
               "manual": [
-                -100.0,
-                340.0,
-                90.0
+                -100,
+                340,
+                90
               ]
             }
           }
@@ -513,9 +513,9 @@ async fn simple_search_two_indexes() {
             "cattos": "pésti",
             "_vectors": {
               "manual": [
-                1.0,
-                2.0,
-                3.0
+                1,
+                2,
+                3
               ]
             }
           },
@@ -535,9 +535,9 @@ async fn simple_search_two_indexes() {
             ],
             "_vectors": {
               "manual": [
-                1.0,
-                2.0,
-                54.0
+                1,
+                2,
+                54
               ]
             }
           }
@@ -585,9 +585,9 @@ async fn federation_two_search_two_indexes() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -613,9 +613,9 @@ async fn federation_two_search_two_indexes() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -640,9 +640,9 @@ async fn federation_two_search_two_indexes() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -707,9 +707,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -735,9 +735,9 @@ async fn federation_multiple_search_multiple_indexes() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -773,9 +773,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -793,9 +793,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              -23.0,
-              32.0
+              10,
+              -23,
+              32
             ]
           },
           "_federation": {
@@ -824,9 +824,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -869,9 +869,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -898,9 +898,9 @@ async fn federation_multiple_search_multiple_indexes() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              231.0,
-              32.0
+              -100,
+              231,
+              32
             ]
           },
           "_federation": {
@@ -1393,9 +1393,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -1414,9 +1414,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -1442,9 +1442,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -1474,9 +1474,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -1522,9 +1522,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -1550,9 +1550,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -1582,9 +1582,9 @@ async fn federation_sort_same_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -1716,9 +1716,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -1748,9 +1748,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -1769,9 +1769,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -1797,9 +1797,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -1845,9 +1845,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -1874,9 +1874,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           "cattos": "pésti",
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -1906,9 +1906,9 @@ async fn federation_sort_same_indexes_different_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              23.0,
-              32.0
+              10,
+              23,
+              32
             ]
           },
           "_federation": {
@@ -2103,9 +2103,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -2124,9 +2124,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              -23.0,
-              32.0
+              10,
+              -23,
+              32
             ]
           },
           "_federation": {
@@ -2145,9 +2145,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -2166,9 +2166,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              231.0,
-              32.0
+              -100,
+              231,
+              32
             ]
           },
           "_federation": {
@@ -2187,9 +2187,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -2228,9 +2228,9 @@ async fn federation_sort_different_indexes_same_criterion_same_direction() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -2415,9 +2415,9 @@ async fn federation_sort_different_ranking_rules() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -2436,9 +2436,9 @@ async fn federation_sort_different_ranking_rules() {
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              -23.0,
-              32.0
+              10,
+              -23,
+              32
             ]
           },
           "_federation": {
@@ -2457,9 +2457,9 @@ async fn federation_sort_different_ranking_rules() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -2478,9 +2478,9 @@ async fn federation_sort_different_ranking_rules() {
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              231.0,
-              32.0
+              -100,
+              231,
+              32
             ]
           },
           "_federation": {
@@ -2499,9 +2499,9 @@ async fn federation_sort_different_ranking_rules() {
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -2716,9 +2716,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -2757,9 +2757,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              10.0,
-              -23.0,
-              32.0
+              10,
+              -23,
+              32
             ]
           },
           "_federation": {
@@ -2778,9 +2778,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              340.0,
-              90.0
+              -100,
+              340,
+              90
             ]
           },
           "_federation": {
@@ -2799,9 +2799,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              -100.0,
-              231.0,
-              32.0
+              -100,
+              231,
+              32
             ]
           },
           "_federation": {
@@ -2820,9 +2820,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              3.0
+              1,
+              2,
+              3
             ]
           },
           "_federation": {
@@ -2881,9 +2881,9 @@ async fn federation_sort_different_indexes_different_criterion_same_direction()
           ],
           "_vectors": {
             "manual": [
-              1.0,
-              2.0,
-              54.0
+              1,
+              2,
+              54
             ]
           },
           "_federation": {
@@ -4346,10 +4346,10 @@ async fn federation_vector_two_indexes() {
 
     let (response, code) = server
         .multi_search(json!({"federation": {}, "queries": [
-        {"indexUid" : "vectors-animal", "vector": [1.0, 0.0, 0.5], "hybrid": {"semanticRatio": 1.0, "embedder": "animal"}},
+        {"indexUid" : "vectors-animal", "vector": [1.0, 0.0, 0.5], "hybrid": {"semanticRatio": 1.0, "embedder": "animal"}, "retrieveVectors": true},
         // joyful and energetic first
-        {"indexUid": "vectors-sentiment", "vector": [0.8, 0.6], "hybrid": {"semanticRatio": 1.0, "embedder": "sentiment"}},
-        {"indexUid": "vectors-sentiment", "q": "dog"},
+        {"indexUid": "vectors-sentiment", "vector": [0.8, 0.6], "hybrid": {"semanticRatio": 1.0, "embedder": "sentiment"}, "retrieveVectors": true},
+        {"indexUid": "vectors-sentiment", "q": "dog", "retrieveVectors": true},
         ]}))
         .await;
     snapshot!(code, @"200 OK");
@@ -4364,7 +4364,16 @@ async fn federation_vector_two_indexes() {
               0.8,
               0.09,
               0.8
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.30000001192092896
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4379,7 +4388,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               0.8,
               0.3
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.09000000357627869,
+                  0.800000011920929
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4394,7 +4413,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -1.0,
               0.1
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.8500000238418579,
+                  0.019999999552965164,
+                  0.10000000149011612
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4410,7 +4439,16 @@ async fn federation_vector_two_indexes() {
               0.9,
               0.8,
               0.05
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -0.10000000149011612,
+                  0.550000011920929
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4426,7 +4464,16 @@ async fn federation_vector_two_indexes() {
               0.85,
               0.02,
               0.1
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -1.0,
+                  0.10000000149011612
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4441,7 +4488,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -0.2,
               0.65
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.8999999761581421,
+                  0.5
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4456,7 +4513,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -0.1,
               0.55
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.8999999761581421,
+                  0.800000011920929,
+                  0.05000000074505806
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4472,7 +4539,16 @@ async fn federation_vector_two_indexes() {
               0.8,
               0.9,
               0.5
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -0.20000000298023224,
+                  0.6499999761581421
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4492,8 +4568,8 @@ async fn federation_vector_two_indexes() {
     // hybrid search, distinct embedder
     let (response, code) = server
         .multi_search(json!({"federation": {}, "queries": [
-          {"indexUid" : "vectors-animal", "vector": [1.0, 0.0, 0.5], "hybrid": {"semanticRatio": 1.0, "embedder": "animal"}, "showRankingScore": true},
-          {"indexUid": "vectors-sentiment", "vector": [-1, 0.6], "q": "beagle", "hybrid": {"semanticRatio": 1.0, "embedder": "sentiment"}, "showRankingScore": true},
+          {"indexUid" : "vectors-animal", "vector": [1.0, 0.0, 0.5], "hybrid": {"semanticRatio": 1.0, "embedder": "animal"}, "showRankingScore": true, "retrieveVectors": true},
+          {"indexUid": "vectors-sentiment", "vector": [-1, 0.6], "q": "beagle", "hybrid": {"semanticRatio": 1.0, "embedder": "sentiment"}, "showRankingScore": true, "retrieveVectors": true,},
         ]}))
         .await;
     snapshot!(code, @"200 OK");
@@ -4507,7 +4583,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               0.8,
               0.3
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.09000000357627869,
+                  0.800000011920929
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4523,7 +4609,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -1.0,
               0.1
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.8500000238418579,
+                  0.019999999552965164,
+                  0.10000000149011612
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4540,7 +4636,16 @@ async fn federation_vector_two_indexes() {
               0.85,
               0.02,
               0.1
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -1.0,
+                  0.10000000149011612
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4557,7 +4662,16 @@ async fn federation_vector_two_indexes() {
               0.8,
               0.9,
               0.5
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -0.20000000298023224,
+                  0.6499999761581421
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4573,7 +4687,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -0.2,
               0.65
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.8999999761581421,
+                  0.5
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4589,7 +4713,17 @@ async fn federation_vector_two_indexes() {
             "sentiment": [
               -0.1,
               0.55
-            ]
+            ],
+            "animal": {
+              "embeddings": [
+                [
+                  0.8999999761581421,
+                  0.800000011920929,
+                  0.05000000074505806
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-animal",
@@ -4606,7 +4740,16 @@ async fn federation_vector_two_indexes() {
               0.9,
               0.8,
               0.05
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  -0.10000000149011612,
+                  0.550000011920929
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",
@@ -4623,7 +4766,16 @@ async fn federation_vector_two_indexes() {
               0.8,
               0.09,
               0.8
-            ]
+            ],
+            "sentiment": {
+              "embeddings": [
+                [
+                  0.800000011920929,
+                  0.30000001192092896
+                ]
+              ],
+              "regenerate": false
+            }
           },
           "_federation": {
             "indexUid": "vectors-sentiment",

crates/meilisearch/tests/vector/mod.rs

@@ -250,7 +250,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Missing field `regenerate` inside `.manual`",
+        "message": "Bad embedder configuration in the document with id: `0`. Missing field `._vectors.manual.regenerate`\n  - note: `._vectors.manual` must be an array of floats, an array of arrays of floats, or an object with field `regenerate`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -280,7 +280,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Missing field `regenerate` inside `.manual`",
+        "message": "Bad embedder configuration in the document with id: `0`. Missing field `._vectors.manual.regenerate`\n  - note: `._vectors.manual` must be an array of floats, an array of arrays of floats, or an object with field `regenerate`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -311,7 +311,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.regenerate`: expected a boolean, but found a string: `\"yes please\"`",
+        "message": "Bad embedder configuration in the document with id: `0`. Could not parse `._vectors.manual.regenerate`: invalid type: string \"yes please\", expected a boolean at line 1 column 26",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -323,8 +323,7 @@ async fn user_provided_embeddings_error() {
     }
     "###);
 
-    let documents =
-        json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": true }}});
+    let documents = json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": true, "regenerate": true }}});
     let (value, code) = index.add_documents(documents, None).await;
     snapshot!(code, @"202 Accepted");
     let task = index.wait_task(value.uid()).await;
@@ -341,7 +340,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings`: expected null or an array, but found a boolean: `true`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings`: expected null or an array, but found a boolean: `true`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -353,8 +352,7 @@ async fn user_provided_embeddings_error() {
     }
     "###);
 
-    let documents =
-        json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": [true] }}});
+    let documents = json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": [true], "regenerate": true }}});
     let (value, code) = index.add_documents(documents, None).await;
     snapshot!(code, @"202 Accepted");
     let task = index.wait_task(value.uid()).await;
@@ -371,7 +369,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings[0]`: expected a number or an array, but found a boolean: `true`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings[0]`: expected a number or an array, but found a boolean: `true`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -383,8 +381,7 @@ async fn user_provided_embeddings_error() {
     }
     "###);
 
-    let documents =
-        json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": [[true]] }}});
+    let documents = json!({"id": 0, "name": "kefir", "_vectors": { "manual": { "embeddings": [[true]], "regenerate": false }}});
     let (value, code) = index.add_documents(documents, None).await;
     snapshot!(code, @"202 Accepted");
     let task = index.wait_task(value.uid()).await;
@@ -401,7 +398,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings[0][0]`: expected a number, but found a boolean: `true`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings[0][0]`: expected a number, but found a boolean: `true`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -443,7 +440,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings[1]`: expected a number, but found an array: `[0.2,0.3]`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings[1]`: expected a number, but found an array: `[0.2,0.3]`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -472,7 +469,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings[1]`: expected an array, but found a number: `0.3`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings[1]`: expected an array, but found a number: `0.3`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -501,7 +498,7 @@ async fn user_provided_embeddings_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "Bad embedder configuration in the document with id: `\"0\"`. Invalid value type at `.manual.embeddings[0][1]`: expected a number, but found a boolean: `true`",
+        "message": "Bad embedder configuration in the document with id: `0`. Invalid value type at `._vectors.manual.embeddings[0][1]`: expected a number, but found a boolean: `true`",
         "code": "invalid_vectors_type",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#invalid_vectors_type"
@@ -542,7 +539,7 @@ async fn user_provided_vectors_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "While embedding documents for embedder `manual`: no vectors provided for document \"40\" and at least 4 other document(s)\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: opt-out for a document with `_vectors.manual: null`",
+        "message": "While embedding documents for embedder `manual`: no vectors provided for document `40` and at least 4 other document(s)\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: opt-out for a document with `_vectors.manual: null`",
         "code": "vector_embedding_error",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#vector_embedding_error"
@@ -572,7 +569,7 @@ async fn user_provided_vectors_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "While embedding documents for embedder `manual`: no vectors provided for document \"42\"\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: try replacing `_vector` by `_vectors` in 1 document(s).",
+        "message": "While embedding documents for embedder `manual`: no vectors provided for document `42`\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: try replacing `_vector` by `_vectors` in 1 document(s).",
         "code": "vector_embedding_error",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#vector_embedding_error"
@@ -602,7 +599,7 @@ async fn user_provided_vectors_error() {
         "indexedDocuments": 0
       },
       "error": {
-        "message": "While embedding documents for embedder `manual`: no vectors provided for document \"42\"\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: try replacing `_vectors.manaul` by `_vectors.manual` in 1 document(s).",
+        "message": "While embedding documents for embedder `manual`: no vectors provided for document `42`\n- Note: `manual` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.manual`.\n- Hint: try replacing `_vectors.manaul` by `_vectors.manual` in 1 document(s).",
         "code": "vector_embedding_error",
         "type": "invalid_request",
         "link": "https://docs.meilisearch.com/errors#vector_embedding_error"

crates/meilisearch/tests/vector/openai.rs

@@ -137,13 +137,14 @@ fn long_text() -> &'static str {
 }
 
 async fn create_mock_tokenized() -> (MockServer, Value) {
-    create_mock_with_template("{{doc.text}}", ModelDimensions::Large, false).await
+    create_mock_with_template("{{doc.text}}", ModelDimensions::Large, false, false).await
 }
 
 async fn create_mock_with_template(
     document_template: &str,
     model_dimensions: ModelDimensions,
     fallible: bool,
+    slow: bool,
 ) -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
     const API_KEY: &str = "my-api-key";
@@ -154,7 +155,11 @@ async fn create_mock_with_template(
     Mock::given(method("POST"))
         .and(path("/"))
         .respond_with(move |req: &Request| {
-            // 0. maybe return 500
+            // 0. wait for a long time
+            if slow {
+              std::thread::sleep(std::time::Duration::from_secs(1));
+            }
+            // 1. maybe return 500
             if fallible {
              let attempt = attempt.fetch_add(1, Ordering::Relaxed);
              let failed = matches!(attempt % 4, 0 | 1 | 3);
@@ -167,7 +172,7 @@ async fn create_mock_with_template(
                 }))
              }
             }
-            // 1. check API key
+            // 3. check API key
             match req.headers.get("Authorization") {
                 Some(api_key) if api_key == API_KEY_BEARER => {
                     {}
@@ -202,7 +207,7 @@ async fn create_mock_with_template(
                     )
                 }
             }
-            // 2. parse text inputs
+            // 3. parse text inputs
             let query: serde_json::Value = match req.body_json() {
                 Ok(query) => query,
                 Err(_error) => return ResponseTemplate::new(400).set_body_json(
@@ -223,7 +228,7 @@ async fn create_mock_with_template(
                 panic!("Expected {model_dimensions:?}, got {query_model_dimensions:?}")
             }
 
-            // 3. for each text, find embedding in responses
+            // 4. for each text, find embedding in responses
             let serde_json::Value::Array(inputs) = &query["input"] else {
                 panic!("Unexpected `input` value")
             };
@@ -283,7 +288,7 @@ async fn create_mock_with_template(
                 "embedding": embedding,
             })).collect();
 
-            // 4. produce output from embeddings
+            // 5. produce output from embeddings
             ResponseTemplate::new(200).set_body_json(json!({
                 "object": "list",
                 "data": data,
@@ -317,23 +322,27 @@ const DOGGO_TEMPLATE: &str = r#"{%- if doc.gender == "F" -%}Une chienne nommée
         {%- endif %}, de race {{doc.breed}}."#;
 
 async fn create_mock() -> (MockServer, Value) {
-    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large, false).await
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large, false, false).await
 }
 
 async fn create_mock_dimensions() -> (MockServer, Value) {
-    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large512, false).await
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large512, false, false).await
 }
 
 async fn create_mock_small_embedding_model() -> (MockServer, Value) {
-    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Small, false).await
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Small, false, false).await
 }
 
 async fn create_mock_legacy_embedding_model() -> (MockServer, Value) {
-    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Ada, false).await
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Ada, false, false).await
 }
 
 async fn create_fallible_mock() -> (MockServer, Value) {
-    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large, true).await
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large, true, false).await
+}
+
+async fn create_slow_mock() -> (MockServer, Value) {
+    create_mock_with_template(DOGGO_TEMPLATE, ModelDimensions::Large, true, true).await
 }
 
 // basic test "it works"
@@ -1884,4 +1893,115 @@ async fn it_still_works() {
     ]
     "###);
 }
+
+// test with a server that responds 500 on 3 out of 4 calls
+#[actix_rt::test]
+async fn timeout() {
+    let (_mock, setting) = create_slow_mock().await;
+    let server = get_server_vector().await;
+    let index = server.index("doggo");
+
+    let (response, code) = index
+        .update_settings(json!({
+          "embedders": {
+              "default": setting,
+          },
+        }))
+        .await;
+    snapshot!(code, @"202 Accepted");
+    let task = server.wait_task(response.uid()).await;
+    snapshot!(task["status"], @r###""succeeded""###);
+    let documents = json!([
+      {"id": 0, "name": "kefir", "gender": "M", "birthyear": 2023, "breed": "Patou"},
+    ]);
+    let (value, code) = index.add_documents(documents, None).await;
+    snapshot!(code, @"202 Accepted");
+    let task = index.wait_task(value.uid()).await;
+    snapshot!(task, @r###"
+    {
+      "uid": "[uid]",
+      "batchUid": "[batch_uid]",
+      "indexUid": "doggo",
+      "status": "succeeded",
+      "type": "documentAdditionOrUpdate",
+      "canceledBy": null,
+      "details": {
+        "receivedDocuments": 1,
+        "indexedDocuments": 1
+      },
+      "error": null,
+      "duration": "[duration]",
+      "enqueuedAt": "[date]",
+      "startedAt": "[date]",
+      "finishedAt": "[date]"
+    }
+    "###);
+
+    let (documents, _code) = index
+        .get_all_documents(GetAllDocumentsOptions { retrieve_vectors: true, ..Default::default() })
+        .await;
+    snapshot!(json_string!(documents, {".results.*._vectors.default.embeddings" => "[vector]"}), @r###"
+    {
+      "results": [
+        {
+          "id": 0,
+          "name": "kefir",
+          "gender": "M",
+          "birthyear": 2023,
+          "breed": "Patou",
+          "_vectors": {
+            "default": {
+              "embeddings": "[vector]",
+              "regenerate": true
+            }
+          }
+        }
+      ],
+      "offset": 0,
+      "limit": 20,
+      "total": 1
+    }
+    "###);
+
+    let (response, code) = index
+        .search_post(json!({
+            "q": "grand chien de berger des montagnes",
+            "hybrid": {"semanticRatio": 0.99, "embedder": "default"}
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["semanticHitCount"]), @"0");
+    snapshot!(json_string!(response["hits"]), @"[]");
+
+    let (response, code) = index
+        .search_post(json!({
+            "q": "grand chien de berger des montagnes",
+            "hybrid": {"semanticRatio": 0.99, "embedder": "default"}
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["semanticHitCount"]), @"1");
+    snapshot!(json_string!(response["hits"]), @r###"
+    [
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
+      }
+    ]
+    "###);
+
+    let (response, code) = index
+        .search_post(json!({
+            "q": "grand chien de berger des montagnes",
+            "hybrid": {"semanticRatio": 0.99, "embedder": "default"}
+        }))
+        .await;
+    snapshot!(code, @"200 OK");
+    snapshot!(json_string!(response["semanticHitCount"]), @"0");
+    snapshot!(json_string!(response["hits"]), @"[]");
+}
+
 // test with a server that wrongly responds 400

crates/meilisearch/tests/vector/rest.rs

@@ -1,5 +1,4 @@
 use std::collections::BTreeMap;
-use std::sync::atomic::{AtomicUsize, Ordering};
 
 use meili_snap::{json_string, snapshot};
 use reqwest::IntoUrl;
@@ -13,13 +12,22 @@ use crate::vector::{get_server_vector, GetAllDocumentsOptions};
 async fn create_mock() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
-    let counter = AtomicUsize::new(0);
+    let text_to_embedding: BTreeMap<_, _> = vec![
+        // text -> embedding
+        ("kefir", [0.0, 0.0, 0.0]),
+        ("intel", [1.0, 1.0, 1.0]),
+    ]
+    // turn into btree
+    .into_iter()
+    .collect();
 
     Mock::given(method("POST"))
         .and(path("/"))
-        .respond_with(move |_req: &Request| {
-            let counter = counter.fetch_add(1, Ordering::Relaxed);
-            ResponseTemplate::new(200).set_body_json(json!({ "data": vec![counter; 3] }))
+        .respond_with(move |req: &Request| {
+            let text: String = req.body_json().unwrap();
+            ResponseTemplate::new(200).set_body_json(
+                json!({ "data": text_to_embedding.get(text.as_str()).unwrap_or(&[99., 99., 99.]) }),
+            )
         })
         .mount(&mock_server)
         .await;
@@ -32,13 +40,14 @@ async fn create_mock() -> (MockServer, Value) {
         "request": "{{text}}",
         "response": {
           "data": "{{embedding}}"
-        }
+        },
+        "documentTemplate": "{{doc.name}}",
     });
 
     (mock_server, embedder_settings)
 }
 
-async fn create_mock_map() -> (MockServer, Value) {
+async fn create_mock_default_template() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
     let text_to_embedding: BTreeMap<_, _> = vec![
@@ -97,7 +106,14 @@ struct SingleResponse {
 async fn create_mock_multiple() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
-    let counter = AtomicUsize::new(0);
+    let text_to_embedding: BTreeMap<_, _> = vec![
+        // text -> embedding
+        ("kefir", [0.0, 0.0, 0.0]),
+        ("intel", [1.0, 1.0, 1.0]),
+    ]
+    // turn into btree
+    .into_iter()
+    .collect();
 
     Mock::given(method("POST"))
         .and(path("/"))
@@ -115,8 +131,11 @@ async fn create_mock_multiple() -> (MockServer, Value) {
                 .input
                 .into_iter()
                 .map(|text| SingleResponse {
+                    embedding: text_to_embedding
+                        .get(text.as_str())
+                        .unwrap_or(&[99., 99., 99.])
+                        .to_vec(),
                     text,
-                    embedding: vec![counter.fetch_add(1, Ordering::Relaxed) as f32; 3],
                 })
                 .collect();
 
@@ -142,7 +161,8 @@ async fn create_mock_multiple() -> (MockServer, Value) {
             },
             "{{..}}"
           ]
-        }
+        },
+        "documentTemplate": "{{doc.name}}"
     });
 
     (mock_server, embedder_settings)
@@ -156,7 +176,14 @@ struct SingleRequest {
 async fn create_mock_single_response_in_array() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
-    let counter = AtomicUsize::new(0);
+    let text_to_embedding: BTreeMap<_, _> = vec![
+        // text -> embedding
+        ("kefir", [0.0, 0.0, 0.0]),
+        ("intel", [1.0, 1.0, 1.0]),
+    ]
+    // turn into btree
+    .into_iter()
+    .collect();
 
     Mock::given(method("POST"))
         .and(path("/"))
@@ -171,8 +198,11 @@ async fn create_mock_single_response_in_array() -> (MockServer, Value) {
             };
 
             let output = vec![SingleResponse {
+                embedding: text_to_embedding
+                    .get(req.input.as_str())
+                    .unwrap_or(&[99., 99., 99.])
+                    .to_vec(),
                 text: req.input,
-                embedding: vec![counter.fetch_add(1, Ordering::Relaxed) as f32; 3],
             }];
 
             let response = MultipleResponse { output };
@@ -196,7 +226,8 @@ async fn create_mock_single_response_in_array() -> (MockServer, Value) {
               "embedding": "{{embedding}}"
             }
           ]
-        }
+        },
+        "documentTemplate": "{{doc.name}}"
     });
 
     (mock_server, embedder_settings)
@@ -205,7 +236,14 @@ async fn create_mock_single_response_in_array() -> (MockServer, Value) {
 async fn create_mock_raw_with_custom_header() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
-    let counter = AtomicUsize::new(0);
+    let text_to_embedding: BTreeMap<_, _> = vec![
+        // text -> embedding
+        ("kefir", [0.0, 0.0, 0.0]),
+        ("intel", [1.0, 1.0, 1.0]),
+    ]
+    // turn into btree
+    .into_iter()
+    .collect();
 
     Mock::given(method("POST"))
         .and(path("/"))
@@ -223,7 +261,7 @@ async fn create_mock_raw_with_custom_header() -> (MockServer, Value) {
                 }
             }
 
-            let _req: String = match req.body_json() {
+            let req: String = match req.body_json() {
                 Ok(req) => req,
                 Err(error) => {
                     return ResponseTemplate::new(400).set_body_json(json!({
@@ -232,7 +270,7 @@ async fn create_mock_raw_with_custom_header() -> (MockServer, Value) {
                 }
             };
 
-            let output = vec![counter.fetch_add(1, Ordering::Relaxed) as f32; 3];
+            let output = text_to_embedding.get(req.as_str()).unwrap_or(&[99., 99., 99.]).to_vec();
 
             ResponseTemplate::new(200).set_body_json(output)
         })
@@ -245,7 +283,8 @@ async fn create_mock_raw_with_custom_header() -> (MockServer, Value) {
         "url": url,
         "request": "{{text}}",
         "response": "{{embedding}}",
-        "headers": {"my-nonstandard-auth": "bearer of the ring"}
+        "headers": {"my-nonstandard-auth": "bearer of the ring"},
+        "documentTemplate": "{{doc.name}}"
     });
 
     (mock_server, embedder_settings)
@@ -254,12 +293,19 @@ async fn create_mock_raw_with_custom_header() -> (MockServer, Value) {
 async fn create_mock_raw() -> (MockServer, Value) {
     let mock_server = MockServer::start().await;
 
-    let counter = AtomicUsize::new(0);
+    let text_to_embedding: BTreeMap<_, _> = vec![
+        // text -> embedding
+        ("kefir", [0.0, 0.0, 0.0]),
+        ("intel", [1.0, 1.0, 1.0]),
+    ]
+    // turn into btree
+    .into_iter()
+    .collect();
 
     Mock::given(method("POST"))
         .and(path("/"))
         .respond_with(move |req: &Request| {
-            let _req: String = match req.body_json() {
+            let req: String = match req.body_json() {
                 Ok(req) => req,
                 Err(error) => {
                     return ResponseTemplate::new(400).set_body_json(json!({
@@ -268,7 +314,7 @@ async fn create_mock_raw() -> (MockServer, Value) {
                 }
             };
 
-            let output = vec![counter.fetch_add(1, Ordering::Relaxed) as f32; 3];
+            let output = text_to_embedding.get(req.as_str()).unwrap_or(&[99., 99., 99.]).to_vec();
 
             ResponseTemplate::new(200).set_body_json(output)
         })
@@ -281,29 +327,30 @@ async fn create_mock_raw() -> (MockServer, Value) {
         "url": url,
         "dimensions": 3,
         "request": "{{text}}",
-        "response": "{{embedding}}"
+        "response": "{{embedding}}",
+        "documentTemplate": "{{doc.name}}"
     });
 
     (mock_server, embedder_settings)
 }
 
-pub async fn post<T: IntoUrl>(url: T) -> reqwest::Result<reqwest::Response> {
-    reqwest::Client::builder().build()?.post(url).send().await
+pub async fn post<T: IntoUrl>(url: T, text: &str) -> reqwest::Result<reqwest::Response> {
+    reqwest::Client::builder().build()?.post(url).json(&json!(text)).send().await
 }
 
 #[actix_rt::test]
 async fn dummy_testing_the_mock() {
     let (mock, _setting) = create_mock().await;
-    let body = post(&mock.uri()).await.unwrap().text().await.unwrap();
-    snapshot!(body, @r###"{"data":[0,0,0]}"###);
-    let body = post(&mock.uri()).await.unwrap().text().await.unwrap();
-    snapshot!(body, @r###"{"data":[1,1,1]}"###);
-    let body = post(&mock.uri()).await.unwrap().text().await.unwrap();
-    snapshot!(body, @r###"{"data":[2,2,2]}"###);
-    let body = post(&mock.uri()).await.unwrap().text().await.unwrap();
-    snapshot!(body, @r###"{"data":[3,3,3]}"###);
-    let body = post(&mock.uri()).await.unwrap().text().await.unwrap();
-    snapshot!(body, @r###"{"data":[4,4,4]}"###);
+    let body = post(&mock.uri(), "kefir").await.unwrap().text().await.unwrap();
+    snapshot!(body, @r###"{"data":[0.0,0.0,0.0]}"###);
+    let body = post(&mock.uri(), "intel").await.unwrap().text().await.unwrap();
+    snapshot!(body, @r###"{"data":[1.0,1.0,1.0]}"###);
+    let body = post(&mock.uri(), "kefir").await.unwrap().text().await.unwrap();
+    snapshot!(body, @r###"{"data":[0.0,0.0,0.0]}"###);
+    let body = post(&mock.uri(), "kefir").await.unwrap().text().await.unwrap();
+    snapshot!(body, @r###"{"data":[0.0,0.0,0.0]}"###);
+    let body = post(&mock.uri(), "intel").await.unwrap().text().await.unwrap();
+    snapshot!(body, @r###"{"data":[1.0,1.0,1.0]}"###);
 }
 
 #[actix_rt::test]
@@ -954,7 +1001,7 @@ async fn bad_settings() {
     let (response, code) = index
         .update_settings(json!({
           "embedders": {
-              "rest": json!({ "source": "rest", "url": mock.uri(), "request": "{{text}}", "response": { "data": "{{embedding}}" }, "dimensions": 2 }),
+              "rest": json!({ "source": "rest", "url": mock.uri(), "request": "{{text}}", "response": { "data": "{{embedding}}" }, "dimensions": 2, "documentTemplate": "{{doc.name}}" }),
           },
         }))
         .await;
@@ -1937,6 +1984,7 @@ async fn server_custom_header() {
         "embedders": {
           "rest": {
             "source": "rest",
+            "documentTemplate": "{{doc.name}}",
             "url": "[url]",
             "request": "{{text}}",
             "response": "{{embedding}}",
@@ -1957,7 +2005,7 @@ async fn server_custom_header() {
 
 #[actix_rt::test]
 async fn searchable_reindex() {
-    let (_mock, setting) = create_mock_map().await;
+    let (_mock, setting) = create_mock_default_template().await;
     let server = get_server_vector().await;
     let index = server.index("doggo");
 

crates/meilitool/src/main.rs

@@ -264,7 +264,7 @@ fn export_a_dump(
                         format!("While iterating on content file {:?}", content_file_uuid)
                     })? {
                         dump_content_file
-                            .push_document(&obkv_to_object(&doc, &documents_batch_index)?)?;
+                            .push_document(&obkv_to_object(doc, &documents_batch_index)?)?;
                     }
                     dump_content_file.flush()?;
                     count += 1;

crates/milli/Cargo.toml

@@ -12,12 +12,14 @@ readme.workspace = true
 license.workspace = true
 
 [dependencies]
+big_s = "1.0.2"
 bimap = { version = "0.6.3", features = ["serde"] }
 bincode = "1.3.3"
 bstr = "1.9.1"
 bytemuck = { version = "1.18.0", features = ["extern_crate_alloc"] }
 byteorder = "1.5.0"
-charabia = { version = "0.9.1", default-features = false }
+# charabia = { version = "0.9.0", default-features = false }
+charabia = { git = "https://github.com/meilisearch/charabia", branch = "mutualize-char-normalizer", default-features = false }
 concat-arrays = "0.1.2"
 crossbeam-channel = "0.5.13"
 deserr = "0.6.2"
@@ -27,9 +29,9 @@ fst = "0.4.7"
 fxhash = "0.2.1"
 geoutils = "0.5.1"
 grenad = { version = "0.4.7", default-features = false, features = [
-    "rayon",
+    "rayon",    # TODO Should we keep this feature
     "tempfile",
-] }
+], git = "https://github.com/meilisearch/grenad", branch = "various-improvements" }
 heed = { version = "0.20.3", default-features = false, features = [
     "serde-json",
     "serde-bincode",
@@ -40,14 +42,14 @@ json-depth-checker = { path = "../json-depth-checker" }
 levenshtein_automata = { version = "0.2.1", features = ["fst_automaton"] }
 memchr = "2.5.0"
 memmap2 = "0.9.4"
-obkv = "0.2.2"
+obkv = { git = "https://github.com/kerollmops/obkv", branch = "unsized-kvreader" }
 once_cell = "1.19.0"
 ordered-float = "4.2.1"
 rayon = "1.10.0"
 roaring = { version = "0.10.6", features = ["serde"] }
 rstar = { version = "0.12.0", features = ["serde"] }
 serde = { version = "1.0.204", features = ["derive"] }
-serde_json = { version = "1.0.120", features = ["preserve_order"] }
+serde_json = { version = "1.0.120", features = ["preserve_order", "raw_value"] }
 slice-group-by = "0.3.1"
 smallstr = { version = "0.3.0", features = ["serde"] }
 smallvec = "1.13.2"
@@ -79,17 +81,30 @@ hf-hub = { git = "https://github.com/dureuill/hf-hub.git", branch = "rust_tls",
 ] }
 tiktoken-rs = "0.5.9"
 liquid = "0.26.6"
-rhai = { git = "https://github.com/rhaiscript/rhai", rev = "ef3df63121d27aacd838f366f2b83fd65f20a1e4", features = ["serde", "no_module", "no_custom_syntax", "no_time", "sync"] }
+rhai = { git = "https://github.com/rhaiscript/rhai", rev = "ef3df63121d27aacd838f366f2b83fd65f20a1e4", features = [
+    "serde",
+    "no_module",
+    "no_custom_syntax",
+    "no_time",
+    "sync",
+] }
 arroy = "0.5.0"
 rand = "0.8.5"
 tracing = "0.1.40"
 ureq = { version = "2.10.0", features = ["json"] }
 url = "2.5.2"
 rayon-par-bridge = "0.1.0"
+hashbrown = "0.15.0"
+raw-collections = { git = "https://github.com/dureuill/raw-collections.git", version = "0.1.0" }
+bumpalo = "3.16.0"
+thread_local = "1.1.8"
+allocator-api2 = "0.2.18"
+rustc-hash = "2.0.0"
+uell = "0.1.0"
+enum-iterator = "2.1.0"
 
 [dev-dependencies]
 mimalloc = { version = "0.1.43", default-features = false }
-big_s = "1.0.2"
 insta = "1.39.0"
 maplit = "1.0.2"
 md5 = "0.7.0"

crates/milli/examples/index.rs

@@ -1,114 +0,0 @@
-use std::error::Error;
-use std::fs::File;
-use std::io::{BufRead, BufReader, Cursor, Seek};
-use std::path::Path;
-
-use heed::EnvOpenOptions;
-use milli::documents::{DocumentsBatchBuilder, DocumentsBatchReader};
-use milli::update::{IndexDocuments, IndexDocumentsConfig, IndexerConfig, Settings};
-use milli::{Index, Object};
-
-fn usage(error: &str, program_name: &str) -> String {
-    format!(
-        "{}. Usage: {} <PATH-TO-INDEX> <PATH-TO-DATASET> [searchable_fields] [filterable_fields]",
-        error, program_name
-    )
-}
-
-fn main() -> Result<(), Box<dyn Error>> {
-    let mut args = std::env::args();
-    let program_name = args.next().expect("No program name");
-    let index_path =
-        args.next().unwrap_or_else(|| panic!("{}", usage("Missing path to index.", &program_name)));
-    let dataset_path = args
-        .next()
-        .unwrap_or_else(|| panic!("{}", usage("Missing path to source dataset.", &program_name)));
-    // let primary_key = args.next().unwrap_or_else(|| "id".into());
-    // "title overview"
-    let searchable_fields: Vec<String> = args
-        .next()
-        .map(|arg| arg.split_whitespace().map(ToString::to_string).collect())
-        .unwrap_or_default();
-
-    println!("{searchable_fields:?}");
-    // "release_date genres"
-    let filterable_fields: Vec<String> = args
-        .next()
-        .map(|arg| arg.split_whitespace().map(ToString::to_string).collect())
-        .unwrap_or_default();
-
-    let mut options = EnvOpenOptions::new();
-    options.map_size(100 * 1024 * 1024 * 1024); // 100 GB
-
-    std::fs::create_dir_all(&index_path).unwrap();
-    let index = Index::new(options, index_path).unwrap();
-    let mut wtxn = index.write_txn().unwrap();
-
-    let config = IndexerConfig::default();
-    let mut builder = Settings::new(&mut wtxn, &index, &config);
-    // builder.set_primary_key(primary_key);
-    let searchable_fields = searchable_fields.iter().map(|s| s.to_string()).collect();
-    builder.set_searchable_fields(searchable_fields);
-    let filterable_fields = filterable_fields.iter().map(|s| s.to_string()).collect();
-    builder.set_filterable_fields(filterable_fields);
-
-    builder.execute(|_| (), || false).unwrap();
-
-    let config = IndexerConfig::default();
-    let indexing_config = IndexDocumentsConfig::default();
-
-    let builder =
-        IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| (), || false).unwrap();
-
-    let documents = documents_from(
-        &dataset_path,
-        Path::new(&dataset_path).extension().unwrap_or_default().to_str().unwrap_or_default(),
-    );
-    let (builder, user_error) = builder.add_documents(documents).unwrap();
-    user_error.unwrap();
-    builder.execute().unwrap();
-    wtxn.commit().unwrap();
-
-    index.prepare_for_closing().wait();
-    Ok(())
-}
-fn documents_from(filename: &str, filetype: &str) -> DocumentsBatchReader<impl BufRead + Seek> {
-    let reader = File::open(filename)
-        .unwrap_or_else(|_| panic!("could not find the dataset in: {}", filename));
-    let reader = BufReader::new(reader);
-    let documents = match filetype {
-        "csv" => documents_from_csv(reader).unwrap(),
-        "json" => documents_from_json(reader).unwrap(),
-        "jsonl" => documents_from_jsonl(reader).unwrap(),
-        otherwise => panic!("invalid update format {:?}", otherwise),
-    };
-    DocumentsBatchReader::from_reader(Cursor::new(documents)).unwrap()
-}
-
-fn documents_from_jsonl(reader: impl BufRead) -> milli::Result<Vec<u8>> {
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
-
-    for result in serde_json::Deserializer::from_reader(reader).into_iter::<Object>() {
-        let object = result.unwrap();
-        documents.append_json_object(&object)?;
-    }
-
-    documents.into_inner().map_err(Into::into)
-}
-
-fn documents_from_json(reader: impl BufRead) -> milli::Result<Vec<u8>> {
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
-
-    documents.append_json_array(reader)?;
-
-    documents.into_inner().map_err(Into::into)
-}
-
-fn documents_from_csv(reader: impl BufRead) -> milli::Result<Vec<u8>> {
-    let csv = csv::Reader::from_reader(reader);
-
-    let mut documents = DocumentsBatchBuilder::new(Vec::new());
-    documents.append_csv(csv)?;
-
-    documents.into_inner().map_err(Into::into)
-}

crates/milli/examples/search.rs

@@ -1,124 +0,0 @@
-use std::error::Error;
-use std::io::stdin;
-use std::path::Path;
-use std::time::Instant;
-
-use heed::EnvOpenOptions;
-use milli::{
-    execute_search, filtered_universe, DefaultSearchLogger, GeoSortStrategy, Index, SearchContext,
-    SearchLogger, TermsMatchingStrategy, TimeBudget,
-};
-
-#[global_allocator]
-static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
-
-fn main() -> Result<(), Box<dyn Error>> {
-    let mut args = std::env::args();
-    let program_name = args.next().expect("No program name");
-    let dataset = args.next().unwrap_or_else(|| {
-        panic!(
-            "Missing path to index. Usage: {} <PATH-TO-INDEX> [<logger-dir>] [print-documents]",
-            program_name
-        )
-    });
-    let detailed_logger_dir = args.next();
-    let print_documents: bool =
-        if let Some(arg) = args.next() { arg == "print-documents" } else { false };
-
-    let mut options = EnvOpenOptions::new();
-    options.map_size(100 * 1024 * 1024 * 1024); // 100 GB
-
-    let index = Index::new(options, dataset)?;
-    let txn = index.read_txn()?;
-    let mut query = String::new();
-    while stdin().read_line(&mut query)? > 0 {
-        for _ in 0..2 {
-            let mut default_logger = DefaultSearchLogger;
-            // FIXME: consider resetting the state of the logger between search executions as otherwise panics are possible.
-            // Workaround'd here by recreating the logger on each iteration of the loop
-            let mut detailed_logger = detailed_logger_dir
-                .as_ref()
-                .map(|logger_dir| (milli::VisualSearchLogger::default(), logger_dir));
-            let logger: &mut dyn SearchLogger<_> =
-                if let Some((detailed_logger, _)) = detailed_logger.as_mut() {
-                    detailed_logger
-                } else {
-                    &mut default_logger
-                };
-
-            let start = Instant::now();
-
-            let mut ctx = SearchContext::new(&index, &txn)?;
-            let universe = filtered_universe(ctx.index, ctx.txn, &None)?;
-
-            let docs = execute_search(
-                &mut ctx,
-                (!query.trim().is_empty()).then(|| query.trim()),
-                TermsMatchingStrategy::Last,
-                milli::score_details::ScoringStrategy::Skip,
-                false,
-                universe,
-                &None,
-                &None,
-                GeoSortStrategy::default(),
-                0,
-                20,
-                None,
-                &mut DefaultSearchLogger,
-                logger,
-                TimeBudget::max(),
-                None,
-                None,
-            )?;
-            if let Some((logger, dir)) = detailed_logger {
-                logger.finish(&mut ctx, Path::new(dir))?;
-            }
-            let elapsed = start.elapsed();
-            println!("new: {}us, docids: {:?}", elapsed.as_micros(), docs.documents_ids);
-            if print_documents {
-                let documents = index
-                    .documents(&txn, docs.documents_ids.iter().copied())
-                    .unwrap()
-                    .into_iter()
-                    .map(|(id, obkv)| {
-                        let mut object = serde_json::Map::default();
-                        for (fid, fid_name) in index.fields_ids_map(&txn).unwrap().iter() {
-                            let value = obkv.get(fid).unwrap();
-                            let value: serde_json::Value = serde_json::from_slice(value).unwrap();
-                            object.insert(fid_name.to_owned(), value);
-                        }
-                        (id, serde_json::to_string_pretty(&object).unwrap())
-                    })
-                    .collect::<Vec<_>>();
-
-                for (id, document) in documents {
-                    println!("{id}:");
-                    println!("{document}");
-                }
-
-                let documents = index
-                    .documents(&txn, docs.documents_ids.iter().copied())
-                    .unwrap()
-                    .into_iter()
-                    .map(|(id, obkv)| {
-                        let mut object = serde_json::Map::default();
-                        for (fid, fid_name) in index.fields_ids_map(&txn).unwrap().iter() {
-                            let value = obkv.get(fid).unwrap();
-                            let value: serde_json::Value = serde_json::from_slice(value).unwrap();
-                            object.insert(fid_name.to_owned(), value);
-                        }
-                        (id, serde_json::to_string_pretty(&object).unwrap())
-                    })
-                    .collect::<Vec<_>>();
-                println!("{}us: {:?}", elapsed.as_micros(), docs.documents_ids);
-                for (id, document) in documents {
-                    println!("{id}:");
-                    println!("{document}");
-                }
-            }
-        }
-        query.clear();
-    }
-
-    Ok(())
-}

crates/milli/examples/settings.rs

@@ -1,33 +0,0 @@
-// use big_s::S;
-use heed::EnvOpenOptions;
-// use maplit::hashset;
-use milli::{
-    update::{IndexerConfig, Settings},
-    Criterion, Index,
-};
-
-fn main() {
-    let mut options = EnvOpenOptions::new();
-    options.map_size(100 * 1024 * 1024 * 1024); // 100 GB
-
-    let index = Index::new(options, "data_movies.ms").unwrap();
-    let mut wtxn = index.write_txn().unwrap();
-
-    let config = IndexerConfig::default();
-    let mut builder = Settings::new(&mut wtxn, &index, &config);
-
-    // builder.set_min_word_len_one_typo(5);
-    // builder.set_min_word_len_two_typos(7);
-    // builder.set_sortable_fields(hashset! { S("release_date") });
-    builder.set_criteria(vec![
-        Criterion::Words,
-        Criterion::Typo,
-        Criterion::Proximity,
-        Criterion::Attribute,
-        Criterion::Sort,
-        Criterion::Exactness,
-    ]);
-
-    builder.execute(|_| (), || false).unwrap();
-    wtxn.commit().unwrap();
-}

crates/milli/src/documents/builder.rs

@@ -292,7 +292,7 @@ mod test {
             .unwrap()
             .into_cursor_and_fields_index();
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -321,7 +321,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -348,7 +348,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -375,7 +375,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -402,7 +402,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -429,7 +429,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -456,7 +456,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -483,7 +483,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -510,7 +510,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,
@@ -555,7 +555,7 @@ mod test {
             .into_cursor_and_fields_index();
 
         let doc = cursor.next_document().unwrap().unwrap();
-        let val = obkv_to_object(&doc, &index).map(Value::from).unwrap();
+        let val = obkv_to_object(doc, &index).map(Value::from).unwrap();
 
         assert_eq!(
             val,

crates/milli/src/documents/enriched.rs

@@ -69,7 +69,7 @@ impl<R: io::Read + io::Seek> EnrichedDocumentsBatchReader<R> {
 
 #[derive(Debug, Clone)]
 pub struct EnrichedDocument<'a> {
-    pub document: KvReader<'a, FieldId>,
+    pub document: &'a KvReader<FieldId>,
     pub document_id: DocumentId,
 }
 

crates/milli/src/documents/mod.rs

@@ -13,8 +13,8 @@ pub use builder::DocumentsBatchBuilder;
 pub use enriched::{EnrichedDocument, EnrichedDocumentsBatchCursor, EnrichedDocumentsBatchReader};
 use obkv::KvReader;
 pub use primary_key::{
-    validate_document_id_value, DocumentIdExtractionError, FieldIdMapper, PrimaryKey,
-    DEFAULT_PRIMARY_KEY,
+    validate_document_id_str, validate_document_id_value, DocumentIdExtractionError, FieldIdMapper,
+    PrimaryKey, DEFAULT_PRIMARY_KEY,
 };
 pub use reader::{DocumentsBatchCursor, DocumentsBatchCursorError, DocumentsBatchReader};
 use serde::{Deserialize, Serialize};
@@ -27,7 +27,7 @@ use crate::{FieldId, Object, Result};
 const DOCUMENTS_BATCH_INDEX_KEY: [u8; 8] = u64::MAX.to_be_bytes();
 
 /// Helper function to convert an obkv reader into a JSON object.
-pub fn obkv_to_object(obkv: &KvReader<'_, FieldId>, index: &DocumentsBatchIndex) -> Result<Object> {
+pub fn obkv_to_object(obkv: &KvReader<FieldId>, index: &DocumentsBatchIndex) -> Result<Object> {
     obkv.iter()
         .map(|(field_id, value)| {
             let field_name = index
@@ -76,7 +76,7 @@ impl DocumentsBatchIndex {
         self.0.get_by_right(name).cloned()
     }
 
-    pub fn recreate_json(&self, document: &obkv::KvReaderU16<'_>) -> Result<Object> {
+    pub fn recreate_json(&self, document: &obkv::KvReaderU16) -> Result<Object> {
         let mut map = Object::new();
 
         for (k, v) in document.iter() {
@@ -96,6 +96,10 @@ impl FieldIdMapper for DocumentsBatchIndex {
     fn id(&self, name: &str) -> Option<FieldId> {
         self.id(name)
     }
+
+    fn name(&self, id: FieldId) -> Option<&str> {
+        self.name(id)
+    }
 }
 
 #[derive(Debug, thiserror::Error)]
@@ -146,11 +150,33 @@ pub fn objects_from_json_value(json: serde_json::Value) -> Vec<crate::Object> {
 macro_rules! documents {
     ($data:tt) => {{
         let documents = serde_json::json!($data);
-        let documents = $crate::documents::objects_from_json_value(documents);
-        $crate::documents::documents_batch_reader_from_objects(documents)
+        let mut file = tempfile::tempfile().unwrap();
+
+        match documents {
+            serde_json::Value::Array(vec) => {
+                for document in vec {
+                    serde_json::to_writer(&mut file, &document).unwrap();
+                }
+            }
+            serde_json::Value::Object(document) => {
+                serde_json::to_writer(&mut file, &document).unwrap();
+            }
+            _ => unimplemented!("The `documents!` macro only support Objects and Array"),
+        }
+        file.sync_all().unwrap();
+        unsafe { memmap2::Mmap::map(&file).unwrap() }
     }};
 }
 
+pub fn mmap_from_objects(objects: impl IntoIterator<Item = Object>) -> memmap2::Mmap {
+    let mut writer = tempfile::tempfile().map(std::io::BufWriter::new).unwrap();
+    for object in objects {
+        serde_json::to_writer(&mut writer, &object).unwrap();
+    }
+    let file = writer.into_inner().unwrap();
+    unsafe { memmap2::Mmap::map(&file).unwrap() }
+}
+
 pub fn documents_batch_reader_from_objects(
     objects: impl IntoIterator<Item = Object>,
 ) -> DocumentsBatchReader<std::io::Cursor<Vec<u8>>> {
@@ -166,7 +192,7 @@ pub fn documents_batch_reader_from_objects(
 mod test {
     use std::io::Cursor;
 
-    use serde_json::{json, Value};
+    use serde_json::json;
 
     use super::*;
 
@@ -220,20 +246,6 @@ mod test {
         assert!(documents.next_document().unwrap().is_none());
     }
 
-    #[test]
-    fn test_nested() {
-        let docs_reader = documents!([{
-            "hello": {
-                "toto": ["hello"]
-            }
-        }]);
-
-        let (mut cursor, _) = docs_reader.into_cursor_and_fields_index();
-        let doc = cursor.next_document().unwrap().unwrap();
-        let nested: Value = serde_json::from_slice(doc.get(0).unwrap()).unwrap();
-        assert_eq!(nested, json!({ "toto": ["hello"] }));
-    }
-
     #[test]
     fn out_of_order_json_fields() {
         let _documents = documents!([

crates/milli/src/documents/primary_key.rs

@@ -1,8 +1,14 @@
 use std::iter;
+use std::ops::ControlFlow;
 use std::result::Result as StdResult;
 
+use bumpalo::Bump;
+use serde_json::value::RawValue;
 use serde_json::Value;
 
+use crate::fields_ids_map::MutFieldIdMapper;
+use crate::update::new::indexer::de::{match_component, DeOrBumpStr};
+use crate::update::new::KvReaderFieldId;
 use crate::{FieldId, InternalError, Object, Result, UserError};
 
 /// The symbol used to define levels in a nested primary key.
@@ -17,6 +23,21 @@ pub trait FieldIdMapper {
     ///
     /// `None` if the field with this name was not found.
     fn id(&self, name: &str) -> Option<FieldId>;
+
+    fn name(&self, id: FieldId) -> Option<&str>;
+}
+
+impl<T> FieldIdMapper for &T
+where
+    T: FieldIdMapper,
+{
+    fn id(&self, name: &str) -> Option<FieldId> {
+        T::id(self, name)
+    }
+
+    fn name(&self, id: FieldId) -> Option<&str> {
+        T::name(self, id)
+    }
 }
 
 /// A type that represent the type of primary key that has been set
@@ -43,7 +64,19 @@ impl<'a> PrimaryKey<'a> {
         })
     }
 
-    pub fn name(&self) -> &str {
+    pub fn new_or_insert(
+        path: &'a str,
+        fields: &mut impl MutFieldIdMapper,
+    ) -> StdResult<Self, UserError> {
+        Ok(if path.contains(PRIMARY_KEY_SPLIT_SYMBOL) {
+            Self::Nested { name: path }
+        } else {
+            let field_id = fields.insert(path).ok_or(UserError::AttributeLimitReached)?;
+            Self::Flat { name: path, field_id }
+        })
+    }
+
+    pub fn name(&self) -> &'a str {
         match self {
             PrimaryKey::Flat { name, .. } => name,
             PrimaryKey::Nested { name } => name,
@@ -52,7 +85,7 @@ impl<'a> PrimaryKey<'a> {
 
     pub fn document_id(
         &self,
-        document: &obkv::KvReader<'_, FieldId>,
+        document: &obkv::KvReader<FieldId>,
         fields: &impl FieldIdMapper,
     ) -> Result<StdResult<String, DocumentIdExtractionError>> {
         match self {
@@ -100,9 +133,105 @@ impl<'a> PrimaryKey<'a> {
         }
     }
 
+    pub fn extract_docid_from_db<'pl, 'bump: 'pl, Mapper: FieldIdMapper>(
+        &self,
+        document: &'pl KvReaderFieldId,
+        db_fields_ids_map: &Mapper,
+        indexer: &'bump Bump,
+    ) -> Result<DeOrBumpStr<'pl, 'bump>> {
+        use serde::Deserializer as _;
+
+        match self {
+            PrimaryKey::Flat { name: _, field_id } => {
+                let Some(document_id) = document.get(*field_id) else {
+                    return Err(InternalError::DocumentsError(
+                        crate::documents::Error::InvalidDocumentFormat,
+                    )
+                    .into());
+                };
+
+                let document_id: &RawValue =
+                    serde_json::from_slice(document_id).map_err(InternalError::SerdeJson)?;
+
+                let document_id = document_id
+                    .deserialize_any(crate::update::new::indexer::de::DocumentIdVisitor(indexer))
+                    .map_err(InternalError::SerdeJson)?;
+
+                let external_document_id = match document_id {
+                    Ok(document_id) => Ok(document_id),
+                    Err(_) => Err(InternalError::DocumentsError(
+                        crate::documents::Error::InvalidDocumentFormat,
+                    )),
+                }?;
+
+                Ok(external_document_id)
+            }
+            nested @ PrimaryKey::Nested { name: _ } => {
+                let mut docid = None;
+                for (first_level, right) in nested.possible_level_names() {
+                    let Some(fid) = db_fields_ids_map.id(first_level) else { continue };
+
+                    let Some(value) = document.get(fid) else { continue };
+                    let value: &RawValue =
+                        serde_json::from_slice(value).map_err(InternalError::SerdeJson)?;
+                    match match_component(first_level, right, value, indexer, &mut docid) {
+                        ControlFlow::Continue(()) => continue,
+                        ControlFlow::Break(Ok(_)) => {
+                            return Err(InternalError::DocumentsError(
+                                crate::documents::Error::InvalidDocumentFormat,
+                            )
+                            .into())
+                        }
+                        ControlFlow::Break(Err(err)) => {
+                            return Err(InternalError::SerdeJson(err).into())
+                        }
+                    }
+                }
+                Ok(docid.ok_or(InternalError::DocumentsError(
+                    crate::documents::Error::InvalidDocumentFormat,
+                ))?)
+            }
+        }
+    }
+
+    pub fn extract_fields_and_docid<'pl, 'bump: 'pl, Mapper: MutFieldIdMapper>(
+        &self,
+        document: &'pl RawValue,
+        new_fields_ids_map: &mut Mapper,
+        indexer: &'bump Bump,
+    ) -> Result<DeOrBumpStr<'pl, 'bump>> {
+        use serde::Deserializer as _;
+        let res = document
+            .deserialize_map(crate::update::new::indexer::de::FieldAndDocidExtractor::new(
+                new_fields_ids_map,
+                self,
+                indexer,
+            ))
+            .map_err(UserError::SerdeJson)??;
+
+        let external_document_id = match res {
+            Ok(document_id) => Ok(document_id),
+            Err(DocumentIdExtractionError::InvalidDocumentId(e)) => Err(e),
+            Err(DocumentIdExtractionError::MissingDocumentId) => {
+                Err(UserError::MissingDocumentId {
+                    primary_key: self.name().to_string(),
+                    document: serde_json::from_str(document.get()).unwrap(),
+                })
+            }
+            Err(DocumentIdExtractionError::TooManyDocumentIds(_)) => {
+                Err(UserError::TooManyDocumentIds {
+                    primary_key: self.name().to_string(),
+                    document: serde_json::from_str(document.get()).unwrap(),
+                })
+            }
+        }?;
+
+        Ok(external_document_id)
+    }
+
     /// Returns an `Iterator` that gives all the possible fields names the primary key
     /// can have depending of the first level name and depth of the objects.
-    pub fn possible_level_names(&self) -> impl Iterator<Item = (&str, &str)> + '_ {
+    pub fn possible_level_names(&self) -> impl Iterator<Item = (&'a str, &'a str)> + '_ {
         let name = self.name();
         name.match_indices(PRIMARY_KEY_SPLIT_SYMBOL)
             .map(move |(i, _)| (&name[..i], &name[i + PRIMARY_KEY_SPLIT_SYMBOL.len_utf8()..]))
@@ -149,7 +278,7 @@ fn starts_with(selector: &str, key: &str) -> bool {
 
 // FIXME: move to a DocumentId struct
 
-fn validate_document_id(document_id: &str) -> Option<&str> {
+pub fn validate_document_id_str(document_id: &str) -> Option<&str> {
     if document_id.is_empty()
         || document_id.len() > 512
         || !document_id.chars().all(|c| c.is_ascii_alphanumeric() || c == '-' || c == '_')
@@ -162,7 +291,7 @@ fn validate_document_id(document_id: &str) -> Option<&str> {
 
 pub fn validate_document_id_value(document_id: Value) -> StdResult<String, UserError> {
     match document_id {
-        Value::String(string) => match validate_document_id(&string) {
+        Value::String(string) => match validate_document_id_str(&string) {
             Some(s) if s.len() == string.len() => Ok(string),
             Some(s) => Ok(s.to_string()),
             None => Err(UserError::InvalidDocumentId { document_id: Value::String(string) }),

crates/milli/src/documents/reader.rs

@@ -72,15 +72,24 @@ impl<R> DocumentsBatchCursor<R> {
 }
 
 impl<R: io::Read + io::Seek> DocumentsBatchCursor<R> {
+    /// Returns a single document from the database.
+    pub fn get(
+        &mut self,
+        offset: u32,
+    ) -> Result<Option<&KvReader<FieldId>>, DocumentsBatchCursorError> {
+        match self.cursor.move_on_key_equal_to(offset.to_be_bytes())? {
+            Some((key, value)) if key != DOCUMENTS_BATCH_INDEX_KEY => Ok(Some(value.into())),
+            _otherwise => Ok(None),
+        }
+    }
+
     /// Returns the next document, starting from the first one. Subsequent calls to
     /// `next_document` advance the document reader until all the documents have been read.
     pub fn next_document(
         &mut self,
-    ) -> Result<Option<KvReader<'_, FieldId>>, DocumentsBatchCursorError> {
+    ) -> Result<Option<&KvReader<FieldId>>, DocumentsBatchCursorError> {
         match self.cursor.move_on_next()? {
-            Some((key, value)) if key != DOCUMENTS_BATCH_INDEX_KEY => {
-                Ok(Some(KvReader::new(value)))
-            }
+            Some((key, value)) if key != DOCUMENTS_BATCH_INDEX_KEY => Ok(Some(value.into())),
             _otherwise => Ok(None),
         }
     }

crates/milli/src/error.rs

@@ -31,23 +31,23 @@ pub enum Error {
 pub enum InternalError {
     #[error("{}", HeedError::DatabaseClosing)]
     DatabaseClosing,
-    #[error("Missing {} in the {db_name} database.", key.unwrap_or("key"))]
+    #[error("missing {} in the {db_name} database", key.unwrap_or("key"))]
     DatabaseMissingEntry { db_name: &'static str, key: Option<&'static str> },
-    #[error("Missing {key} in the fieldids weights mapping.")]
+    #[error("missing {key} in the fieldids weights mapping")]
     FieldidsWeightsMapMissingEntry { key: FieldId },
     #[error(transparent)]
     FieldIdMapMissingEntry(#[from] FieldIdMapMissingEntry),
-    #[error("Missing {key} in the field id mapping.")]
+    #[error("missing {key} in the field id mapping")]
     FieldIdMappingMissingEntry { key: FieldId },
     #[error(transparent)]
     Fst(#[from] fst::Error),
     #[error(transparent)]
     DocumentsError(#[from] documents::Error),
-    #[error("Invalid compression type have been specified to grenad")]
+    #[error("invalid compression type have been specified to grenad")]
     GrenadInvalidCompressionType,
-    #[error("Invalid grenad file with an invalid version format")]
+    #[error("invalid grenad file with an invalid version format")]
     GrenadInvalidFormatVersion,
-    #[error("Invalid merge while processing {process}")]
+    #[error("invalid merge while processing {process}")]
     IndexingMergingKeys { process: &'static str },
     #[error(transparent)]
     RayonThreadPool(#[from] ThreadPoolBuildError),
@@ -56,6 +56,8 @@ pub enum InternalError {
     #[error(transparent)]
     SerdeJson(#[from] serde_json::Error),
     #[error(transparent)]
+    BincodeError(#[from] bincode::Error),
+    #[error(transparent)]
     Serialization(#[from] SerializationError),
     #[error(transparent)]
     Store(#[from] MdbError),
@@ -122,7 +124,7 @@ and can not be more than 512 bytes.", .document_id.to_string()
     #[error("The `_vectors` field in the document with id: `{document_id}` is not an object. Was expecting an object with a key for each embedder with manually provided vectors, but instead got `{value}`")]
     InvalidVectorsMapType { document_id: String, value: Value },
     #[error("Bad embedder configuration in the document with id: `{document_id}`. {error}")]
-    InvalidVectorsEmbedderConf { document_id: String, error: deserr::errors::JsonError },
+    InvalidVectorsEmbedderConf { document_id: String, error: String },
     #[error("{0}")]
     InvalidFilter(String),
     #[error("Invalid type for filter subexpression: expected: {}, found: {1}.", .0.join(", "))]

crates/milli/src/fields_ids_map.rs

@@ -4,6 +4,10 @@ use serde::{Deserialize, Serialize};
 
 use crate::FieldId;
 
+mod global;
+pub mod metadata;
+pub use global::GlobalFieldsIdsMap;
+
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct FieldsIdsMap {
     names_ids: BTreeMap<String, FieldId>,
@@ -95,6 +99,20 @@ impl crate::documents::FieldIdMapper for FieldsIdsMap {
     fn id(&self, name: &str) -> Option<FieldId> {
         self.id(name)
     }
+
+    fn name(&self, id: FieldId) -> Option<&str> {
+        self.name(id)
+    }
+}
+
+pub trait MutFieldIdMapper {
+    fn insert(&mut self, name: &str) -> Option<FieldId>;
+}
+
+impl MutFieldIdMapper for FieldsIdsMap {
+    fn insert(&mut self, name: &str) -> Option<FieldId> {
+        self.insert(name)
+    }
 }
 
 #[cfg(test)]

crates/milli/src/fields_ids_map/global.rs

@@ -0,0 +1,129 @@
+use std::collections::BTreeMap;
+use std::sync::RwLock;
+
+use super::metadata::{FieldIdMapWithMetadata, Metadata};
+use super::MutFieldIdMapper;
+use crate::documents::FieldIdMapper;
+use crate::FieldId;
+
+/// A fields ids map that can be globally updated to add fields
+#[derive(Debug, Clone)]
+pub struct GlobalFieldsIdsMap<'indexing> {
+    global: &'indexing RwLock<FieldIdMapWithMetadata>,
+    local: LocalFieldsIdsMap,
+}
+
+#[derive(Debug, Clone)]
+pub struct LocalFieldsIdsMap {
+    names_ids: BTreeMap<String, FieldId>,
+    ids_names: BTreeMap<FieldId, String>,
+    metadata: BTreeMap<FieldId, Metadata>,
+}
+
+impl FieldIdMapper for LocalFieldsIdsMap {
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.id(name)
+    }
+
+    fn name(&self, id: FieldId) -> Option<&str> {
+        self.name(id)
+    }
+}
+
+impl LocalFieldsIdsMap {
+    fn new(global: &RwLock<FieldIdMapWithMetadata>) -> Self {
+        let global = global.read().unwrap();
+        Self {
+            names_ids: global.as_fields_ids_map().names_ids.clone(),
+            ids_names: global.as_fields_ids_map().ids_names.clone(),
+            metadata: global.iter_id_metadata().collect(),
+        }
+    }
+
+    fn insert(&mut self, name: &str, field_id: FieldId, metadata: Metadata) {
+        self.names_ids.insert(name.to_owned(), field_id);
+        self.ids_names.insert(field_id, name.to_owned());
+        self.metadata.insert(field_id, metadata);
+    }
+
+    fn name(&self, id: FieldId) -> Option<&str> {
+        self.ids_names.get(&id).map(String::as_str)
+    }
+
+    fn id(&self, name: &str) -> Option<FieldId> {
+        self.names_ids.get(name).copied()
+    }
+
+    fn id_with_metadata(&self, name: &str) -> Option<(FieldId, Metadata)> {
+        let id = self.id(name)?;
+        Some((id, self.metadata(id).unwrap()))
+    }
+
+    fn metadata(&self, id: FieldId) -> Option<Metadata> {
+        self.metadata.get(&id).copied()
+    }
+}
+
+impl<'indexing> GlobalFieldsIdsMap<'indexing> {
+    pub fn new(global: &'indexing RwLock<FieldIdMapWithMetadata>) -> Self {
+        Self { local: LocalFieldsIdsMap::new(global), global }
+    }
+
+    /// Returns the field id related to a field name, it will create a new field id if the
+    /// name is not already known. Returns `None` if the maximum field id as been reached.
+    pub fn id_or_insert(&mut self, name: &str) -> Option<FieldId> {
+        self.id_with_metadata_or_insert(name).map(|(fid, _meta)| fid)
+    }
+
+    pub fn id_with_metadata_or_insert(&mut self, name: &str) -> Option<(FieldId, Metadata)> {
+        if let Some(entry) = self.local.id_with_metadata(name) {
+            return Some(entry);
+        }
+
+        {
+            // optimistically lookup the global map
+            let global = self.global.read().unwrap();
+
+            if let Some((field_id, metadata)) = global.id_with_metadata(name) {
+                self.local.insert(name, field_id, metadata);
+                return Some((field_id, metadata));
+            }
+        }
+
+        {
+            let mut global = self.global.write().unwrap();
+
+            if let Some((field_id, metadata)) = global.id_with_metadata(name) {
+                self.local.insert(name, field_id, metadata);
+                return Some((field_id, metadata));
+            }
+
+            let field_id = global.insert(name)?;
+            let metadata = global.metadata(field_id).unwrap();
+            self.local.insert(name, field_id, metadata);
+            Some((field_id, metadata))
+        }
+    }
+
+    /// Get the name of a field based on its id.
+    pub fn name(&mut self, id: FieldId) -> Option<&str> {
+        if self.local.name(id).is_none() {
+            let global = self.global.read().unwrap();
+
+            let (name, metadata) = global.name_with_metadata(id)?;
+            self.local.insert(name, id, metadata);
+        }
+
+        self.local.name(id)
+    }
+
+    pub fn local_map(&self) -> &LocalFieldsIdsMap {
+        &self.local
+    }
+}
+
+impl<'indexing> MutFieldIdMapper for GlobalFieldsIdsMap<'indexing> {
+    fn insert(&mut self, name: &str) -> Option<FieldId> {
+        self.id_or_insert(name)
+    }
+}

crates/milli/src/fields_ids_map/metadata.rs

@@ -0,0 +1,183 @@
+use std::collections::{BTreeMap, HashSet};
+use std::num::NonZeroU16;
+
+use charabia::Language;
+use heed::RoTxn;
+
+use super::FieldsIdsMap;
+use crate::{FieldId, Index, LocalizedAttributesRule, Result};
+
+#[derive(Debug, Clone, Copy)]
+pub struct Metadata {
+    pub searchable: bool,
+    pub filterable: bool,
+    pub sortable: bool,
+    localized_attributes_rule_id: Option<NonZeroU16>,
+}
+
+#[derive(Debug, Clone)]
+pub struct FieldIdMapWithMetadata {
+    fields_ids_map: FieldsIdsMap,
+    builder: MetadataBuilder,
+    metadata: BTreeMap<FieldId, Metadata>,
+}
+
+impl FieldIdMapWithMetadata {
+    pub fn new(existing_fields_ids_map: FieldsIdsMap, builder: MetadataBuilder) -> Self {
+        let metadata = existing_fields_ids_map
+            .iter()
+            .map(|(id, name)| (id, builder.metadata_for_field(name)))
+            .collect();
+        Self { fields_ids_map: existing_fields_ids_map, builder, metadata }
+    }
+
+    pub fn as_fields_ids_map(&self) -> &FieldsIdsMap {
+        &self.fields_ids_map
+    }
+
+    /// Returns the number of fields ids in the map.
+    pub fn len(&self) -> usize {
+        self.fields_ids_map.len()
+    }
+
+    /// Returns `true` if the map is empty.
+    pub fn is_empty(&self) -> bool {
+        self.fields_ids_map.is_empty()
+    }
+
+    /// Returns the field id related to a field name, it will create a new field id if the
+    /// name is not already known. Returns `None` if the maximum field id as been reached.
+    pub fn insert(&mut self, name: &str) -> Option<FieldId> {
+        let id = self.fields_ids_map.insert(name)?;
+        self.metadata.insert(id, self.builder.metadata_for_field(name));
+        Some(id)
+    }
+
+    /// Get the id of a field based on its name.
+    pub fn id(&self, name: &str) -> Option<FieldId> {
+        self.fields_ids_map.id(name)
+    }
+
+    pub fn id_with_metadata(&self, name: &str) -> Option<(FieldId, Metadata)> {
+        let id = self.fields_ids_map.id(name)?;
+        Some((id, self.metadata(id).unwrap()))
+    }
+
+    /// Get the name of a field based on its id.
+    pub fn name(&self, id: FieldId) -> Option<&str> {
+        self.fields_ids_map.name(id)
+    }
+
+    /// Get the name of a field based on its id.
+    pub fn name_with_metadata(&self, id: FieldId) -> Option<(&str, Metadata)> {
+        let name = self.fields_ids_map.name(id)?;
+        Some((name, self.metadata(id).unwrap()))
+    }
+
+    pub fn metadata(&self, id: FieldId) -> Option<Metadata> {
+        self.metadata.get(&id).copied()
+    }
+
+    /// Iterate over the ids and names in the ids order.
+    pub fn iter(&self) -> impl Iterator<Item = (FieldId, &str, Metadata)> {
+        self.fields_ids_map.iter().map(|(id, name)| (id, name, self.metadata(id).unwrap()))
+    }
+
+    pub fn iter_id_metadata(&self) -> impl Iterator<Item = (FieldId, Metadata)> + '_ {
+        self.metadata.iter().map(|(k, v)| (*k, *v))
+    }
+
+    pub fn iter_metadata(&self) -> impl Iterator<Item = Metadata> + '_ {
+        self.metadata.values().copied()
+    }
+
+    pub fn metadata_builder(&self) -> &MetadataBuilder {
+        &self.builder
+    }
+}
+
+impl Metadata {
+    pub fn locales<'rules>(
+        &self,
+        rules: &'rules [LocalizedAttributesRule],
+    ) -> Option<&'rules [Language]> {
+        let localized_attributes_rule_id = self.localized_attributes_rule_id?.get();
+        let rule = rules.get((localized_attributes_rule_id - 1) as usize).unwrap();
+        Some(rule.locales())
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct MetadataBuilder {
+    searchable_attributes: Vec<String>,
+    filterable_attributes: HashSet<String>,
+    sortable_attributes: HashSet<String>,
+    localized_attributes: Option<Vec<LocalizedAttributesRule>>,
+}
+
+impl MetadataBuilder {
+    pub fn from_index(index: &Index, rtxn: &RoTxn) -> Result<Self> {
+        let searchable_attributes =
+            index.searchable_fields(rtxn)?.into_iter().map(|s| s.to_string()).collect();
+        let filterable_attributes = index.filterable_fields(rtxn)?;
+        let sortable_attributes = index.sortable_fields(rtxn)?;
+        let localized_attributes = index.localized_attributes_rules(rtxn)?;
+
+        Ok(Self {
+            searchable_attributes,
+            filterable_attributes,
+            sortable_attributes,
+            localized_attributes,
+        })
+    }
+
+    pub fn new(
+        searchable_attributes: Vec<String>,
+        filterable_attributes: HashSet<String>,
+        sortable_attributes: HashSet<String>,
+        localized_attributes: Option<Vec<LocalizedAttributesRule>>,
+    ) -> Self {
+        Self {
+            searchable_attributes,
+            filterable_attributes,
+            sortable_attributes,
+            localized_attributes,
+        }
+    }
+
+    pub fn metadata_for_field(&self, field: &str) -> Metadata {
+        let searchable = self
+            .searchable_attributes
+            .iter()
+            .any(|attribute| attribute == "*" || attribute == field);
+
+        let filterable = self.filterable_attributes.contains(field);
+
+        let sortable = self.sortable_attributes.contains(field);
+
+        let localized_attributes_rule_id = self
+            .localized_attributes
+            .iter()
+            .flat_map(|v| v.iter())
+            .position(|rule| rule.match_str(field))
+            .map(|id| NonZeroU16::new(id.saturating_add(1).try_into().unwrap()).unwrap());
+
+        Metadata { searchable, filterable, sortable, localized_attributes_rule_id }
+    }
+
+    pub fn searchable_attributes(&self) -> &[String] {
+        self.searchable_attributes.as_slice()
+    }
+
+    pub fn sortable_attributes(&self) -> &HashSet<String> {
+        &self.sortable_attributes
+    }
+
+    pub fn filterable_attributes(&self) -> &HashSet<String> {
+        &self.filterable_attributes
+    }
+
+    pub fn localized_attributes_rules(&self) -> Option<&[LocalizedAttributesRule]> {
+        self.localized_attributes.as_deref()
+    }
+}

crates/milli/src/heed_codec/facet/ordered_f64_codec.rs

@@ -27,17 +27,34 @@ impl heed::BytesEncode<'_> for OrderedF64Codec {
     fn bytes_encode(f: &Self::EItem) -> Result<Cow<'_, [u8]>, BoxedError> {
         let mut buffer = [0u8; 16];
 
-        // write the globally ordered float
-        let bytes = f64_into_bytes(*f).ok_or(InvalidGloballyOrderedFloatError { float: *f })?;
-        buffer[..8].copy_from_slice(&bytes[..]);
-        // Then the f64 value just to be able to read it back
-        let bytes = f.to_be_bytes();
-        buffer[8..16].copy_from_slice(&bytes[..]);
+        encode_f64_into_ordered_bytes(*f, &mut buffer)?;
 
         Ok(Cow::Owned(buffer.to_vec()))
     }
 }
 
+impl OrderedF64Codec {
+    pub fn serialize_into(
+        f: f64,
+        buffer: &mut [u8; 16],
+    ) -> Result<(), InvalidGloballyOrderedFloatError> {
+        encode_f64_into_ordered_bytes(f, buffer)
+    }
+}
+
+fn encode_f64_into_ordered_bytes(
+    f: f64,
+    buffer: &mut [u8; 16],
+) -> Result<(), InvalidGloballyOrderedFloatError> {
+    let bytes = f64_into_bytes(f).ok_or(InvalidGloballyOrderedFloatError { float: f })?;
+    buffer[..8].copy_from_slice(&bytes[..]);
+    // Then the f64 value just to be able to read it back
+    let bytes = f.to_be_bytes();
+    buffer[8..16].copy_from_slice(&bytes[..]);
+
+    Ok(())
+}
+
 #[derive(Error, Debug)]
 #[error("the float {float} cannot be converted to a globally ordered representation")]
 pub struct InvalidGloballyOrderedFloatError {

crates/milli/src/heed_codec/obkv_codec.rs

@@ -6,10 +6,10 @@ use obkv::{KvReaderU16, KvWriterU16};
 pub struct ObkvCodec;
 
 impl<'a> heed::BytesDecode<'a> for ObkvCodec {
-    type DItem = KvReaderU16<'a>;
+    type DItem = &'a KvReaderU16;
 
     fn bytes_decode(bytes: &'a [u8]) -> Result<Self::DItem, BoxedError> {
-        Ok(KvReaderU16::new(bytes))
+        Ok(KvReaderU16::from_slice(bytes))
     }
 }
 

crates/milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs

@@ -122,7 +122,7 @@ impl CboRoaringBitmapCodec {
 
     /// Merges a DelAdd delta into a CboRoaringBitmap.
     pub fn merge_deladd_into<'a>(
-        deladd: KvReaderDelAdd<'_>,
+        deladd: &KvReaderDelAdd,
         previous: &[u8],
         buffer: &'a mut Vec<u8>,
     ) -> io::Result<Option<&'a [u8]>> {

crates/milli/src/index.rs

@@ -1251,12 +1251,20 @@ impl Index {
 
     /* documents */
 
+    /// Returns a document by using the document id.
+    pub fn document<'t>(&self, rtxn: &'t RoTxn, id: DocumentId) -> Result<&'t obkv::KvReaderU16> {
+        self.documents
+            .get(rtxn, &id)?
+            .ok_or(UserError::UnknownInternalDocumentId { document_id: id })
+            .map_err(Into::into)
+    }
+
     /// Returns an iterator over the requested documents. The next item will be an error if a document is missing.
     pub fn iter_documents<'a, 't: 'a>(
         &'a self,
         rtxn: &'t RoTxn<'t>,
         ids: impl IntoIterator<Item = DocumentId> + 'a,
-    ) -> Result<impl Iterator<Item = Result<(DocumentId, obkv::KvReaderU16<'t>)>> + 'a> {
+    ) -> Result<impl Iterator<Item = Result<(DocumentId, &'t obkv::KvReaderU16)>> + 'a> {
         Ok(ids.into_iter().map(move |id| {
             let kv = self
                 .documents
@@ -1271,7 +1279,7 @@ impl Index {
         &self,
         rtxn: &'t RoTxn<'t>,
         ids: impl IntoIterator<Item = DocumentId>,
-    ) -> Result<Vec<(DocumentId, obkv::KvReaderU16<'t>)>> {
+    ) -> Result<Vec<(DocumentId, &'t obkv::KvReaderU16)>> {
         self.iter_documents(rtxn, ids)?.collect()
     }
 
@@ -1279,7 +1287,7 @@ impl Index {
     pub fn all_documents<'a, 't: 'a>(
         &'a self,
         rtxn: &'t RoTxn<'t>,
-    ) -> Result<impl Iterator<Item = Result<(DocumentId, obkv::KvReaderU16<'t>)>> + 'a> {
+    ) -> Result<impl Iterator<Item = Result<(DocumentId, &'t obkv::KvReaderU16)>> + 'a> {
         self.iter_documents(rtxn, self.documents_ids(rtxn)?)
     }
 
@@ -1303,7 +1311,7 @@ impl Index {
         })?;
         Ok(self.iter_documents(rtxn, ids)?.map(move |entry| -> Result<_> {
             let (_docid, obkv) = entry?;
-            match primary_key.document_id(&obkv, &fields)? {
+            match primary_key.document_id(obkv, &fields)? {
                 Ok(document_id) => Ok(document_id),
                 Err(_) => Err(InternalError::DocumentsError(
                     crate::documents::Error::InvalidDocumentFormat,
@@ -1638,6 +1646,14 @@ impl Index {
         }
         Ok(res)
     }
+
+    pub fn prefix_settings(&self, _rtxn: &RoTxn<'_>) -> Result<PrefixSettings> {
+        Ok(PrefixSettings {
+            compute_prefixes: true,
+            max_prefix_length: 4,
+            prefix_count_threshold: 100,
+        })
+    }
 }
 
 #[derive(Debug, Deserialize, Serialize)]
@@ -1647,6 +1663,13 @@ pub struct IndexEmbeddingConfig {
     pub user_provided: RoaringBitmap,
 }
 
+#[derive(Debug, Deserialize, Serialize)]
+pub struct PrefixSettings {
+    pub prefix_count_threshold: u64,
+    pub max_prefix_length: usize,
+    pub compute_prefixes: bool,
+}
+
 #[derive(Serialize, Deserialize)]
 #[serde(transparent)]
 struct OffsetDateTime(#[serde(with = "time::serde::rfc3339")] time::OffsetDateTime);
@@ -1657,19 +1680,24 @@ pub(crate) mod tests {
     use std::ops::Deref;
 
     use big_s::S;
+    use bumpalo::Bump;
     use heed::{EnvOpenOptions, RwTxn};
     use maplit::{btreemap, hashset};
+    use memmap2::Mmap;
     use tempfile::TempDir;
 
-    use crate::documents::DocumentsBatchReader;
     use crate::error::{Error, InternalError};
     use crate::index::{DEFAULT_MIN_WORD_LEN_ONE_TYPO, DEFAULT_MIN_WORD_LEN_TWO_TYPOS};
+    use crate::update::new::indexer;
+    use crate::update::settings::InnerIndexSettings;
     use crate::update::{
-        self, IndexDocuments, IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Setting,
-        Settings,
+        self, IndexDocumentsConfig, IndexDocumentsMethod, IndexerConfig, Setting, Settings,
     };
     use crate::vector::settings::{EmbedderSource, EmbeddingSettings};
-    use crate::{db_snap, obkv_to_json, Filter, Index, Search, SearchResult};
+    use crate::vector::EmbeddingConfigs;
+    use crate::{
+        db_snap, obkv_to_json, Filter, Index, Search, SearchResult, ThreadPoolNoAbortBuilder,
+    };
 
     pub(crate) struct TempIndex {
         pub inner: Index,
@@ -1702,35 +1730,67 @@ pub(crate) mod tests {
         pub fn new() -> Self {
             Self::new_with_map_size(4096 * 2000)
         }
-        pub fn add_documents_using_wtxn<'t, R>(
+
+        pub fn add_documents_using_wtxn<'t>(
             &'t self,
             wtxn: &mut RwTxn<'t>,
-            documents: DocumentsBatchReader<R>,
-        ) -> Result<(), crate::error::Error>
-        where
-            R: std::io::Read + std::io::Seek,
-        {
-            let builder = IndexDocuments::new(
-                wtxn,
-                self,
-                &self.indexer_config,
-                self.index_documents_config.clone(),
-                |_| (),
-                || false,
-            )
-            .unwrap();
-            let (builder, user_error) = builder.add_documents(documents).unwrap();
-            user_error?;
-            builder.execute()?;
+            documents: Mmap,
+        ) -> Result<(), crate::error::Error> {
+            let local_pool;
+            let indexer_config = &self.indexer_config;
+            let pool = match &indexer_config.thread_pool {
+                Some(pool) => pool,
+                None => {
+                    local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                    &local_pool
+                }
+            };
+
+            let rtxn = self.inner.read_txn()?;
+            let db_fields_ids_map = self.inner.fields_ids_map(&rtxn)?;
+            let mut new_fields_ids_map = db_fields_ids_map.clone();
+
+            let embedders =
+                InnerIndexSettings::from_index(&self.inner, &rtxn, None)?.embedding_configs;
+            let mut indexer =
+                indexer::DocumentOperation::new(self.index_documents_config.update_method);
+            indexer.add_documents(&documents).unwrap();
+
+            let indexer_alloc = Bump::new();
+            let (document_changes, operation_stats, primary_key) = indexer.into_changes(
+                &indexer_alloc,
+                &self.inner,
+                &rtxn,
+                None,
+                &mut new_fields_ids_map,
+                &|| false,
+                &|_progress| (),
+            )?;
+
+            if let Some(error) = operation_stats.into_iter().find_map(|stat| stat.error) {
+                return Err(error.into());
+            }
+
+            pool.install(|| {
+                indexer::index(
+                    wtxn,
+                    &self.inner,
+                    indexer_config.grenad_parameters(),
+                    &db_fields_ids_map,
+                    new_fields_ids_map,
+                    primary_key,
+                    &document_changes,
+                    embedders,
+                    &|| false,
+                    &|_| (),
+                )
+            })
+            .unwrap()?;
+
             Ok(())
         }
-        pub fn add_documents<R>(
-            &self,
-            documents: DocumentsBatchReader<R>,
-        ) -> Result<(), crate::error::Error>
-        where
-            R: std::io::Read + std::io::Seek,
-        {
+
+        pub fn add_documents(&self, documents: Mmap) -> Result<(), crate::error::Error> {
             let mut wtxn = self.write_txn().unwrap();
             self.add_documents_using_wtxn(&mut wtxn, documents)?;
             wtxn.commit().unwrap();
@@ -1746,6 +1806,7 @@ pub(crate) mod tests {
             wtxn.commit().unwrap();
             Ok(())
         }
+
         pub fn update_settings_using_wtxn<'t>(
             &'t self,
             wtxn: &mut RwTxn<'t>,
@@ -1761,25 +1822,68 @@ pub(crate) mod tests {
             &'t self,
             wtxn: &mut RwTxn<'t>,
             external_document_ids: Vec<String>,
-        ) {
-            let builder = IndexDocuments::new(
-                wtxn,
-                self,
-                &self.indexer_config,
-                self.index_documents_config.clone(),
-                |_| (),
-                || false,
-            )
-            .unwrap();
-            let (builder, user_error) = builder.remove_documents(external_document_ids).unwrap();
-            user_error.unwrap();
-            builder.execute().unwrap();
+        ) -> Result<(), crate::error::Error> {
+            let local_pool;
+            let indexer_config = &self.indexer_config;
+            let pool = match &indexer_config.thread_pool {
+                Some(pool) => pool,
+                None => {
+                    local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                    &local_pool
+                }
+            };
+
+            let rtxn = self.inner.read_txn()?;
+            let db_fields_ids_map = self.inner.fields_ids_map(&rtxn)?;
+            let mut new_fields_ids_map = db_fields_ids_map.clone();
+
+            let embedders =
+                InnerIndexSettings::from_index(&self.inner, &rtxn, None)?.embedding_configs;
+
+            let mut indexer =
+                indexer::DocumentOperation::new(self.index_documents_config.update_method);
+            let external_document_ids: Vec<_> =
+                external_document_ids.iter().map(AsRef::as_ref).collect();
+            indexer.delete_documents(external_document_ids.as_slice());
+
+            let indexer_alloc = Bump::new();
+            let (document_changes, operation_stats, primary_key) = indexer.into_changes(
+                &indexer_alloc,
+                &self.inner,
+                &rtxn,
+                None,
+                &mut new_fields_ids_map,
+                &|| false,
+                &|_progress| (),
+            )?;
+
+            if let Some(error) = operation_stats.into_iter().find_map(|stat| stat.error) {
+                return Err(error.into());
+            }
+
+            pool.install(|| {
+                indexer::index(
+                    wtxn,
+                    &self.inner,
+                    indexer_config.grenad_parameters(),
+                    &db_fields_ids_map,
+                    new_fields_ids_map,
+                    primary_key,
+                    &document_changes,
+                    embedders,
+                    &|| false,
+                    &|_| (),
+                )
+            })
+            .unwrap()?;
+
+            Ok(())
         }
 
         pub fn delete_documents(&self, external_document_ids: Vec<String>) {
             let mut wtxn = self.write_txn().unwrap();
 
-            self.delete_documents_using_wtxn(&mut wtxn, external_document_ids);
+            self.delete_documents_using_wtxn(&mut wtxn, external_document_ids).unwrap();
 
             wtxn.commit().unwrap();
         }
@@ -1796,29 +1900,63 @@ pub(crate) mod tests {
 
         let index = TempIndex::new();
         let mut wtxn = index.inner.write_txn().unwrap();
-
         let should_abort = AtomicBool::new(false);
-        let builder = IndexDocuments::new(
-            &mut wtxn,
-            &index.inner,
-            &index.indexer_config,
-            index.index_documents_config.clone(),
-            |_| (),
-            || should_abort.load(Relaxed),
-        )
-        .unwrap();
 
-        let (builder, user_error) = builder
-            .add_documents(documents!([
-                { "id": 1, "name": "kevin" },
-                { "id": 2, "name": "bob", "age": 20 },
-                { "id": 2, "name": "bob", "age": 20 },
-            ]))
+        let local_pool;
+        let indexer_config = &index.indexer_config;
+        let pool = match &indexer_config.thread_pool {
+            Some(pool) => pool,
+            None => {
+                local_pool = ThreadPoolNoAbortBuilder::new().build().unwrap();
+                &local_pool
+            }
+        };
+
+        let rtxn = index.inner.read_txn().unwrap();
+        let db_fields_ids_map = index.inner.fields_ids_map(&rtxn).unwrap();
+        let mut new_fields_ids_map = db_fields_ids_map.clone();
+
+        let embedders = EmbeddingConfigs::default();
+        let mut indexer = indexer::DocumentOperation::new(IndexDocumentsMethod::ReplaceDocuments);
+        let payload = documents!([
+            { "id": 1, "name": "kevin" },
+            { "id": 2, "name": "bob", "age": 20 },
+            { "id": 2, "name": "bob", "age": 20 },
+        ]);
+        indexer.add_documents(&payload).unwrap();
+
+        let indexer_alloc = Bump::new();
+        let (document_changes, _operation_stats, primary_key) = indexer
+            .into_changes(
+                &indexer_alloc,
+                &index.inner,
+                &rtxn,
+                None,
+                &mut new_fields_ids_map,
+                &|| false,
+                &|_progress| (),
+            )
             .unwrap();
-        user_error.unwrap();
 
         should_abort.store(true, Relaxed);
-        let err = builder.execute().unwrap_err();
+
+        let err = pool
+            .install(|| {
+                indexer::index(
+                    &mut wtxn,
+                    &index.inner,
+                    indexer_config.grenad_parameters(),
+                    &db_fields_ids_map,
+                    new_fields_ids_map,
+                    primary_key,
+                    &document_changes,
+                    embedders,
+                    &|| should_abort.load(Relaxed),
+                    &|_| (),
+                )
+            })
+            .unwrap()
+            .unwrap_err();
 
         assert!(matches!(err, Error::InternalError(InternalError::AbortedIndexation)));
     }
@@ -2314,7 +2452,16 @@ pub(crate) mod tests {
 
         // And adding lots of documents afterwards instead of just one.
         // These extra subtests don't add much, but it's better than nothing.
-        index.add_documents(documents!([{ "primary_key": 38 }, { "primary_key": 39 }, { "primary_key": 41 }, { "primary_key": 40 }, { "primary_key": 41 }, { "primary_key": 42 }])).unwrap();
+        index
+            .add_documents(documents!([
+                { "primary_key": 38 },
+                { "primary_key": 39 },
+                { "primary_key": 41 },
+                { "primary_key": 40 },
+                { "primary_key": 41 },
+                { "primary_key": 42 },
+            ]))
+            .unwrap();
 
         db_snap!(index, documents_ids, @"[0, 1, 2, 3, 4, 5, ]");
         db_snap!(index, external_documents_ids, 7, @r###"
@@ -2701,7 +2848,7 @@ pub(crate) mod tests {
                 documents!({ "id" : "doggo", "_geo": { "lat": 1, "lng": 2, "doggo": "are the best" }}),
             )
             .unwrap_err();
-        insta::assert_snapshot!(err, @r###"The `_geo` field in the document with the id: `"\"doggo\""` contains the following unexpected fields: `{"doggo":"are the best"}`."###);
+        insta::assert_snapshot!(err, @r###"The `_geo` field in the document with the id: `"doggo"` contains the following unexpected fields: `{"doggo":"are the best"}`."###);
 
         db_snap!(index, geo_faceted_documents_ids); // ensure that no documents were inserted
 
@@ -2711,7 +2858,7 @@ pub(crate) mod tests {
                 documents!({ "id" : "doggo", "_geo": { "lat": 1, "lng": 2, "doggo": "are the best", "and": { "all": ["cats", { "are": "beautiful" } ] } } }),
             )
             .unwrap_err();
-        insta::assert_snapshot!(err, @r###"The `_geo` field in the document with the id: `"\"doggo\""` contains the following unexpected fields: `{"and":{"all":["cats",{"are":"beautiful"}]},"doggo":"are the best"}`."###);
+        insta::assert_snapshot!(err, @r###"The `_geo` field in the document with the id: `"doggo"` contains the following unexpected fields: `{"and":{"all":["cats",{"are":"beautiful"}]},"doggo":"are the best"}`."###);
 
         db_snap!(index, geo_faceted_documents_ids); // ensure that no documents were inserted
     }
@@ -2830,7 +2977,6 @@ pub(crate) mod tests {
         db_snap!(index, fields_ids_map, @r###"
         0   id               |
         1   _vectors         |
-        2   _vectors.doggo   |
         "###);
         db_snap!(index, searchable_fields, @r###"["id"]"###);
         db_snap!(index, fieldids_weights_map, @r###"

crates/milli/src/lib.rs

@@ -55,7 +55,7 @@ pub use self::error::{
 };
 pub use self::external_documents_ids::ExternalDocumentsIds;
 pub use self::fieldids_weights_map::FieldidsWeightsMap;
-pub use self::fields_ids_map::FieldsIdsMap;
+pub use self::fields_ids_map::{FieldsIdsMap, GlobalFieldsIdsMap};
 pub use self::heed_codec::{
     BEU16StrCodec, BEU32StrCodec, BoRoaringBitmapCodec, BoRoaringBitmapLenCodec,
     CboRoaringBitmapCodec, CboRoaringBitmapLenCodec, FieldIdWordCountCodec, ObkvCodec,
@@ -88,6 +88,7 @@ pub type Object = serde_json::Map<String, serde_json::Value>;
 pub type Position = u32;
 pub type RelativePosition = u16;
 pub type SmallString32 = smallstr::SmallString<[u8; 32]>;
+pub type Prefix = smallstr::SmallString<[u8; 16]>;
 pub type SmallVec16<T> = smallvec::SmallVec<[T; 16]>;
 pub type SmallVec32<T> = smallvec::SmallVec<[T; 32]>;
 pub type SmallVec8<T> = smallvec::SmallVec<[T; 8]>;
@@ -214,7 +215,7 @@ pub fn bucketed_position(relative: u16) -> u16 {
 pub fn obkv_to_json(
     displayed_fields: &[FieldId],
     fields_ids_map: &FieldsIdsMap,
-    obkv: obkv::KvReaderU16<'_>,
+    obkv: &obkv::KvReaderU16,
 ) -> Result<Object> {
     displayed_fields
         .iter()
@@ -232,10 +233,7 @@ pub fn obkv_to_json(
 }
 
 /// Transform every field of a raw obkv store into a JSON Object.
-pub fn all_obkv_to_json(
-    obkv: obkv::KvReaderU16<'_>,
-    fields_ids_map: &FieldsIdsMap,
-) -> Result<Object> {
+pub fn all_obkv_to_json(obkv: &obkv::KvReaderU16, fields_ids_map: &FieldsIdsMap) -> Result<Object> {
     let all_keys = obkv.iter().map(|(k, _v)| k).collect::<Vec<_>>();
     obkv_to_json(all_keys.as_slice(), fields_ids_map, obkv)
 }
@@ -434,7 +432,7 @@ mod tests {
         writer.insert(id1, b"1234").unwrap();
         writer.insert(id2, b"4321").unwrap();
         let contents = writer.into_inner().unwrap();
-        let obkv = obkv::KvReaderU16::new(&contents);
+        let obkv = obkv::KvReaderU16::from_slice(&contents);
 
         let expected = json!({
             "field1": 1234,

crates/milli/src/prompt/context.rs

@@ -3,23 +3,19 @@ use liquid::model::{
 };
 use liquid::{ObjectView, ValueView};
 
-use super::document::Document;
-use super::fields::Fields;
-use super::FieldsIdsMapWithMetadata;
-
 #[derive(Debug, Clone)]
-pub struct Context<'a> {
-    document: &'a Document<'a>,
-    fields: Fields<'a>,
+pub struct Context<'a, D: ObjectView, F: ArrayView> {
+    document: &'a D,
+    fields: &'a F,
 }
 
-impl<'a> Context<'a> {
-    pub fn new(document: &'a Document<'a>, field_id_map: &'a FieldsIdsMapWithMetadata<'a>) -> Self {
-        Self { document, fields: Fields::new(document, field_id_map) }
+impl<'a, D: ObjectView, F: ArrayView> Context<'a, D, F> {
+    pub fn new(document: &'a D, fields: &'a F) -> Self {
+        Self { document, fields }
     }
 }
 
-impl<'a> ObjectView for Context<'a> {
+impl<'a, D: ObjectView, F: ArrayView> ObjectView for Context<'a, D, F> {
     fn as_value(&self) -> &dyn ValueView {
         self
     }
@@ -56,7 +52,7 @@ impl<'a> ObjectView for Context<'a> {
     }
 }
 
-impl<'a> ValueView for Context<'a> {
+impl<'a, D: ObjectView, F: ArrayView> ValueView for Context<'a, D, F> {
     fn as_debug(&self) -> &dyn std::fmt::Debug {
         self
     }

crates/milli/src/prompt/document.rs

@@ -1,10 +1,15 @@
 use std::cell::OnceCell;
 use std::collections::BTreeMap;
+use std::fmt::{self, Debug};
 
+use bumpalo::Bump;
 use liquid::model::{
-    DisplayCow, KString, KStringCow, ObjectRender, ObjectSource, State, Value as LiquidValue,
+    ArrayView, DisplayCow, KString, KStringCow, ObjectRender, ObjectSource, ScalarCow, State,
+    Value as LiquidValue,
 };
 use liquid::{ObjectView, ValueView};
+use raw_collections::{RawMap, RawVec};
+use serde_json::value::RawValue;
 
 use crate::update::del_add::{DelAdd, KvReaderDelAdd};
 use crate::FieldsIdsMap;
@@ -30,13 +35,13 @@ impl ParsedValue {
 
 impl<'a> Document<'a> {
     pub fn new(
-        data: obkv::KvReaderU16<'a>,
+        data: &'a obkv::KvReaderU16,
         side: DelAdd,
         inverted_field_map: &'a FieldsIdsMap,
     ) -> Self {
         let mut out_data = BTreeMap::new();
         for (fid, raw) in data {
-            let obkv = KvReaderDelAdd::new(raw);
+            let obkv = KvReaderDelAdd::from_slice(raw);
             let Some(raw) = obkv.get(side) else {
                 continue;
             };
@@ -93,7 +98,7 @@ impl<'a> ObjectView for Document<'a> {
 }
 
 impl<'a> ValueView for Document<'a> {
-    fn as_debug(&self) -> &dyn std::fmt::Debug {
+    fn as_debug(&self) -> &dyn Debug {
         self
     }
 
@@ -128,4 +133,515 @@ impl<'a> ValueView for Document<'a> {
     fn as_object(&self) -> Option<&dyn ObjectView> {
         Some(self)
     }
+
+    fn is_object(&self) -> bool {
+        true
+    }
+}
+
+/// Implementation for any type that implements the Document trait
+use crate::update::new::document::Document as DocumentTrait;
+
+#[derive(Debug)]
+pub struct ParseableDocument<'doc, D> {
+    document: D,
+    doc_alloc: &'doc Bump,
+}
+
+impl<'doc, D> ParseableDocument<'doc, D> {
+    pub fn new(document: D, doc_alloc: &'doc Bump) -> Self {
+        Self { document, doc_alloc }
+    }
+}
+
+impl<'doc, D: DocumentTrait<'doc> + Debug> ObjectView for ParseableDocument<'doc, D> {
+    fn as_value(&self) -> &dyn ValueView {
+        self
+    }
+
+    fn size(&self) -> i64 {
+        self.document.len() as i64
+    }
+
+    fn keys<'k>(&'k self) -> Box<dyn Iterator<Item = KStringCow<'k>> + 'k> {
+        Box::new(self.document.iter_top_level_fields().map(|res| {
+            let (field, _) = res.unwrap();
+            KStringCow::from_ref(field)
+        }))
+    }
+
+    fn values<'k>(&'k self) -> Box<dyn Iterator<Item = &'k dyn ValueView> + 'k> {
+        Box::new(self.document.iter_top_level_fields().map(|res| {
+            let (_, value) = res.unwrap();
+            ParseableValue::new_bump(value, self.doc_alloc) as _
+        }))
+    }
+
+    fn iter<'k>(&'k self) -> Box<dyn Iterator<Item = (KStringCow<'k>, &'k dyn ValueView)> + 'k> {
+        Box::new(self.document.iter_top_level_fields().map(|res| {
+            let (field, value) = res.unwrap();
+            (KStringCow::from_ref(field), ParseableValue::new_bump(value, self.doc_alloc) as _)
+        }))
+    }
+
+    fn contains_key(&self, index: &str) -> bool {
+        self.document.top_level_field(index).unwrap().is_some()
+    }
+
+    fn get<'s>(&'s self, index: &str) -> Option<&'s dyn ValueView> {
+        let s = self.document.top_level_field(index).unwrap()?;
+        Some(ParseableValue::new_bump(s, self.doc_alloc))
+    }
+}
+
+impl<'doc, D: DocumentTrait<'doc> + Debug> ValueView for ParseableDocument<'doc, D> {
+    fn as_debug(&self) -> &dyn fmt::Debug {
+        self
+    }
+    fn render(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ObjectRender::new(self)))
+    }
+
+    fn source(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ObjectSource::new(self)))
+    }
+
+    fn type_name(&self) -> &'static str {
+        "object"
+    }
+
+    fn query_state(&self, state: liquid::model::State) -> bool {
+        match state {
+            State::Truthy => true,
+            State::DefaultValue | State::Empty | State::Blank => false,
+        }
+    }
+
+    fn to_kstr(&self) -> liquid::model::KStringCow<'_> {
+        let s = ObjectRender::new(self).to_string();
+        KStringCow::from_string(s)
+    }
+
+    fn to_value(&self) -> LiquidValue {
+        LiquidValue::Object(
+            self.document
+                .iter_top_level_fields()
+                .map(|res| {
+                    let (k, v) = res.unwrap();
+                    (k.to_string().into(), ParseableValue::new(v, self.doc_alloc).to_value())
+                })
+                .collect(),
+        )
+    }
+
+    fn as_object(&self) -> Option<&dyn ObjectView> {
+        Some(self)
+    }
+
+    fn is_object(&self) -> bool {
+        true
+    }
+}
+
+#[derive(Debug)]
+struct ParseableValue<'doc> {
+    value: raw_collections::Value<'doc>,
+}
+
+impl<'doc> ParseableValue<'doc> {
+    pub fn new(value: &'doc RawValue, doc_alloc: &'doc Bump) -> Self {
+        let value = raw_collections::Value::from_raw_value(value, doc_alloc).unwrap();
+        Self { value }
+    }
+
+    pub fn new_bump(value: &'doc RawValue, doc_alloc: &'doc Bump) -> &'doc Self {
+        doc_alloc.alloc(Self::new(value, doc_alloc))
+    }
+}
+
+// transparent newtype for implementing ValueView
+#[repr(transparent)]
+#[derive(Debug)]
+struct ParseableMap<'doc>(RawMap<'doc>);
+
+// transparent newtype for implementing ValueView
+#[repr(transparent)]
+#[derive(Debug)]
+struct ParseableArray<'doc>(RawVec<'doc>);
+
+impl<'doc> ParseableMap<'doc> {
+    pub fn as_parseable<'a>(map: &'a RawMap<'doc>) -> &'a ParseableMap<'doc> {
+        // SAFETY: repr(transparent)
+        unsafe { &*(map as *const RawMap as *const Self) }
+    }
+}
+
+impl<'doc> ParseableArray<'doc> {
+    pub fn as_parseable<'a>(array: &'a RawVec<'doc>) -> &'a ParseableArray<'doc> {
+        // SAFETY: repr(transparent)
+        unsafe { &*(array as *const RawVec as *const Self) }
+    }
+}
+
+impl<'doc> ArrayView for ParseableArray<'doc> {
+    fn as_value(&self) -> &dyn ValueView {
+        self
+    }
+
+    fn size(&self) -> i64 {
+        self.0.len() as _
+    }
+
+    fn values<'k>(&'k self) -> Box<dyn Iterator<Item = &'k dyn ValueView> + 'k> {
+        Box::new(self.0.iter().map(|v| ParseableValue::new_bump(v, self.0.bump()) as _))
+    }
+
+    fn contains_key(&self, index: i64) -> bool {
+        let index = convert_index(index, self.size());
+        index < self.size() && index >= 0
+    }
+
+    fn get(&self, index: i64) -> Option<&dyn ValueView> {
+        let index = convert_index(index, self.size());
+        if index <= 0 {
+            return None;
+        }
+        let v = self.0.get(index as usize)?;
+        Some(ParseableValue::new_bump(v, self.0.bump()))
+    }
+}
+
+impl<'doc> ValueView for ParseableArray<'doc> {
+    fn as_debug(&self) -> &dyn std::fmt::Debug {
+        self
+    }
+
+    fn render(&self) -> DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ArrayRender { s: &self.0 }))
+    }
+
+    fn source(&self) -> DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ArraySource { s: &self.0 }))
+    }
+
+    fn type_name(&self) -> &'static str {
+        "array"
+    }
+
+    fn query_state(&self, state: State) -> bool {
+        match state {
+            State::Truthy => true,
+            State::DefaultValue | State::Empty | State::Blank => self.0.is_empty(),
+        }
+    }
+
+    fn to_kstr(&self) -> KStringCow<'_> {
+        let s = ArrayRender { s: &self.0 }.to_string();
+        KStringCow::from_string(s)
+    }
+
+    fn to_value(&self) -> LiquidValue {
+        LiquidValue::Array(self.values().map(|v| v.to_value()).collect())
+    }
+
+    fn is_array(&self) -> bool {
+        true
+    }
+
+    fn as_array(&self) -> Option<&dyn ArrayView> {
+        Some(self as _)
+    }
+}
+
+impl<'doc> ObjectView for ParseableMap<'doc> {
+    fn as_value(&self) -> &dyn ValueView {
+        self
+    }
+
+    fn size(&self) -> i64 {
+        self.0.len() as i64
+    }
+
+    fn keys<'k>(&'k self) -> Box<dyn Iterator<Item = KStringCow<'k>> + 'k> {
+        Box::new(self.0.keys().map(Into::into))
+    }
+
+    fn values<'k>(&'k self) -> Box<dyn Iterator<Item = &'k dyn ValueView> + 'k> {
+        Box::new(self.0.values().map(|value| {
+            let doc_alloc = self.0.bump();
+            ParseableValue::new_bump(value, doc_alloc) as _
+        }))
+    }
+
+    fn iter<'k>(&'k self) -> Box<dyn Iterator<Item = (KStringCow<'k>, &'k dyn ValueView)> + 'k> {
+        Box::new(self.0.iter().map(|(k, v)| {
+            let doc_alloc = self.0.bump();
+            (k.into(), ParseableValue::new_bump(v, doc_alloc) as _)
+        }))
+    }
+
+    fn contains_key(&self, index: &str) -> bool {
+        self.0.get(index).is_some()
+    }
+
+    fn get<'s>(&'s self, index: &str) -> Option<&'s dyn ValueView> {
+        let v = self.0.get(index)?;
+        let doc_alloc = self.0.bump();
+        let value = ParseableValue::new(v, doc_alloc);
+        Some(doc_alloc.alloc(value) as _)
+    }
+}
+
+impl<'doc> ValueView for ParseableMap<'doc> {
+    fn as_debug(&self) -> &dyn std::fmt::Debug {
+        self
+    }
+
+    fn render(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ObjectRender::new(self)))
+    }
+
+    fn source(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ObjectSource::new(self)))
+    }
+
+    fn type_name(&self) -> &'static str {
+        "object"
+    }
+
+    fn query_state(&self, state: liquid::model::State) -> bool {
+        match state {
+            State::Truthy => true,
+            State::DefaultValue | State::Empty | State::Blank => self.0.is_empty(),
+        }
+    }
+
+    fn to_kstr(&self) -> liquid::model::KStringCow<'_> {
+        let s = ObjectRender::new(self).to_string();
+        KStringCow::from_string(s)
+    }
+
+    fn to_value(&self) -> LiquidValue {
+        LiquidValue::Object(
+            self.0
+                .iter()
+                .map(|(k, v)| {
+                    (k.to_string().into(), ParseableValue::new(v, self.0.bump()).to_value())
+                })
+                .collect(),
+        )
+    }
+
+    fn as_object(&self) -> Option<&dyn ObjectView> {
+        Some(self)
+    }
+
+    fn is_object(&self) -> bool {
+        true
+    }
+}
+
+impl<'doc> ValueView for ParseableValue<'doc> {
+    fn as_debug(&self) -> &dyn Debug {
+        self
+    }
+
+    fn render(&self) -> DisplayCow<'_> {
+        use raw_collections::value::Number;
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => LiquidValue::Nil.render(),
+            Value::Bool(v) => v.render(),
+            Value::Number(number) => match number {
+                Number::PosInt(x) => DisplayCow::Borrowed(x),
+                Number::NegInt(x) => x.render(),
+                Number::Finite(x) => x.render(),
+            },
+            Value::String(s) => s.render(),
+            Value::Array(raw_vec) => ParseableArray::as_parseable(raw_vec).render(),
+            Value::Object(raw_map) => ParseableMap::as_parseable(raw_map).render(),
+        }
+    }
+
+    fn source(&self) -> DisplayCow<'_> {
+        use raw_collections::value::Number;
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => LiquidValue::Nil.source(),
+            Value::Bool(v) => ValueView::source(v),
+            Value::Number(number) => match number {
+                Number::PosInt(x) => DisplayCow::Borrowed(x),
+                Number::NegInt(x) => x.source(),
+                Number::Finite(x) => x.source(),
+            },
+            Value::String(s) => s.source(),
+            Value::Array(raw_vec) => ParseableArray::as_parseable(raw_vec).source(),
+            Value::Object(raw_map) => ParseableMap::as_parseable(raw_map).source(),
+        }
+    }
+
+    fn type_name(&self) -> &'static str {
+        use raw_collections::value::Number;
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => LiquidValue::Nil.type_name(),
+            Value::Bool(v) => v.type_name(),
+            Value::Number(number) => match number {
+                Number::PosInt(_x) => "whole positive number",
+                Number::NegInt(x) => x.type_name(),
+                Number::Finite(x) => x.type_name(),
+            },
+            Value::String(s) => s.type_name(),
+            Value::Array(_raw_vec) => "array",
+            Value::Object(_raw_map) => "object",
+        }
+    }
+
+    fn query_state(&self, state: State) -> bool {
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => ValueView::query_state(&LiquidValue::Nil, state),
+            Value::Bool(v) => ValueView::query_state(v, state),
+            Value::Number(_number) => match state {
+                State::Truthy => true,
+                State::DefaultValue => false,
+                State::Empty => false,
+                State::Blank => false,
+            },
+            Value::String(s) => ValueView::query_state(s, state),
+            Value::Array(raw_vec) => ParseableArray::as_parseable(raw_vec).query_state(state),
+            Value::Object(raw_map) => ParseableMap::as_parseable(raw_map).query_state(state),
+        }
+    }
+
+    fn to_kstr(&self) -> KStringCow<'_> {
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => ValueView::to_kstr(&LiquidValue::Nil),
+            Value::Bool(v) => ValueView::to_kstr(v),
+            Value::Number(_number) => self.render().to_string().into(),
+            Value::String(s) => KStringCow::from_ref(s),
+            Value::Array(raw_vec) => ParseableArray::as_parseable(raw_vec).to_kstr(),
+            Value::Object(raw_map) => ParseableMap::as_parseable(raw_map).to_kstr(),
+        }
+    }
+
+    fn to_value(&self) -> LiquidValue {
+        use raw_collections::Value;
+        match &self.value {
+            Value::Null => LiquidValue::Nil,
+            Value::Bool(v) => LiquidValue::Scalar(liquid::model::ScalarCow::new(*v)),
+            Value::Number(number) => match number {
+                raw_collections::value::Number::PosInt(number) => {
+                    let number: i64 = match (*number).try_into() {
+                        Ok(number) => number,
+                        Err(_) => {
+                            return LiquidValue::Scalar(ScalarCow::new(self.render().to_string()))
+                        }
+                    };
+                    LiquidValue::Scalar(ScalarCow::new(number))
+                }
+                raw_collections::value::Number::NegInt(number) => {
+                    LiquidValue::Scalar(ScalarCow::new(*number))
+                }
+                raw_collections::value::Number::Finite(number) => {
+                    LiquidValue::Scalar(ScalarCow::new(*number))
+                }
+            },
+            Value::String(s) => LiquidValue::Scalar(liquid::model::ScalarCow::new(s.to_string())),
+            Value::Array(raw_vec) => ParseableArray::as_parseable(raw_vec).to_value(),
+            Value::Object(raw_map) => ParseableMap::as_parseable(raw_map).to_value(),
+        }
+    }
+
+    fn as_scalar(&self) -> Option<liquid::model::ScalarCow<'_>> {
+        use raw_collections::value::Number;
+        use raw_collections::Value;
+        match &self.value {
+            Value::Bool(v) => Some(liquid::model::ScalarCow::new(*v)),
+            Value::Number(number) => match number {
+                Number::PosInt(number) => {
+                    let number: i64 = match (*number).try_into() {
+                        Ok(number) => number,
+                        Err(_) => return Some(ScalarCow::new(self.render().to_string())),
+                    };
+                    Some(ScalarCow::new(number))
+                }
+                Number::NegInt(number) => Some(ScalarCow::new(*number)),
+                Number::Finite(number) => Some(ScalarCow::new(*number)),
+            },
+            Value::String(s) => Some(ScalarCow::new(*s)),
+            _ => None,
+        }
+    }
+
+    fn is_scalar(&self) -> bool {
+        use raw_collections::Value;
+        matches!(&self.value, Value::Bool(_) | Value::Number(_) | Value::String(_))
+    }
+
+    fn as_array(&self) -> Option<&dyn liquid::model::ArrayView> {
+        if let raw_collections::Value::Array(array) = &self.value {
+            return Some(ParseableArray::as_parseable(array) as _);
+        }
+        None
+    }
+
+    fn is_array(&self) -> bool {
+        matches!(&self.value, raw_collections::Value::Array(_))
+    }
+
+    fn as_object(&self) -> Option<&dyn ObjectView> {
+        if let raw_collections::Value::Object(object) = &self.value {
+            return Some(ParseableMap::as_parseable(object) as _);
+        }
+        None
+    }
+
+    fn is_object(&self) -> bool {
+        matches!(&self.value, raw_collections::Value::Object(_))
+    }
+
+    fn is_nil(&self) -> bool {
+        matches!(&self.value, raw_collections::Value::Null)
+    }
+}
+
+struct ArraySource<'s, 'doc> {
+    s: &'s RawVec<'doc>,
+}
+
+impl<'s, 'doc> fmt::Display for ArraySource<'s, 'doc> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "[")?;
+        for item in self.s {
+            let v = ParseableValue::new(item, self.s.bump());
+            write!(f, "{}, ", v.render())?;
+        }
+        write!(f, "]")?;
+        Ok(())
+    }
+}
+
+struct ArrayRender<'s, 'doc> {
+    s: &'s RawVec<'doc>,
+}
+
+impl<'s, 'doc> fmt::Display for ArrayRender<'s, 'doc> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        for item in self.s {
+            let v = ParseableValue::new(item, self.s.bump());
+
+            write!(f, "{}", v.render())?;
+        }
+        Ok(())
+    }
+}
+
+fn convert_index(index: i64, max_size: i64) -> i64 {
+    if 0 <= index {
+        index
+    } else {
+        max_size + index
+    }
 }

crates/milli/src/prompt/fields.rs

@@ -1,36 +1,23 @@
+use std::cell::RefCell;
+use std::fmt;
+
+use bumpalo::Bump;
 use liquid::model::{
     ArrayView, DisplayCow, KStringCow, ObjectRender, ObjectSource, State, Value as LiquidValue,
 };
 use liquid::{ObjectView, ValueView};
 
-use super::document::Document;
 use super::{FieldMetadata, FieldsIdsMapWithMetadata};
-#[derive(Debug, Clone)]
-pub struct Fields<'a>(Vec<FieldValue<'a>>);
-
-impl<'a> Fields<'a> {
-    pub fn new(document: &'a Document<'a>, field_id_map: &'a FieldsIdsMapWithMetadata<'a>) -> Self {
-        Self(
-            std::iter::repeat(document)
-                .zip(field_id_map.iter())
-                .map(|(document, (fid, name))| FieldValue {
-                    document,
-                    name,
-                    metadata: field_id_map.metadata(fid).unwrap_or_default(),
-                })
-                .collect(),
-        )
-    }
-}
+use crate::GlobalFieldsIdsMap;
 
 #[derive(Debug, Clone, Copy)]
-pub struct FieldValue<'a> {
+pub struct FieldValue<'a, D: ObjectView> {
     name: &'a str,
-    document: &'a Document<'a>,
+    document: &'a D,
     metadata: FieldMetadata,
 }
 
-impl<'a> ValueView for FieldValue<'a> {
+impl<'a, D: ObjectView> ValueView for FieldValue<'a, D> {
     fn as_debug(&self) -> &dyn std::fmt::Debug {
         self
     }
@@ -70,7 +57,7 @@ impl<'a> ValueView for FieldValue<'a> {
     }
 }
 
-impl<'a> FieldValue<'a> {
+impl<'a, D: ObjectView> FieldValue<'a, D> {
     pub fn name(&self) -> &&'a str {
         &self.name
     }
@@ -88,7 +75,7 @@ impl<'a> FieldValue<'a> {
     }
 }
 
-impl<'a> ObjectView for FieldValue<'a> {
+impl<'a, D: ObjectView> ObjectView for FieldValue<'a, D> {
     fn as_value(&self) -> &dyn ValueView {
         self
     }
@@ -127,7 +114,42 @@ impl<'a> ObjectView for FieldValue<'a> {
     }
 }
 
-impl<'a> ArrayView for Fields<'a> {
+#[derive(Debug, Clone)]
+pub struct OwnedFields<'a, D: ObjectView>(Vec<FieldValue<'a, D>>);
+
+#[derive(Debug)]
+pub struct BorrowedFields<'a, 'map, D: ObjectView> {
+    document: &'a D,
+    field_id_map: &'a RefCell<GlobalFieldsIdsMap<'map>>,
+    doc_alloc: &'a Bump,
+}
+
+impl<'a, D: ObjectView> OwnedFields<'a, D> {
+    pub fn new(document: &'a D, field_id_map: &'a FieldsIdsMapWithMetadata<'a>) -> Self {
+        Self(
+            std::iter::repeat(document)
+                .zip(field_id_map.iter())
+                .map(|(document, (fid, name))| FieldValue {
+                    document,
+                    name,
+                    metadata: field_id_map.metadata(fid).unwrap_or_default(),
+                })
+                .collect(),
+        )
+    }
+}
+
+impl<'a, 'map, D: ObjectView> BorrowedFields<'a, 'map, D> {
+    pub fn new(
+        document: &'a D,
+        field_id_map: &'a RefCell<GlobalFieldsIdsMap<'map>>,
+        doc_alloc: &'a Bump,
+    ) -> Self {
+        Self { document, field_id_map, doc_alloc }
+    }
+}
+
+impl<'a, D: ObjectView> ArrayView for OwnedFields<'a, D> {
     fn as_value(&self) -> &dyn ValueView {
         self.0.as_value()
     }
@@ -149,7 +171,91 @@ impl<'a> ArrayView for Fields<'a> {
     }
 }
 
-impl<'a> ValueView for Fields<'a> {
+impl<'a, 'map, D: ObjectView> ArrayView for BorrowedFields<'a, 'map, D> {
+    fn as_value(&self) -> &dyn ValueView {
+        self
+    }
+
+    fn size(&self) -> i64 {
+        self.document.size()
+    }
+
+    fn values<'k>(&'k self) -> Box<dyn Iterator<Item = &'k dyn ValueView> + 'k> {
+        Box::new(self.document.keys().map(|k| {
+            let mut field_id_map = self.field_id_map.borrow_mut();
+            let (_, metadata) = field_id_map.id_with_metadata_or_insert(&k).unwrap();
+            let fv = self.doc_alloc.alloc(FieldValue {
+                name: self.doc_alloc.alloc_str(&k),
+                document: self.document,
+                metadata: FieldMetadata { searchable: metadata.searchable },
+            });
+            fv as _
+        }))
+    }
+
+    fn contains_key(&self, index: i64) -> bool {
+        let index = if index >= 0 { index } else { self.size() + index };
+        index >= 0 && index < self.size()
+    }
+
+    fn get(&self, index: i64) -> Option<&dyn ValueView> {
+        let index = if index >= 0 { index } else { self.size() + index };
+        let index: usize = index.try_into().ok()?;
+        let key = self.document.keys().nth(index)?;
+        let mut field_id_map = self.field_id_map.borrow_mut();
+        let (_, metadata) = field_id_map.id_with_metadata_or_insert(&key)?;
+        let fv = self.doc_alloc.alloc(FieldValue {
+            name: self.doc_alloc.alloc_str(&key),
+            document: self.document,
+            metadata: FieldMetadata { searchable: metadata.searchable },
+        });
+        Some(fv as _)
+    }
+}
+
+impl<'a, 'map, D: ObjectView> ValueView for BorrowedFields<'a, 'map, D> {
+    fn as_debug(&self) -> &dyn std::fmt::Debug {
+        self
+    }
+
+    fn render(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ArrayRender { s: self }))
+    }
+
+    fn source(&self) -> liquid::model::DisplayCow<'_> {
+        DisplayCow::Owned(Box::new(ArraySource { s: self }))
+    }
+
+    fn type_name(&self) -> &'static str {
+        "array"
+    }
+
+    fn query_state(&self, state: liquid::model::State) -> bool {
+        match state {
+            State::Truthy => true,
+            State::DefaultValue | State::Empty | State::Blank => self.document.size() == 0,
+        }
+    }
+
+    fn to_kstr(&self) -> liquid::model::KStringCow<'_> {
+        let s = ArrayRender { s: self }.to_string();
+        KStringCow::from_string(s)
+    }
+
+    fn to_value(&self) -> LiquidValue {
+        LiquidValue::Array(self.values().map(|v| v.to_value()).collect())
+    }
+
+    fn as_array(&self) -> Option<&dyn ArrayView> {
+        Some(self)
+    }
+
+    fn is_array(&self) -> bool {
+        true
+    }
+}
+
+impl<'a, D: ObjectView> ValueView for OwnedFields<'a, D> {
     fn as_debug(&self) -> &dyn std::fmt::Debug {
         self
     }
@@ -182,3 +288,31 @@ impl<'a> ValueView for Fields<'a> {
         Some(self)
     }
 }
+
+struct ArraySource<'a, 'map, D: ObjectView> {
+    s: &'a BorrowedFields<'a, 'map, D>,
+}
+
+impl<'a, 'map, D: ObjectView> fmt::Display for ArraySource<'a, 'map, D> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "[")?;
+        for item in self.s.values() {
+            write!(f, "{}, ", item.render())?;
+        }
+        write!(f, "]")?;
+        Ok(())
+    }
+}
+
+struct ArrayRender<'a, 'map, D: ObjectView> {
+    s: &'a BorrowedFields<'a, 'map, D>,
+}
+
+impl<'a, 'map, D: ObjectView> fmt::Display for ArrayRender<'a, 'map, D> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        for item in self.s.values() {
+            write!(f, "{}", item.render())?;
+        }
+        Ok(())
+    }
+}

crates/milli/src/prompt/mod.rs

@@ -4,17 +4,22 @@ pub(crate) mod error;
 mod fields;
 mod template_checker;
 
+use std::cell::RefCell;
 use std::collections::BTreeMap;
 use std::convert::TryFrom;
+use std::fmt::Debug;
 use std::num::NonZeroUsize;
 use std::ops::Deref;
 
+use bumpalo::Bump;
+use document::ParseableDocument;
 use error::{NewPromptError, RenderPromptError};
+use fields::{BorrowedFields, OwnedFields};
 
 use self::context::Context;
 use self::document::Document;
 use crate::update::del_add::DelAdd;
-use crate::{FieldId, FieldsIdsMap};
+use crate::{FieldId, FieldsIdsMap, GlobalFieldsIdsMap};
 
 pub struct Prompt {
     template: liquid::Template,
@@ -109,14 +114,38 @@ impl Prompt {
         Ok(this)
     }
 
-    pub fn render(
+    pub fn render_document<
+        'a,       // lifetime of the borrow of the document
+        'doc: 'a, // lifetime of the allocator, will live for an entire chunk of documents
+    >(
         &self,
-        document: obkv::KvReaderU16<'_>,
+        document: impl crate::update::new::document::Document<'a> + Debug,
+        field_id_map: &RefCell<GlobalFieldsIdsMap>,
+        doc_alloc: &'doc Bump,
+    ) -> Result<&'doc str, RenderPromptError> {
+        let document = ParseableDocument::new(document, doc_alloc);
+        let fields = BorrowedFields::new(&document, field_id_map, doc_alloc);
+        let context = Context::new(&document, &fields);
+        let mut rendered = bumpalo::collections::Vec::with_capacity_in(
+            self.max_bytes.unwrap_or_else(default_max_bytes).get(),
+            doc_alloc,
+        );
+        self.template
+            .render_to(&mut rendered, &context)
+            .map_err(RenderPromptError::missing_context)?;
+        Ok(std::str::from_utf8(rendered.into_bump_slice())
+            .expect("render can only write UTF-8 because all inputs and processing preserve utf-8"))
+    }
+
+    pub fn render_kvdeladd(
+        &self,
+        document: &obkv::KvReaderU16,
         side: DelAdd,
         field_id_map: &FieldsIdsMapWithMetadata,
     ) -> Result<String, RenderPromptError> {
         let document = Document::new(document, side, field_id_map);
-        let context = Context::new(&document, field_id_map);
+        let fields = OwnedFields::new(&document, field_id_map);
+        let context = Context::new(&document, &fields);
 
         let mut rendered =
             self.template.render(&context).map_err(RenderPromptError::missing_context)?;

crates/milli/src/search/facet/facet_distribution.rs

@@ -407,7 +407,7 @@ mod tests {
     use big_s::S;
     use maplit::hashset;
 
-    use crate::documents::documents_batch_reader_from_objects;
+    use crate::documents::mmap_from_objects;
     use crate::index::tests::TempIndex;
     use crate::{milli_snap, FacetDistribution, OrderBy};
 
@@ -416,17 +416,16 @@ mod tests {
         // All the tests here avoid using the code in `facet_distribution_iter` because there aren't
         // enough candidates.
 
-        let mut index = TempIndex::new();
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new();
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
             .unwrap();
 
         let documents = documents!([
-            { "colour": "Blue" },
-            { "colour": "  blue" },
-            { "colour": "RED" }
+            { "id": 0, "colour": "Blue" },
+            { "id": 1, "colour": "  blue" },
+            { "id": 2, "colour": "RED" }
         ]);
 
         index.add_documents(documents).unwrap();
@@ -488,8 +487,7 @@ mod tests {
 
     #[test]
     fn many_candidates_few_facet_values() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -500,6 +498,7 @@ mod tests {
         let mut documents = vec![];
         for i in 0..10_000 {
             let document = serde_json::json!({
+                "id": i,
                 "colour": facet_values[i % 5],
             })
             .as_object()
@@ -508,8 +507,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();
@@ -574,8 +572,7 @@ mod tests {
 
     #[test]
     fn many_candidates_many_facet_values() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -586,6 +583,7 @@ mod tests {
         let mut documents = vec![];
         for i in 0..10_000 {
             let document = serde_json::json!({
+                "id": i,
                 "colour": facet_values[i % 1000],
             })
             .as_object()
@@ -594,8 +592,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();
@@ -634,8 +631,7 @@ mod tests {
 
     #[test]
     fn facet_stats() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -646,6 +642,7 @@ mod tests {
         let mut documents = vec![];
         for i in 0..1000 {
             let document = serde_json::json!({
+                "id": i,
                 "colour": facet_values[i % 1000],
             })
             .as_object()
@@ -654,8 +651,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();
@@ -686,8 +682,7 @@ mod tests {
 
     #[test]
     fn facet_stats_array() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -698,6 +693,7 @@ mod tests {
         let mut documents = vec![];
         for i in 0..1000 {
             let document = serde_json::json!({
+                "id": i,
                 "colour": [facet_values[i % 1000], facet_values[i % 1000] + 1000],
             })
             .as_object()
@@ -706,8 +702,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();
@@ -738,8 +733,7 @@ mod tests {
 
     #[test]
     fn facet_stats_mixed_array() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -750,6 +744,7 @@ mod tests {
         let mut documents = vec![];
         for i in 0..1000 {
             let document = serde_json::json!({
+                "id": i,
                 "colour": [facet_values[i % 1000], format!("{}", facet_values[i % 1000] + 1000)],
             })
             .as_object()
@@ -758,8 +753,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();
@@ -790,8 +784,7 @@ mod tests {
 
     #[test]
     fn facet_mixed_values() {
-        let mut index = TempIndex::new_with_map_size(4096 * 10_000);
-        index.index_documents_config.autogenerate_docids = true;
+        let index = TempIndex::new_with_map_size(4096 * 10_000);
 
         index
             .update_settings(|settings| settings.set_filterable_fields(hashset! { S("colour") }))
@@ -803,10 +796,12 @@ mod tests {
         for i in 0..1000 {
             let document = if i % 2 == 0 {
                 serde_json::json!({
+                    "id": i,
                     "colour": [facet_values[i % 1000], facet_values[i % 1000] + 1000],
                 })
             } else {
                 serde_json::json!({
+                    "id": i,
                     "colour": format!("{}", facet_values[i % 1000] + 10000),
                 })
             };
@@ -814,8 +809,7 @@ mod tests {
             documents.push(document);
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
-
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         let txn = index.read_txn().unwrap();

crates/milli/src/search/hybrid.rs

@@ -201,7 +201,9 @@ impl<'a> Search<'a> {
                 let span = tracing::trace_span!(target: "search::hybrid", "embed_one");
                 let _entered = span.enter();
 
-                match embedder.embed_one(query) {
+                let deadline = std::time::Instant::now() + std::time::Duration::from_secs(3);
+
+                match embedder.embed_one(query, Some(deadline)) {
                     Ok(embedding) => embedding,
                     Err(error) => {
                         tracing::error!(error=%error, "Embedding failed");

crates/milli/src/search/new/db_cache.rs

@@ -3,6 +3,7 @@ use std::collections::hash_map::Entry;
 use std::hash::Hash;
 
 use fxhash::FxHashMap;
+use grenad::MergeFunction;
 use heed::types::Bytes;
 use heed::{BytesEncode, Database, RoTxn};
 use roaring::RoaringBitmap;
@@ -11,7 +12,7 @@ use super::interner::Interned;
 use super::Word;
 use crate::heed_codec::{BytesDecodeOwned, StrBEU16Codec};
 use crate::proximity::ProximityPrecision;
-use crate::update::{merge_cbo_roaring_bitmaps, MergeFn};
+use crate::update::MergeCboRoaringBitmaps;
 use crate::{
     CboRoaringBitmapCodec, CboRoaringBitmapLenCodec, Result, SearchContext, U8StrStrCodec,
 };
@@ -110,19 +111,21 @@ impl<'ctx> DatabaseCache<'ctx> {
             .map_err(Into::into)
     }
 
-    fn get_value_from_keys<'v, K1, KC>(
+    fn get_value_from_keys<'v, K1, KC, MF>(
         txn: &'ctx RoTxn<'_>,
         cache_key: K1,
         db_keys: &'v [KC::EItem],
         cache: &mut FxHashMap<K1, Option<Cow<'ctx, [u8]>>>,
         db: Database<KC, Bytes>,
         universe: Option<&RoaringBitmap>,
-        merger: MergeFn,
+        merger: MF,
     ) -> Result<Option<RoaringBitmap>>
     where
         K1: Copy + Eq + Hash,
         KC: BytesEncode<'v>,
         KC::EItem: Sized,
+        MF: MergeFunction,
+        crate::Error: From<MF::Error>,
     {
         if let Entry::Vacant(entry) = cache.entry(cache_key) {
             let bitmap_ptr: Option<Cow<'ctx, [u8]>> = match db_keys {
@@ -138,7 +141,7 @@ impl<'ctx> DatabaseCache<'ctx> {
                     if bitmaps.is_empty() {
                         None
                     } else {
-                        Some(merger(&[], &bitmaps[..])?)
+                        Some(merger.merge(&[], &bitmaps[..])?)
                     }
                 }
             };
@@ -213,17 +216,17 @@ impl<'ctx> SearchContext<'ctx> {
                 let keys: Vec<_> =
                     restricted_fids.tolerant.iter().map(|(fid, _)| (interned, *fid)).collect();
 
-                DatabaseCache::get_value_from_keys::<_, _>(
+                DatabaseCache::get_value_from_keys(
                     self.txn,
                     word,
                     &keys[..],
                     &mut self.db_cache.word_docids,
                     self.index.word_fid_docids.remap_data_type::<Bytes>(),
                     universe,
-                    merge_cbo_roaring_bitmaps,
+                    MergeCboRoaringBitmaps,
                 )
             }
-            None => DatabaseCache::get_value::<_, _>(
+            None => DatabaseCache::get_value(
                 self.txn,
                 word,
                 self.word_interner.get(word).as_str(),
@@ -245,17 +248,17 @@ impl<'ctx> SearchContext<'ctx> {
                 let keys: Vec<_> =
                     restricted_fids.exact.iter().map(|(fid, _)| (interned, *fid)).collect();
 
-                DatabaseCache::get_value_from_keys::<_, _>(
+                DatabaseCache::get_value_from_keys(
                     self.txn,
                     word,
                     &keys[..],
                     &mut self.db_cache.exact_word_docids,
                     self.index.word_fid_docids.remap_data_type::<Bytes>(),
                     universe,
-                    merge_cbo_roaring_bitmaps,
+                    MergeCboRoaringBitmaps,
                 )
             }
-            None => DatabaseCache::get_value::<_, _>(
+            None => DatabaseCache::get_value(
                 self.txn,
                 word,
                 self.word_interner.get(word).as_str(),
@@ -302,17 +305,17 @@ impl<'ctx> SearchContext<'ctx> {
                 let keys: Vec<_> =
                     restricted_fids.tolerant.iter().map(|(fid, _)| (interned, *fid)).collect();
 
-                DatabaseCache::get_value_from_keys::<_, _>(
+                DatabaseCache::get_value_from_keys(
                     self.txn,
                     prefix,
                     &keys[..],
                     &mut self.db_cache.word_prefix_docids,
                     self.index.word_prefix_fid_docids.remap_data_type::<Bytes>(),
                     universe,
-                    merge_cbo_roaring_bitmaps,
+                    MergeCboRoaringBitmaps,
                 )
             }
-            None => DatabaseCache::get_value::<_, _>(
+            None => DatabaseCache::get_value(
                 self.txn,
                 prefix,
                 self.word_interner.get(prefix).as_str(),
@@ -334,17 +337,17 @@ impl<'ctx> SearchContext<'ctx> {
                 let keys: Vec<_> =
                     restricted_fids.exact.iter().map(|(fid, _)| (interned, *fid)).collect();
 
-                DatabaseCache::get_value_from_keys::<_, _>(
+                DatabaseCache::get_value_from_keys(
                     self.txn,
                     prefix,
                     &keys[..],
                     &mut self.db_cache.exact_word_prefix_docids,
                     self.index.word_prefix_fid_docids.remap_data_type::<Bytes>(),
                     universe,
-                    merge_cbo_roaring_bitmaps,
+                    MergeCboRoaringBitmaps,
                 )
             }
-            None => DatabaseCache::get_value::<_, _>(
+            None => DatabaseCache::get_value(
                 self.txn,
                 prefix,
                 self.word_interner.get(prefix).as_str(),
@@ -405,7 +408,7 @@ impl<'ctx> SearchContext<'ctx> {
 
                 Ok(docids)
             }
-            ProximityPrecision::ByWord => DatabaseCache::get_value::<_, _>(
+            ProximityPrecision::ByWord => DatabaseCache::get_value(
                 self.txn,
                 (proximity, word1, word2),
                 &(
@@ -538,7 +541,7 @@ impl<'ctx> SearchContext<'ctx> {
             return Ok(None);
         }
 
-        DatabaseCache::get_value::<_, _>(
+        DatabaseCache::get_value(
             self.txn,
             (word, fid),
             &(self.word_interner.get(word).as_str(), fid),
@@ -559,7 +562,7 @@ impl<'ctx> SearchContext<'ctx> {
             return Ok(None);
         }
 
-        DatabaseCache::get_value::<_, _>(
+        DatabaseCache::get_value(
             self.txn,
             (word_prefix, fid),
             &(self.word_interner.get(word_prefix).as_str(), fid),
@@ -629,7 +632,7 @@ impl<'ctx> SearchContext<'ctx> {
         word: Interned<String>,
         position: u16,
     ) -> Result<Option<RoaringBitmap>> {
-        DatabaseCache::get_value::<_, _>(
+        DatabaseCache::get_value(
             self.txn,
             (word, position),
             &(self.word_interner.get(word).as_str(), position),
@@ -645,7 +648,7 @@ impl<'ctx> SearchContext<'ctx> {
         word_prefix: Interned<String>,
         position: u16,
     ) -> Result<Option<RoaringBitmap>> {
-        DatabaseCache::get_value::<_, _>(
+        DatabaseCache::get_value(
             self.txn,
             (word_prefix, position),
             &(self.word_interner.get(word_prefix).as_str(), position),

crates/milli/src/search/new/matches/mod.rs

@@ -3,6 +3,9 @@ mod r#match;
 mod matching_words;
 mod simple_token_kind;
 
+use std::borrow::Cow;
+use std::cmp::{max, min};
+
 use charabia::{Language, SeparatorKind, Token, Tokenizer};
 use either::Either;
 pub use matching_words::MatchingWords;
@@ -10,10 +13,6 @@ use matching_words::{MatchType, PartialMatch};
 use r#match::{Match, MatchPosition};
 use serde::Serialize;
 use simple_token_kind::SimpleTokenKind;
-use std::{
-    borrow::Cow,
-    cmp::{max, min},
-};
 
 const DEFAULT_CROP_MARKER: &str = "…";
 const DEFAULT_HIGHLIGHT_PREFIX: &str = "<em>";

crates/milli/src/search/new/tests/integration.rs

@@ -1,12 +1,14 @@
-use std::io::Cursor;
+use std::io::Write;
 
 use big_s::S;
+use bumpalo::Bump;
 use heed::EnvOpenOptions;
 use maplit::{btreemap, hashset};
 
-use crate::documents::{DocumentsBatchBuilder, DocumentsBatchReader};
-use crate::update::{IndexDocuments, IndexDocumentsConfig, IndexerConfig, Settings};
-use crate::{db_snap, Criterion, Index, Object};
+use crate::update::new::indexer;
+use crate::update::{IndexDocumentsMethod, IndexerConfig, Settings};
+use crate::vector::EmbeddingConfigs;
+use crate::{db_snap, Criterion, Index};
 pub const CONTENT: &str = include_str!("../../../../tests/assets/test_set.ndjson");
 
 pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
@@ -40,30 +42,60 @@ pub fn setup_search_index_with_criteria(criteria: &[Criterion]) -> Index {
     });
     builder.set_searchable_fields(vec![S("title"), S("description")]);
     builder.execute(|_| (), || false).unwrap();
+    wtxn.commit().unwrap();
 
     // index documents
     let config = IndexerConfig { max_memory: Some(10 * 1024 * 1024), ..Default::default() };
-    let indexing_config = IndexDocumentsConfig::default();
+    let rtxn = index.read_txn().unwrap();
+    let mut wtxn = index.write_txn().unwrap();
 
-    let builder =
-        IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| (), || false).unwrap();
-    let mut documents_builder = DocumentsBatchBuilder::new(Vec::new());
-    let reader = Cursor::new(CONTENT.as_bytes());
+    let db_fields_ids_map = index.fields_ids_map(&rtxn).unwrap();
+    let mut new_fields_ids_map = db_fields_ids_map.clone();
 
-    for result in serde_json::Deserializer::from_reader(reader).into_iter::<Object>() {
-        let object = result.unwrap();
-        documents_builder.append_json_object(&object).unwrap();
-    }
+    let embedders = EmbeddingConfigs::default();
+    let mut indexer = indexer::DocumentOperation::new(IndexDocumentsMethod::ReplaceDocuments);
 
-    let vector = documents_builder.into_inner().unwrap();
+    let mut file = tempfile::tempfile().unwrap();
+    file.write_all(CONTENT.as_bytes()).unwrap();
+    file.sync_all().unwrap();
+    let payload = unsafe { memmap2::Mmap::map(&file).unwrap() };
 
     // index documents
-    let content = DocumentsBatchReader::from_reader(Cursor::new(vector)).unwrap();
-    let (builder, user_error) = builder.add_documents(content).unwrap();
-    user_error.unwrap();
-    builder.execute().unwrap();
+    indexer.add_documents(&payload).unwrap();
+
+    let indexer_alloc = Bump::new();
+    let (document_changes, operation_stats, primary_key) = indexer
+        .into_changes(
+            &indexer_alloc,
+            &index,
+            &rtxn,
+            None,
+            &mut new_fields_ids_map,
+            &|| false,
+            &|_progress| (),
+        )
+        .unwrap();
+
+    if let Some(error) = operation_stats.into_iter().find_map(|stat| stat.error) {
+        panic!("{error}");
+    }
+
+    indexer::index(
+        &mut wtxn,
+        &index,
+        config.grenad_parameters(),
+        &db_fields_ids_map,
+        new_fields_ids_map,
+        primary_key,
+        &document_changes,
+        embedders,
+        &|| false,
+        &|_| (),
+    )
+    .unwrap();
 
     wtxn.commit().unwrap();
+    drop(rtxn);
 
     index
 }

crates/milli/src/update/available_documents_ids.rs

@@ -1,65 +0,0 @@
-use std::iter::{Chain, FromIterator};
-use std::ops::RangeInclusive;
-
-use roaring::bitmap::{IntoIter, RoaringBitmap};
-
-pub struct AvailableDocumentsIds {
-    iter: Chain<IntoIter, RangeInclusive<u32>>,
-}
-
-impl AvailableDocumentsIds {
-    pub fn from_documents_ids(docids: &RoaringBitmap) -> AvailableDocumentsIds {
-        match docids.max() {
-            Some(last_id) => {
-                let mut available = RoaringBitmap::from_iter(0..last_id);
-                available -= docids;
-
-                let iter = match last_id.checked_add(1) {
-                    Some(id) => id..=u32::MAX,
-                    #[allow(clippy::reversed_empty_ranges)]
-                    None => 1..=0, // empty range iterator
-                };
-
-                AvailableDocumentsIds { iter: available.into_iter().chain(iter) }
-            }
-            None => {
-                let empty = RoaringBitmap::new().into_iter();
-                AvailableDocumentsIds { iter: empty.chain(0..=u32::MAX) }
-            }
-        }
-    }
-}
-
-impl Iterator for AvailableDocumentsIds {
-    type Item = u32;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        self.iter.next()
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn empty() {
-        let base = RoaringBitmap::new();
-        let left = AvailableDocumentsIds::from_documents_ids(&base);
-        let right = 0..=u32::MAX;
-        left.zip(right).take(500).for_each(|(l, r)| assert_eq!(l, r));
-    }
-
-    #[test]
-    fn scattered() {
-        let mut base = RoaringBitmap::new();
-        base.insert(0);
-        base.insert(10);
-        base.insert(100);
-        base.insert(405);
-
-        let left = AvailableDocumentsIds::from_documents_ids(&base);
-        let right = (0..=u32::MAX).filter(|&n| n != 0 && n != 10 && n != 100 && n != 405);
-        left.zip(right).take(500).for_each(|(l, r)| assert_eq!(l, r));
-    }
-}

crates/milli/src/update/available_ids.rs

@@ -0,0 +1,65 @@
+use std::iter::{Chain, FromIterator};
+use std::ops::RangeInclusive;
+
+use roaring::bitmap::{IntoIter, RoaringBitmap};
+
+pub struct AvailableIds {
+    iter: Chain<IntoIter, RangeInclusive<u32>>,
+}
+
+impl AvailableIds {
+    pub fn new(docids: &RoaringBitmap) -> AvailableIds {
+        match docids.max() {
+            Some(last_id) => {
+                let mut available = RoaringBitmap::from_iter(0..last_id);
+                available -= docids;
+
+                let iter = match last_id.checked_add(1) {
+                    Some(id) => id..=u32::MAX,
+                    #[allow(clippy::reversed_empty_ranges)]
+                    None => 1..=0, // empty range iterator
+                };
+
+                AvailableIds { iter: available.into_iter().chain(iter) }
+            }
+            None => {
+                let empty = RoaringBitmap::new().into_iter();
+                AvailableIds { iter: empty.chain(0..=u32::MAX) }
+            }
+        }
+    }
+}
+
+impl Iterator for AvailableIds {
+    type Item = u32;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        self.iter.next()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn empty() {
+        let base = RoaringBitmap::new();
+        let left = AvailableIds::new(&base);
+        let right = 0..=u32::MAX;
+        left.zip(right).take(500).for_each(|(l, r)| assert_eq!(l, r));
+    }
+
+    #[test]
+    fn scattered() {
+        let mut base = RoaringBitmap::new();
+        base.insert(0);
+        base.insert(10);
+        base.insert(100);
+        base.insert(405);
+
+        let left = AvailableIds::new(&base);
+        let right = (0..=u32::MAX).filter(|&n| n != 0 && n != 10 && n != 100 && n != 405);
+        left.zip(right).take(500).for_each(|(l, r)| assert_eq!(l, r));
+    }
+}

crates/milli/src/update/concurrent_available_ids.rs

@@ -0,0 +1,59 @@
+use std::sync::atomic::{AtomicBool, AtomicU32, AtomicU64, Ordering};
+
+use roaring::RoaringBitmap;
+
+/// A concurrent ID generate that will never return the same ID twice.
+#[derive(Debug)]
+pub struct ConcurrentAvailableIds {
+    /// The current tree node ID we should use if there is no other IDs available.
+    current: AtomicU32,
+    /// The total number of tree node IDs used.
+    used: AtomicU64,
+
+    /// A list of IDs to exhaust before picking IDs from `current`.
+    available: RoaringBitmap,
+    /// The current Nth ID to select in the bitmap.
+    select_in_bitmap: AtomicU32,
+    /// Tells if you should look in the roaring bitmap or if all the IDs are already exhausted.
+    look_into_bitmap: AtomicBool,
+}
+
+impl ConcurrentAvailableIds {
+    /// Creates an ID generator returning unique IDs, avoiding the specified used IDs.
+    pub fn new(used: RoaringBitmap) -> ConcurrentAvailableIds {
+        let last_id = used.max().map_or(0, |id| id + 1);
+        let used_ids = used.len();
+        let available = RoaringBitmap::from_sorted_iter(0..last_id).unwrap() - used;
+
+        ConcurrentAvailableIds {
+            current: AtomicU32::new(last_id),
+            used: AtomicU64::new(used_ids),
+            select_in_bitmap: AtomicU32::new(0),
+            look_into_bitmap: AtomicBool::new(!available.is_empty()),
+            available,
+        }
+    }
+
+    /// Returns a new unique ID and increase the count of IDs used.
+    pub fn next(&self) -> Option<u32> {
+        if self.used.fetch_add(1, Ordering::Relaxed) > u32::MAX as u64 {
+            None
+        } else if self.look_into_bitmap.load(Ordering::Relaxed) {
+            let current = self.select_in_bitmap.fetch_add(1, Ordering::Relaxed);
+            match self.available.select(current) {
+                Some(id) => Some(id),
+                None => {
+                    self.look_into_bitmap.store(false, Ordering::Relaxed);
+                    Some(self.current.fetch_add(1, Ordering::Relaxed))
+                }
+            }
+        } else {
+            Some(self.current.fetch_add(1, Ordering::Relaxed))
+        }
+    }
+
+    /// Returns the number of used ids in total.
+    pub fn used(&self) -> u64 {
+        self.used.load(Ordering::Relaxed)
+    }
+}

crates/milli/src/update/del_add.rs

@@ -1,7 +1,7 @@
 use obkv::Key;
 
 pub type KvWriterDelAdd<W> = obkv::KvWriter<W, DelAdd>;
-pub type KvReaderDelAdd<'a> = obkv::KvReader<'a, DelAdd>;
+pub type KvReaderDelAdd = obkv::KvReader<DelAdd>;
 
 /// DelAdd defines the new value to add in the database and old value to delete from the database.
 ///
@@ -36,7 +36,7 @@ impl Key for DelAdd {
 /// Addition: put all the values under DelAdd::Addition,
 /// DeletionAndAddition: put all the values under DelAdd::Deletion and DelAdd::Addition,
 pub fn into_del_add_obkv<K: obkv::Key + PartialOrd>(
-    reader: obkv::KvReader<'_, K>,
+    reader: &obkv::KvReader<K>,
     operation: DelAddOperation,
     buffer: &mut Vec<u8>,
 ) -> Result<(), std::io::Error> {
@@ -46,7 +46,7 @@ pub fn into_del_add_obkv<K: obkv::Key + PartialOrd>(
 /// Akin to the [into_del_add_obkv] function but lets you
 /// conditionally define the `DelAdd` variant based on the obkv key.
 pub fn into_del_add_obkv_conditional_operation<K, F>(
-    reader: obkv::KvReader<'_, K>,
+    reader: &obkv::KvReader<K>,
     buffer: &mut Vec<u8>,
     operation: F,
 ) -> std::io::Result<()>
@@ -86,8 +86,8 @@ pub enum DelAddOperation {
 /// putting each deletion obkv's keys under an DelAdd::Deletion
 /// and putting each addition obkv's keys under an DelAdd::Addition
 pub fn del_add_from_two_obkvs<K: obkv::Key + PartialOrd + Ord>(
-    deletion: &obkv::KvReader<'_, K>,
-    addition: &obkv::KvReader<'_, K>,
+    deletion: &obkv::KvReader<K>,
+    addition: &obkv::KvReader<K>,
     buffer: &mut Vec<u8>,
 ) -> Result<(), std::io::Error> {
     use itertools::merge_join_by;
@@ -121,7 +121,7 @@ pub fn del_add_from_two_obkvs<K: obkv::Key + PartialOrd + Ord>(
     writer.finish()
 }
 
-pub fn is_noop_del_add_obkv(del_add: KvReaderDelAdd<'_>) -> bool {
+pub fn is_noop_del_add_obkv(del_add: &KvReaderDelAdd) -> bool {
     del_add.get(DelAdd::Deletion) == del_add.get(DelAdd::Addition)
 }
 
@@ -136,5 +136,5 @@ pub fn deladd_serialize_add_side<'a>(
     obkv: &'a [u8],
     _buffer: &mut Vec<u8>,
 ) -> crate::Result<&'a [u8]> {
-    Ok(KvReaderDelAdd::new(obkv).get(DelAdd::Addition).unwrap_or_default())
+    Ok(KvReaderDelAdd::from_slice(obkv).get(DelAdd::Addition).unwrap_or_default())
 }

crates/milli/src/update/facet/bulk.rs

@@ -14,7 +14,7 @@ use crate::heed_codec::facet::{
 use crate::heed_codec::BytesRefCodec;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd};
 use crate::update::index_documents::{create_writer, valid_lmdb_key, writer_into_reader};
-use crate::update::MergeFn;
+use crate::update::MergeDeladdCboRoaringBitmaps;
 use crate::{CboRoaringBitmapCodec, CboRoaringBitmapLenCodec, FieldId, Index, Result};
 
 /// Algorithm to insert elememts into the `facet_id_(string/f64)_docids` databases
@@ -29,7 +29,7 @@ pub struct FacetsUpdateBulk<'i> {
     facet_type: FacetType,
     field_ids: Vec<FieldId>,
     // None if level 0 does not need to be updated
-    delta_data: Option<Merger<BufReader<File>, MergeFn>>,
+    delta_data: Option<Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>>,
 }
 
 impl<'i> FacetsUpdateBulk<'i> {
@@ -37,7 +37,7 @@ impl<'i> FacetsUpdateBulk<'i> {
         index: &'i Index,
         field_ids: Vec<FieldId>,
         facet_type: FacetType,
-        delta_data: Merger<BufReader<File>, MergeFn>,
+        delta_data: Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>,
         group_size: u8,
         min_level_size: u8,
     ) -> FacetsUpdateBulk<'i> {
@@ -90,7 +90,7 @@ impl<'i> FacetsUpdateBulk<'i> {
 /// Implementation of `FacetsUpdateBulk` that is independent of milli's `Index` type
 pub(crate) struct FacetsUpdateBulkInner<R: std::io::Read + std::io::Seek> {
     pub db: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
-    pub delta_data: Option<Merger<R, MergeFn>>,
+    pub delta_data: Option<Merger<R, MergeDeladdCboRoaringBitmaps>>,
     pub group_size: u8,
     pub min_level_size: u8,
 }
@@ -135,7 +135,7 @@ impl<R: std::io::Read + std::io::Seek> FacetsUpdateBulkInner<R> {
                 if !valid_lmdb_key(key) {
                     continue;
                 }
-                let value = KvReaderDelAdd::new(value);
+                let value = KvReaderDelAdd::from_slice(value);
 
                 // DB is empty, it is safe to ignore Del operations
                 let Some(value) = value.get(DelAdd::Addition) else {
@@ -161,7 +161,7 @@ impl<R: std::io::Read + std::io::Seek> FacetsUpdateBulkInner<R> {
                     continue;
                 }
 
-                let value = KvReaderDelAdd::new(value);
+                let value = KvReaderDelAdd::from_slice(value);
 
                 // the value is a CboRoaringBitmap, but I still need to prepend the
                 // group size for level 0 (= 1) to it
@@ -369,7 +369,7 @@ mod tests {
     use maplit::hashset;
     use roaring::RoaringBitmap;
 
-    use crate::documents::documents_batch_reader_from_objects;
+    use crate::documents::mmap_from_objects;
     use crate::heed_codec::facet::OrderedF64Codec;
     use crate::heed_codec::StrRefCodec;
     use crate::index::tests::TempIndex;
@@ -492,7 +492,7 @@ mod tests {
             );
         }
 
-        let documents = documents_batch_reader_from_objects(documents);
+        let documents = mmap_from_objects(documents);
         index.add_documents(documents).unwrap();
 
         db_snap!(index, facet_id_f64_docids, "initial", @"c34f499261f3510d862fa0283bbe843a");

crates/milli/src/update/facet/incremental.rs

@@ -15,7 +15,7 @@ use crate::heed_codec::BytesRefCodec;
 use crate::search::facet::get_highest_level;
 use crate::update::del_add::DelAdd;
 use crate::update::index_documents::valid_lmdb_key;
-use crate::update::MergeFn;
+use crate::update::MergeDeladdCboRoaringBitmaps;
 use crate::{CboRoaringBitmapCodec, Index, Result};
 
 /// Enum used as a return value for the facet incremental indexing.
@@ -57,14 +57,14 @@ enum ModificationResult {
 /// `facet_id_(string/f64)_docids` databases.
 pub struct FacetsUpdateIncremental {
     inner: FacetsUpdateIncrementalInner,
-    delta_data: Merger<BufReader<File>, MergeFn>,
+    delta_data: Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>,
 }
 
 impl FacetsUpdateIncremental {
     pub fn new(
         index: &Index,
         facet_type: FacetType,
-        delta_data: Merger<BufReader<File>, MergeFn>,
+        delta_data: Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>,
         group_size: u8,
         min_level_size: u8,
         max_group_size: u8,
@@ -109,7 +109,7 @@ impl FacetsUpdateIncremental {
             }
             current_field_id = Some(key.field_id);
 
-            let value = KvReader::new(value);
+            let value = KvReader::from_slice(value);
             let docids_to_delete = value
                 .get(DelAdd::Deletion)
                 .map(CboRoaringBitmapCodec::bytes_decode)

crates/milli/src/update/facet/mod.rs

@@ -86,12 +86,11 @@ use time::OffsetDateTime;
 use tracing::debug;
 
 use self::incremental::FacetsUpdateIncremental;
-use super::FacetsUpdateBulk;
+use super::{FacetsUpdateBulk, MergeDeladdBtreesetString, MergeDeladdCboRoaringBitmaps};
 use crate::facet::FacetType;
 use crate::heed_codec::facet::{FacetGroupKey, FacetGroupKeyCodec, FacetGroupValueCodec};
 use crate::heed_codec::BytesRefCodec;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd};
-use crate::update::MergeFn;
 use crate::{try_split_array_at, FieldId, Index, Result};
 
 pub mod bulk;
@@ -105,8 +104,8 @@ pub struct FacetsUpdate<'i> {
     index: &'i Index,
     database: heed::Database<FacetGroupKeyCodec<BytesRefCodec>, FacetGroupValueCodec>,
     facet_type: FacetType,
-    delta_data: Merger<BufReader<File>, MergeFn>,
-    normalized_delta_data: Option<Merger<BufReader<File>, MergeFn>>,
+    delta_data: Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>,
+    normalized_delta_data: Option<Merger<BufReader<File>, MergeDeladdBtreesetString>>,
     group_size: u8,
     max_group_size: u8,
     min_level_size: u8,
@@ -116,8 +115,8 @@ impl<'i> FacetsUpdate<'i> {
     pub fn new(
         index: &'i Index,
         facet_type: FacetType,
-        delta_data: Merger<BufReader<File>, MergeFn>,
-        normalized_delta_data: Option<Merger<BufReader<File>, MergeFn>>,
+        delta_data: Merger<BufReader<File>, MergeDeladdCboRoaringBitmaps>,
+        normalized_delta_data: Option<Merger<BufReader<File>, MergeDeladdBtreesetString>>,
         data_size: u64,
     ) -> Self {
         let database = match facet_type {
@@ -182,12 +181,12 @@ impl<'i> FacetsUpdate<'i> {
 
 fn index_facet_search(
     wtxn: &mut heed::RwTxn<'_>,
-    normalized_delta_data: Merger<BufReader<File>, MergeFn>,
+    normalized_delta_data: Merger<BufReader<File>, MergeDeladdBtreesetString>,
     index: &Index,
 ) -> Result<()> {
     let mut iter = normalized_delta_data.into_stream_merger_iter()?;
     while let Some((key_bytes, delta_bytes)) = iter.next()? {
-        let deladd_reader = KvReaderDelAdd::new(delta_bytes);
+        let deladd_reader = KvReaderDelAdd::from_slice(delta_bytes);
 
         let database_set = index
             .facet_id_normalized_string_strings
@@ -298,8 +297,8 @@ pub(crate) mod test_helpers {
     use crate::search::facet::get_highest_level;
     use crate::snapshot_tests::display_bitmap;
     use crate::update::del_add::{DelAdd, KvWriterDelAdd};
-    use crate::update::index_documents::merge_deladd_cbo_roaring_bitmaps;
-    use crate::update::{FacetsUpdateIncrementalInner, MergeFn};
+    use crate::update::index_documents::MergeDeladdCboRoaringBitmaps;
+    use crate::update::FacetsUpdateIncrementalInner;
     use crate::CboRoaringBitmapCodec;
 
     /// Utility function to generate a string whose position in a lexicographically
@@ -484,7 +483,7 @@ pub(crate) mod test_helpers {
             }
             writer.finish().unwrap();
             let reader = grenad::Reader::new(std::io::Cursor::new(new_data)).unwrap();
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             builder.push(reader.into_cursor().unwrap());
             let merger = builder.build();
 

crates/milli/src/update/index_documents/enrich.rs

@@ -47,7 +47,7 @@ pub fn enrich_documents_batch<R: Read + Seek>(
                 return match cursor.next_document()? {
                     Some(first_document) => Ok(Err(UserError::MissingDocumentId {
                         primary_key: primary_key.to_string(),
-                        document: obkv_to_object(&first_document, &documents_batch_index)?,
+                        document: obkv_to_object(first_document, &documents_batch_index)?,
                     })),
                     None => unreachable!("Called with reader.is_empty()"),
                 };
@@ -106,7 +106,7 @@ pub fn enrich_documents_batch<R: Read + Seek>(
     let mut count = 0;
     while let Some(document) = cursor.next_document()? {
         let document_id = match fetch_or_generate_document_id(
-            &document,
+            document,
             &documents_batch_index,
             primary_key,
             autogenerate_docids,
@@ -145,7 +145,7 @@ pub fn enrich_documents_batch<R: Read + Seek>(
 #[tracing::instrument(level = "trace", skip(uuid_buffer, documents_batch_index, document)
 target = "indexing::documents")]
 fn fetch_or_generate_document_id(
-    document: &obkv::KvReader<'_, FieldId>,
+    document: &obkv::KvReader<FieldId>,
     documents_batch_index: &DocumentsBatchIndex,
     primary_key: PrimaryKey<'_>,
     autogenerate_docids: bool,

crates/milli/src/update/index_documents/extract/extract_docid_word_positions.rs

@@ -8,7 +8,7 @@ use obkv::{KvReader, KvWriterU16};
 use roaring::RoaringBitmap;
 use serde_json::Value;
 
-use super::helpers::{create_sorter, keep_latest_obkv, sorter_into_reader, GrenadParameters};
+use super::helpers::{create_sorter, sorter_into_reader, GrenadParameters, KeepLatestObkv};
 use crate::error::{InternalError, SerializationError};
 use crate::update::del_add::{del_add_from_two_obkvs, DelAdd, KvReaderDelAdd};
 use crate::update::settings::{InnerIndexSettings, InnerIndexSettingsDiff};
@@ -35,11 +35,12 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
     let mut documents_ids = RoaringBitmap::new();
     let mut docid_word_positions_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        keep_latest_obkv,
+        KeepLatestObkv,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory,
+        true,
     );
 
     // initialize buffers.
@@ -80,10 +81,10 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
             .try_into()
             .map(u32::from_be_bytes)
             .map_err(|_| SerializationError::InvalidNumberSerialization)?;
-        let obkv = KvReader::<FieldId>::new(value);
+        let obkv = KvReader::<FieldId>::from_slice(value);
 
         // if the searchable fields didn't change, skip the searchable indexing for this document.
-        if !force_reindexing && !searchable_fields_changed(&obkv, settings_diff) {
+        if !force_reindexing && !searchable_fields_changed(obkv, settings_diff) {
             continue;
         }
 
@@ -98,7 +99,7 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
             || {
                 // deletions
                 tokens_from_document(
-                    &obkv,
+                    obkv,
                     &settings_diff.old,
                     &del_tokenizer,
                     max_positions_per_attributes,
@@ -109,7 +110,7 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
             || {
                 // additions
                 tokens_from_document(
-                    &obkv,
+                    obkv,
                     &settings_diff.new,
                     &add_tokenizer,
                     max_positions_per_attributes,
@@ -126,13 +127,13 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
         // transforming two KV<FieldId, KV<u16, String>> into one KV<FieldId, KV<DelAdd, KV<u16, String>>>
         value_buffer.clear();
         del_add_from_two_obkvs(
-            &KvReader::<FieldId>::new(del_obkv),
-            &KvReader::<FieldId>::new(add_obkv),
+            KvReader::<FieldId>::from_slice(del_obkv),
+            KvReader::<FieldId>::from_slice(add_obkv),
             &mut value_buffer,
         )?;
 
         // write each KV<DelAdd, KV<u16, String>> into the sorter, field by field.
-        let obkv = KvReader::<FieldId>::new(&value_buffer);
+        let obkv = KvReader::<FieldId>::from_slice(&value_buffer);
         for (field_id, value) in obkv.iter() {
             key_buffer.truncate(mem::size_of::<u32>());
             key_buffer.extend_from_slice(&field_id.to_be_bytes());
@@ -146,13 +147,13 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
 
 /// Check if any searchable fields of a document changed.
 fn searchable_fields_changed(
-    obkv: &KvReader<'_, FieldId>,
+    obkv: &KvReader<FieldId>,
     settings_diff: &InnerIndexSettingsDiff,
 ) -> bool {
     let searchable_fields = &settings_diff.new.searchable_fields_ids;
     for (field_id, field_bytes) in obkv.iter() {
         if searchable_fields.contains(&field_id) {
-            let del_add = KvReaderDelAdd::new(field_bytes);
+            let del_add = KvReaderDelAdd::from_slice(field_bytes);
             match (del_add.get(DelAdd::Deletion), del_add.get(DelAdd::Addition)) {
                 // if both fields are None, check the next field.
                 (None, None) => (),
@@ -189,7 +190,7 @@ fn tokenizer_builder<'a>(
 
 /// Extract words mapped with their positions of a document.
 fn tokens_from_document<'a>(
-    obkv: &KvReader<'a, FieldId>,
+    obkv: &'a KvReader<FieldId>,
     settings: &InnerIndexSettings,
     tokenizer: &Tokenizer<'_>,
     max_positions_per_attributes: u32,
@@ -202,7 +203,7 @@ fn tokens_from_document<'a>(
         // if field is searchable.
         if settings.searchable_fields_ids.contains(&field_id) {
             // extract deletion or addition only.
-            if let Some(field_bytes) = KvReaderDelAdd::new(field_bytes).get(del_add) {
+            if let Some(field_bytes) = KvReaderDelAdd::from_slice(field_bytes).get(del_add) {
                 // parse json.
                 let value =
                     serde_json::from_slice(field_bytes).map_err(InternalError::SerdeJson)?;

crates/milli/src/update/index_documents/extract/extract_facet_number_docids.rs

@@ -4,7 +4,7 @@ use std::io::{self, BufReader};
 use heed::{BytesDecode, BytesEncode};
 
 use super::helpers::{
-    create_sorter, merge_deladd_cbo_roaring_bitmaps, sorter_into_reader, GrenadParameters,
+    create_sorter, sorter_into_reader, GrenadParameters, MergeDeladdCboRoaringBitmaps,
 };
 use crate::heed_codec::facet::{
     FacetGroupKey, FacetGroupKeyCodec, FieldDocIdFacetF64Codec, OrderedF64Codec,
@@ -27,11 +27,12 @@ pub fn extract_facet_number_docids<R: io::Read + io::Seek>(
 
     let mut facet_number_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory,
+        true,
     );
 
     let mut buffer = Vec::new();
@@ -45,7 +46,7 @@ pub fn extract_facet_number_docids<R: io::Read + io::Seek>(
 
         buffer.clear();
         let mut obkv = KvWriterDelAdd::new(&mut buffer);
-        for (deladd_key, _) in KvReaderDelAdd::new(deladd_obkv_bytes).iter() {
+        for (deladd_key, _) in KvReaderDelAdd::from_slice(deladd_obkv_bytes).iter() {
             obkv.insert(deladd_key, document_id.to_ne_bytes())?;
         }
         obkv.finish()?;

crates/milli/src/update/index_documents/extract/extract_facet_string_docids.rs

@@ -15,7 +15,7 @@ use crate::heed_codec::{BEU16StrCodec, StrRefCodec};
 use crate::localized_attributes_rules::LocalizedFieldIds;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::index_documents::helpers::{
-    merge_deladd_btreeset_string, merge_deladd_cbo_roaring_bitmaps,
+    MergeDeladdBtreesetString, MergeDeladdCboRoaringBitmaps,
 };
 use crate::update::settings::InnerIndexSettingsDiff;
 use crate::{FieldId, Result, MAX_FACET_VALUE_LENGTH};
@@ -56,26 +56,28 @@ fn extract_facet_string_docids_document_update<R: io::Read + io::Seek>(
 
     let mut facet_string_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     let mut normalized_facet_string_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        merge_deladd_btreeset_string,
+        MergeDeladdBtreesetString,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     let mut buffer = Vec::new();
     let mut cursor = docid_fid_facet_string.into_cursor()?;
     while let Some((key, deladd_original_value_bytes)) = cursor.move_on_next()? {
-        let deladd_reader = KvReaderDelAdd::new(deladd_original_value_bytes);
+        let deladd_reader = KvReaderDelAdd::from_slice(deladd_original_value_bytes);
 
         let is_same_value = deladd_reader.get(DelAdd::Deletion).is_some()
             && deladd_reader.get(DelAdd::Addition).is_some();
@@ -144,26 +146,28 @@ fn extract_facet_string_docids_settings<R: io::Read + io::Seek>(
 
     let mut facet_string_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     let mut normalized_facet_string_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        merge_deladd_btreeset_string,
+        MergeDeladdBtreesetString,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     let mut buffer = Vec::new();
     let mut cursor = docid_fid_facet_string.into_cursor()?;
     while let Some((key, deladd_original_value_bytes)) = cursor.move_on_next()? {
-        let deladd_reader = KvReaderDelAdd::new(deladd_original_value_bytes);
+        let deladd_reader = KvReaderDelAdd::from_slice(deladd_original_value_bytes);
 
         let is_same_value = deladd_reader.get(DelAdd::Deletion).is_some()
             && deladd_reader.get(DelAdd::Addition).is_some();

crates/milli/src/update/index_documents/extract/extract_fid_docid_facet_values.rs

@@ -1,10 +1,8 @@
-use std::borrow::Cow;
 use std::collections::{BTreeMap, BTreeSet};
 use std::convert::TryInto;
 use std::fs::File;
 use std::io::{self, BufReader};
 use std::mem::size_of;
-use std::result::Result as StdResult;
 
 use bytemuck::bytes_of;
 use grenad::Sorter;
@@ -15,13 +13,13 @@ use roaring::RoaringBitmap;
 use serde_json::{from_slice, Value};
 use FilterableValues::{Empty, Null, Values};
 
-use super::helpers::{create_sorter, keep_first, sorter_into_reader, GrenadParameters};
+use super::helpers::{create_sorter, sorter_into_reader, GrenadParameters, KeepFirst};
 use crate::error::InternalError;
 use crate::facet::value_encoding::f64_into_bytes;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::index_documents::{create_writer, writer_into_reader};
 use crate::update::settings::InnerIndexSettingsDiff;
-use crate::{CboRoaringBitmapCodec, DocumentId, Error, FieldId, Result, MAX_FACET_VALUE_LENGTH};
+use crate::{CboRoaringBitmapCodec, DocumentId, FieldId, Result, MAX_FACET_VALUE_LENGTH};
 
 /// The length of the elements that are always in the buffer when inserting new values.
 const TRUNCATE_SIZE: usize = size_of::<FieldId>() + size_of::<DocumentId>();
@@ -50,20 +48,22 @@ pub fn extract_fid_docid_facet_values<R: io::Read + io::Seek>(
 
     let mut fid_docid_facet_numbers_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        keep_first,
+        KeepFirst,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     let mut fid_docid_facet_strings_sorter = create_sorter(
         grenad::SortAlgorithm::Stable,
-        keep_first,
+        KeepFirst,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 2),
+        true,
     );
 
     // The tuples represents the Del and Add side for a bitmap
@@ -83,10 +83,10 @@ pub fn extract_fid_docid_facet_values<R: io::Read + io::Seek>(
     if !settings_diff.settings_update_only || old_faceted_fids != new_faceted_fids {
         let mut cursor = obkv_documents.into_cursor()?;
         while let Some((docid_bytes, value)) = cursor.move_on_next()? {
-            let obkv = obkv::KvReader::new(value);
+            let obkv = obkv::KvReader::from_slice(value);
             let get_document_json_value = move |field_id, side| {
                 obkv.get(field_id)
-                    .map(KvReaderDelAdd::new)
+                    .map(KvReaderDelAdd::from_slice)
                     .and_then(|kv| kv.get(side))
                     .map(from_slice)
                     .transpose()
@@ -330,15 +330,12 @@ fn truncate_str(s: &str) -> &str {
 
 /// Computes the diff between both Del and Add numbers and
 /// only inserts the parts that differ in the sorter.
-fn insert_numbers_diff<MF>(
-    fid_docid_facet_numbers_sorter: &mut Sorter<MF>,
+fn insert_numbers_diff(
+    fid_docid_facet_numbers_sorter: &mut Sorter<KeepFirst>,
     key_buffer: &mut Vec<u8>,
     mut del_numbers: Vec<f64>,
     mut add_numbers: Vec<f64>,
-) -> Result<()>
-where
-    MF: for<'a> Fn(&[u8], &[Cow<'a, [u8]>]) -> StdResult<Cow<'a, [u8]>, Error>,
-{
+) -> Result<()> {
     // We sort and dedup the float numbers
     del_numbers.sort_unstable_by_key(|f| OrderedFloat(*f));
     add_numbers.sort_unstable_by_key(|f| OrderedFloat(*f));
@@ -390,15 +387,12 @@ where
 
 /// Computes the diff between both Del and Add strings and
 /// only inserts the parts that differ in the sorter.
-fn insert_strings_diff<MF>(
-    fid_docid_facet_strings_sorter: &mut Sorter<MF>,
+fn insert_strings_diff(
+    fid_docid_facet_strings_sorter: &mut Sorter<KeepFirst>,
     key_buffer: &mut Vec<u8>,
     mut del_strings: Vec<(String, String)>,
     mut add_strings: Vec<(String, String)>,
-) -> Result<()>
-where
-    MF: for<'a> Fn(&[u8], &[Cow<'a, [u8]>]) -> StdResult<Cow<'a, [u8]>, Error>,
-{
+) -> Result<()> {
     // We sort and dedup the normalized and original strings
     del_strings.sort_unstable();
     add_strings.sort_unstable();

crates/milli/src/update/index_documents/extract/extract_fid_word_count_docids.rs

@@ -4,8 +4,8 @@ use std::io::{self, BufReader};
 use obkv::KvReaderU16;
 
 use super::helpers::{
-    create_sorter, merge_deladd_cbo_roaring_bitmaps, sorter_into_reader, try_split_array_at,
-    GrenadParameters,
+    create_sorter, sorter_into_reader, try_split_array_at, GrenadParameters,
+    MergeDeladdCboRoaringBitmaps,
 };
 use crate::error::SerializationError;
 use crate::index::db_name::DOCID_WORD_POSITIONS;
@@ -30,11 +30,12 @@ pub fn extract_fid_word_count_docids<R: io::Read + io::Seek>(
 
     let mut fid_word_count_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory,
+        true,
     );
 
     let mut key_buffer = Vec::new();
@@ -45,19 +46,23 @@ pub fn extract_fid_word_count_docids<R: io::Read + io::Seek>(
             .ok_or(SerializationError::Decoding { db_name: Some(DOCID_WORD_POSITIONS) })?;
         let document_id = u32::from_be_bytes(document_id_bytes);
 
-        let del_add_reader = KvReaderDelAdd::new(value);
+        let del_add_reader = KvReaderDelAdd::from_slice(value);
         let deletion = del_add_reader
             // get deleted words
             .get(DelAdd::Deletion)
             // count deleted words
-            .map(|deletion| KvReaderU16::new(deletion).iter().take(MAX_COUNTED_WORDS + 1).count())
+            .map(|deletion| {
+                KvReaderU16::from_slice(deletion).iter().take(MAX_COUNTED_WORDS + 1).count()
+            })
             // keep the count if under or equal to MAX_COUNTED_WORDS
             .filter(|&word_count| word_count <= MAX_COUNTED_WORDS);
         let addition = del_add_reader
             // get added words
             .get(DelAdd::Addition)
             // count added words
-            .map(|addition| KvReaderU16::new(addition).iter().take(MAX_COUNTED_WORDS + 1).count())
+            .map(|addition| {
+                KvReaderU16::from_slice(addition).iter().take(MAX_COUNTED_WORDS + 1).count()
+            })
             // keep the count if under or equal to MAX_COUNTED_WORDS
             .filter(|&word_count| word_count <= MAX_COUNTED_WORDS);
 

crates/milli/src/update/index_documents/extract/extract_geo_points.rs

@@ -29,22 +29,20 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
 
     let mut cursor = obkv_documents.into_cursor()?;
     while let Some((docid_bytes, value)) = cursor.move_on_next()? {
-        let obkv = obkv::KvReader::new(value);
+        let obkv = obkv::KvReader::from_slice(value);
         // since we only need the primary key when we throw an error
         // we create this getter to lazily get it when needed
         let document_id = || -> Value {
-            let reader = KvReaderDelAdd::new(obkv.get(primary_key_id).unwrap());
+            let reader = KvReaderDelAdd::from_slice(obkv.get(primary_key_id).unwrap());
             let document_id =
                 reader.get(DelAdd::Deletion).or(reader.get(DelAdd::Addition)).unwrap();
             serde_json::from_slice(document_id).unwrap()
         };
 
         // extract old version
-        let del_lat_lng =
-            extract_lat_lng(&obkv, &settings_diff.old, DelAdd::Deletion, document_id)?;
+        let del_lat_lng = extract_lat_lng(obkv, &settings_diff.old, DelAdd::Deletion, document_id)?;
         // extract new version
-        let add_lat_lng =
-            extract_lat_lng(&obkv, &settings_diff.new, DelAdd::Addition, document_id)?;
+        let add_lat_lng = extract_lat_lng(obkv, &settings_diff.new, DelAdd::Addition, document_id)?;
 
         if del_lat_lng != add_lat_lng {
             let mut obkv = KvWriterDelAdd::memory();
@@ -68,15 +66,17 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
 
 /// Extract the finite floats lat and lng from two bytes slices.
 fn extract_lat_lng(
-    document: &obkv::KvReader<'_, FieldId>,
+    document: &obkv::KvReader<FieldId>,
     settings: &InnerIndexSettings,
     deladd: DelAdd,
     document_id: impl Fn() -> Value,
 ) -> Result<Option<[f64; 2]>> {
     match settings.geo_fields_ids {
         Some((lat_fid, lng_fid)) => {
-            let lat = document.get(lat_fid).map(KvReaderDelAdd::new).and_then(|r| r.get(deladd));
-            let lng = document.get(lng_fid).map(KvReaderDelAdd::new).and_then(|r| r.get(deladd));
+            let lat =
+                document.get(lat_fid).map(KvReaderDelAdd::from_slice).and_then(|r| r.get(deladd));
+            let lng =
+                document.get(lng_fid).map(KvReaderDelAdd::from_slice).and_then(|r| r.get(deladd));
             let (lat, lng) = match (lat, lng) {
                 (Some(lat), Some(lng)) => (lat, lng),
                 (Some(_), None) => {

crates/milli/src/update/index_documents/extract/extract_vector_points.rs

@@ -21,7 +21,7 @@ use crate::update::settings::InnerIndexSettingsDiff;
 use crate::vector::error::{EmbedErrorKind, PossibleEmbeddingMistakes, UnusedVectorsDistribution};
 use crate::vector::parsed_vectors::{ParsedVectorsDiff, VectorState, RESERVED_VECTORS_FIELD_NAME};
 use crate::vector::settings::ReindexAction;
-use crate::vector::{Embedder, Embeddings};
+use crate::vector::{Embedder, Embedding};
 use crate::{try_split_array_at, DocumentId, FieldId, Result, ThreadPoolNoAbort};
 
 /// The length of the elements that are always in the buffer when inserting new values.
@@ -313,7 +313,7 @@ pub fn extract_vector_points<R: io::Read + io::Seek>(
         debug_assert!(from_utf8(external_id_bytes).is_ok());
         let docid = DocumentId::from_be_bytes(docid_bytes);
 
-        let obkv = obkv::KvReader::new(value);
+        let obkv = obkv::KvReader::from_slice(value);
         key_buffer.clear();
         key_buffer.extend_from_slice(docid_bytes.as_slice());
 
@@ -481,7 +481,7 @@ pub fn extract_vector_points<R: io::Read + io::Seek>(
 #[allow(clippy::too_many_arguments)] // feel free to find efficient way to factor arguments
 fn extract_vector_document_diff(
     docid: DocumentId,
-    obkv: obkv::KvReader<'_, FieldId>,
+    obkv: &obkv::KvReader<FieldId>,
     prompt: &Prompt,
     (add_to_user_provided, remove_from_user_provided): (&mut RoaringBitmap, &mut RoaringBitmap),
     (old, new): (VectorState, VectorState),
@@ -526,7 +526,7 @@ fn extract_vector_document_diff(
             // Do we keep this document?
             let document_is_kept = obkv
                 .iter()
-                .map(|(_, deladd)| KvReaderDelAdd::new(deladd))
+                .map(|(_, deladd)| KvReaderDelAdd::from_slice(deladd))
                 .any(|deladd| deladd.get(DelAdd::Addition).is_some());
 
             if document_is_kept {
@@ -536,9 +536,11 @@ fn extract_vector_document_diff(
                 }
                 // Don't give up if the old prompt was failing
                 let old_prompt = Some(&prompt).map(|p| {
-                    p.render(obkv, DelAdd::Deletion, old_fields_ids_map).unwrap_or_default()
+                    p.render_kvdeladd(obkv, DelAdd::Deletion, old_fields_ids_map)
+                        .unwrap_or_default()
                 });
-                let new_prompt = prompt.render(obkv, DelAdd::Addition, new_fields_ids_map)?;
+                let new_prompt =
+                    prompt.render_kvdeladd(obkv, DelAdd::Addition, new_fields_ids_map)?;
                 if old_prompt.as_ref() != Some(&new_prompt) {
                     let old_prompt = old_prompt.unwrap_or_default();
                     tracing::trace!(
@@ -562,7 +564,7 @@ fn extract_vector_document_diff(
             // Do we keep this document?
             let document_is_kept = obkv
                 .iter()
-                .map(|(_, deladd)| KvReaderDelAdd::new(deladd))
+                .map(|(_, deladd)| KvReaderDelAdd::from_slice(deladd))
                 .any(|deladd| deladd.get(DelAdd::Addition).is_some());
             if document_is_kept {
                 if embedder_is_manual {
@@ -570,7 +572,7 @@ fn extract_vector_document_diff(
                     return Ok(VectorStateDelta::NoChange);
                 }
                 // becomes autogenerated
-                VectorStateDelta::NowGenerated(prompt.render(
+                VectorStateDelta::NowGenerated(prompt.render_kvdeladd(
                     obkv,
                     DelAdd::Addition,
                     new_fields_ids_map,
@@ -588,7 +590,7 @@ fn extract_vector_document_diff(
             // Do we keep this document?
             let document_is_kept = obkv
                 .iter()
-                .map(|(_, deladd)| KvReaderDelAdd::new(deladd))
+                .map(|(_, deladd)| KvReaderDelAdd::from_slice(deladd))
                 .any(|deladd| deladd.get(DelAdd::Addition).is_some());
             if document_is_kept {
                 // if the new version of documents has the vectors in the DB,
@@ -606,16 +608,17 @@ fn extract_vector_document_diff(
 }
 
 fn regenerate_if_prompt_changed(
-    obkv: obkv::KvReader<'_, FieldId>,
+    obkv: &obkv::KvReader<FieldId>,
     (old_prompt, new_prompt): (&Prompt, &Prompt),
     (old_fields_ids_map, new_fields_ids_map): (
         &FieldsIdsMapWithMetadata,
         &FieldsIdsMapWithMetadata,
     ),
 ) -> Result<VectorStateDelta> {
-    let old_prompt =
-        old_prompt.render(obkv, DelAdd::Deletion, old_fields_ids_map).unwrap_or(Default::default());
-    let new_prompt = new_prompt.render(obkv, DelAdd::Addition, new_fields_ids_map)?;
+    let old_prompt = old_prompt
+        .render_kvdeladd(obkv, DelAdd::Deletion, old_fields_ids_map)
+        .unwrap_or(Default::default());
+    let new_prompt = new_prompt.render_kvdeladd(obkv, DelAdd::Addition, new_fields_ids_map)?;
 
     if new_prompt == old_prompt {
         return Ok(VectorStateDelta::NoChange);
@@ -624,11 +627,11 @@ fn regenerate_if_prompt_changed(
 }
 
 fn regenerate_prompt(
-    obkv: obkv::KvReader<'_, FieldId>,
+    obkv: &obkv::KvReader<FieldId>,
     prompt: &Prompt,
     new_fields_ids_map: &FieldsIdsMapWithMetadata,
 ) -> Result<VectorStateDelta> {
-    let prompt = prompt.render(obkv, DelAdd::Addition, new_fields_ids_map)?;
+    let prompt = prompt.render_kvdeladd(obkv, DelAdd::Addition, new_fields_ids_map)?;
 
     Ok(VectorStateDelta::NowGenerated(prompt))
 }
@@ -738,7 +741,7 @@ pub fn extract_embeddings<R: io::Read + io::Seek>(
                 .flat_map(|docids| docids.iter())
                 .zip(chunked_embeds.iter().flat_map(|embeds| embeds.iter()))
             {
-                state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings.as_inner()))?;
+                state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings))?;
             }
             chunks_ids.clear();
         }
@@ -759,7 +762,7 @@ pub fn extract_embeddings<R: io::Read + io::Seek>(
             .flat_map(|docids| docids.iter())
             .zip(chunked_embeds.iter().flat_map(|embeds| embeds.iter()))
         {
-            state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings.as_inner()))?;
+            state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings))?;
         }
     }
 
@@ -775,7 +778,7 @@ pub fn extract_embeddings<R: io::Read + io::Seek>(
 
         if let Some(embeds) = embeds.first() {
             for (docid, embeddings) in current_chunk_ids.iter().zip(embeds.iter()) {
-                state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings.as_inner()))?;
+                state_writer.insert(docid.to_be_bytes(), cast_slice(embeddings))?;
             }
         }
     }
@@ -790,7 +793,7 @@ fn embed_chunks(
     possible_embedding_mistakes: &PossibleEmbeddingMistakes,
     unused_vectors_distribution: &UnusedVectorsDistribution,
     request_threads: &ThreadPoolNoAbort,
-) -> Result<Vec<Vec<Embeddings<f32>>>> {
+) -> Result<Vec<Vec<Embedding>>> {
     match embedder.embed_chunks(text_chunks, request_threads) {
         Ok(chunks) => Ok(chunks),
         Err(error) => {

crates/milli/src/update/index_documents/extract/extract_word_docids.rs

@@ -7,8 +7,8 @@ use obkv::KvReaderU16;
 use roaring::RoaringBitmap;
 
 use super::helpers::{
-    create_sorter, create_writer, merge_deladd_cbo_roaring_bitmaps, try_split_array_at,
-    writer_into_reader, GrenadParameters,
+    create_sorter, create_writer, try_split_array_at, writer_into_reader, GrenadParameters,
+    MergeDeladdCboRoaringBitmaps,
 };
 use crate::error::SerializationError;
 use crate::heed_codec::StrBEU16Codec;
@@ -16,7 +16,6 @@ use crate::index::db_name::DOCID_WORD_POSITIONS;
 use crate::update::del_add::{is_noop_del_add_obkv, DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::index_documents::helpers::sorter_into_reader;
 use crate::update::settings::InnerIndexSettingsDiff;
-use crate::update::MergeFn;
 use crate::{CboRoaringBitmapCodec, DocumentId, FieldId, Result};
 
 /// Extracts the word and the documents ids where this word appear.
@@ -40,11 +39,12 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
 
     let mut word_fid_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 3),
+        true,
     );
     let mut key_buffer = Vec::new();
     let mut del_words = BTreeSet::new();
@@ -58,17 +58,17 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
         let document_id = u32::from_be_bytes(document_id_bytes);
         let fid = u16::from_be_bytes(fid_bytes);
 
-        let del_add_reader = KvReaderDelAdd::new(value);
+        let del_add_reader = KvReaderDelAdd::from_slice(value);
         // extract all unique words to remove.
         if let Some(deletion) = del_add_reader.get(DelAdd::Deletion) {
-            for (_pos, word) in KvReaderU16::new(deletion).iter() {
+            for (_pos, word) in KvReaderU16::from_slice(deletion).iter() {
                 del_words.insert(word.to_vec());
             }
         }
 
         // extract all unique additional words.
         if let Some(addition) = del_add_reader.get(DelAdd::Addition) {
-            for (_pos, word) in KvReaderU16::new(addition).iter() {
+            for (_pos, word) in KvReaderU16::from_slice(addition).iter() {
                 add_words.insert(word.to_vec());
             }
         }
@@ -94,20 +94,22 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
 
     let mut word_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 3),
+        true,
     );
 
     let mut exact_word_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory.map(|m| m / 3),
+        true,
     );
 
     let mut iter = word_fid_docids_sorter.into_stream_merger_iter()?;
@@ -115,7 +117,7 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
     // NOTE: replacing sorters by bitmap merging is less efficient, so, use sorters.
     while let Some((key, value)) = iter.next()? {
         // only keep the value if their is a change to apply in the DB.
-        if !is_noop_del_add_obkv(KvReaderDelAdd::new(value)) {
+        if !is_noop_del_add_obkv(KvReaderDelAdd::from_slice(value)) {
             word_fid_docids_writer.insert(key, value)?;
         }
 
@@ -123,7 +125,7 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
             .map_err(|_| SerializationError::Decoding { db_name: Some(DOCID_WORD_POSITIONS) })?;
 
         // merge all deletions
-        let obkv = KvReaderDelAdd::new(value);
+        let obkv = KvReaderDelAdd::from_slice(value);
         if let Some(value) = obkv.get(DelAdd::Deletion) {
             let delete_from_exact = settings_diff.old.exact_attributes.contains(&fid);
             buffer.clear();
@@ -163,7 +165,7 @@ fn words_into_sorter(
     key_buffer: &mut Vec<u8>,
     del_words: &BTreeSet<Vec<u8>>,
     add_words: &BTreeSet<Vec<u8>>,
-    word_fid_docids_sorter: &mut grenad::Sorter<MergeFn>,
+    word_fid_docids_sorter: &mut grenad::Sorter<MergeDeladdCboRoaringBitmaps>,
 ) -> Result<()> {
     use itertools::merge_join_by;
     use itertools::EitherOrBoth::{Both, Left, Right};

crates/milli/src/update/index_documents/extract/extract_word_pair_proximity_docids.rs

@@ -6,8 +6,8 @@ use std::{cmp, io};
 use obkv::KvReaderU16;
 
 use super::helpers::{
-    create_sorter, create_writer, merge_deladd_cbo_roaring_bitmaps, try_split_array_at,
-    writer_into_reader, GrenadParameters, MergeFn,
+    create_sorter, create_writer, try_split_array_at, writer_into_reader, GrenadParameters,
+    MergeDeladdCboRoaringBitmaps,
 };
 use crate::error::SerializationError;
 use crate::index::db_name::DOCID_WORD_POSITIONS;
@@ -44,11 +44,12 @@ pub fn extract_word_pair_proximity_docids<R: io::Read + io::Seek>(
         .map(|_| {
             create_sorter(
                 grenad::SortAlgorithm::Unstable,
-                merge_deladd_cbo_roaring_bitmaps,
+                MergeDeladdCboRoaringBitmaps,
                 indexer.chunk_compression_type,
                 indexer.chunk_compression_level,
                 indexer.max_nb_chunks,
                 max_memory.map(|m| m / MAX_DISTANCE as usize),
+                true,
             )
         })
         .collect();
@@ -92,8 +93,8 @@ pub fn extract_word_pair_proximity_docids<R: io::Read + io::Seek>(
                 }
 
                 // deletions
-                if let Some(deletion) = KvReaderDelAdd::new(value).get(DelAdd::Deletion) {
-                    for (position, word) in KvReaderU16::new(deletion).iter() {
+                if let Some(deletion) = KvReaderDelAdd::from_slice(value).get(DelAdd::Deletion) {
+                    for (position, word) in KvReaderU16::from_slice(deletion).iter() {
                         // drain the proximity window until the head word is considered close to the word we are inserting.
                         while del_word_positions.front().map_or(false, |(_w, p)| {
                             index_proximity(*p as u32, position as u32) >= MAX_DISTANCE
@@ -125,8 +126,8 @@ pub fn extract_word_pair_proximity_docids<R: io::Read + io::Seek>(
                 }
 
                 // additions
-                if let Some(addition) = KvReaderDelAdd::new(value).get(DelAdd::Addition) {
-                    for (position, word) in KvReaderU16::new(addition).iter() {
+                if let Some(addition) = KvReaderDelAdd::from_slice(value).get(DelAdd::Addition) {
+                    for (position, word) in KvReaderU16::from_slice(addition).iter() {
                         // drain the proximity window until the head word is considered close to the word we are inserting.
                         while add_word_positions.front().map_or(false, |(_w, p)| {
                             index_proximity(*p as u32, position as u32) >= MAX_DISTANCE
@@ -197,7 +198,7 @@ fn document_word_positions_into_sorter(
     document_id: DocumentId,
     del_word_pair_proximity: &BTreeMap<(String, String), u8>,
     add_word_pair_proximity: &BTreeMap<(String, String), u8>,
-    word_pair_proximity_docids_sorters: &mut [grenad::Sorter<MergeFn>],
+    word_pair_proximity_docids_sorters: &mut [grenad::Sorter<MergeDeladdCboRoaringBitmaps>],
 ) -> Result<()> {
     use itertools::merge_join_by;
     use itertools::EitherOrBoth::{Both, Left, Right};

crates/milli/src/update/index_documents/extract/extract_word_position_docids.rs

@@ -5,14 +5,13 @@ use std::io::{self, BufReader};
 use obkv::KvReaderU16;
 
 use super::helpers::{
-    create_sorter, merge_deladd_cbo_roaring_bitmaps, sorter_into_reader, try_split_array_at,
-    GrenadParameters,
+    create_sorter, sorter_into_reader, try_split_array_at, GrenadParameters,
+    MergeDeladdCboRoaringBitmaps,
 };
 use crate::error::SerializationError;
 use crate::index::db_name::DOCID_WORD_POSITIONS;
 use crate::update::del_add::{DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::settings::InnerIndexSettingsDiff;
-use crate::update::MergeFn;
 use crate::{bucketed_position, DocumentId, Result};
 
 /// Extracts the word positions and the documents ids where this word appear.
@@ -29,11 +28,12 @@ pub fn extract_word_position_docids<R: io::Read + io::Seek>(
 
     let mut word_position_docids_sorter = create_sorter(
         grenad::SortAlgorithm::Unstable,
-        merge_deladd_cbo_roaring_bitmaps,
+        MergeDeladdCboRoaringBitmaps,
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
         indexer.max_nb_chunks,
         max_memory,
+        true,
     );
 
     let mut del_word_positions: BTreeSet<(u16, Vec<u8>)> = BTreeSet::new();
@@ -60,10 +60,10 @@ pub fn extract_word_position_docids<R: io::Read + io::Seek>(
 
         current_document_id = Some(document_id);
 
-        let del_add_reader = KvReaderDelAdd::new(value);
+        let del_add_reader = KvReaderDelAdd::from_slice(value);
         // extract all unique words to remove.
         if let Some(deletion) = del_add_reader.get(DelAdd::Deletion) {
-            for (position, word_bytes) in KvReaderU16::new(deletion).iter() {
+            for (position, word_bytes) in KvReaderU16::from_slice(deletion).iter() {
                 let position = bucketed_position(position);
                 del_word_positions.insert((position, word_bytes.to_vec()));
             }
@@ -71,7 +71,7 @@ pub fn extract_word_position_docids<R: io::Read + io::Seek>(
 
         // extract all unique additional words.
         if let Some(addition) = del_add_reader.get(DelAdd::Addition) {
-            for (position, word_bytes) in KvReaderU16::new(addition).iter() {
+            for (position, word_bytes) in KvReaderU16::from_slice(addition).iter() {
                 let position = bucketed_position(position);
                 add_word_positions.insert((position, word_bytes.to_vec()));
             }
@@ -100,7 +100,7 @@ fn words_position_into_sorter(
     key_buffer: &mut Vec<u8>,
     del_word_positions: &BTreeSet<(u16, Vec<u8>)>,
     add_word_positions: &BTreeSet<(u16, Vec<u8>)>,
-    word_position_docids_sorter: &mut grenad::Sorter<MergeFn>,
+    word_position_docids_sorter: &mut grenad::Sorter<MergeDeladdCboRoaringBitmaps>,
 ) -> Result<()> {
     use itertools::merge_join_by;
     use itertools::EitherOrBoth::{Both, Left, Right};

crates/milli/src/update/index_documents/helpers/grenad_helpers.rs

@@ -1,11 +1,10 @@
-use std::borrow::Cow;
 use std::fs::File;
 use std::io::{self, BufReader, BufWriter, Seek};
 
-use grenad::{CompressionType, Sorter};
+use grenad::{CompressionType, MergeFunction, Sorter};
 use heed::types::Bytes;
 
-use super::{ClonableMmap, MergeFn};
+use super::ClonableMmap;
 use crate::update::index_documents::valid_lmdb_key;
 use crate::Result;
 
@@ -31,14 +30,15 @@ pub fn create_writer<R: io::Write>(
 /// A helper function that creates a grenad sorter
 /// with the given parameters. The max memory is
 /// clamped to something reasonable.
-pub fn create_sorter(
+pub fn create_sorter<MF: MergeFunction>(
     sort_algorithm: grenad::SortAlgorithm,
-    merge: MergeFn,
+    merge: MF,
     chunk_compression_type: grenad::CompressionType,
     chunk_compression_level: Option<u32>,
     max_nb_chunks: Option<usize>,
     max_memory: Option<usize>,
-) -> grenad::Sorter<MergeFn> {
+    sort_in_parallel: bool,
+) -> grenad::Sorter<MF> {
     let mut builder = grenad::Sorter::builder(merge);
     builder.chunk_compression_type(chunk_compression_type);
     if let Some(level) = chunk_compression_level {
@@ -52,15 +52,19 @@ pub fn create_sorter(
         builder.allow_realloc(false);
     }
     builder.sort_algorithm(sort_algorithm);
-    builder.sort_in_parallel(true);
+    builder.sort_in_parallel(sort_in_parallel);
     builder.build()
 }
 
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::grenad")]
-pub fn sorter_into_reader(
-    sorter: grenad::Sorter<MergeFn>,
+pub fn sorter_into_reader<MF>(
+    sorter: grenad::Sorter<MF>,
     indexer: GrenadParameters,
-) -> Result<grenad::Reader<BufReader<File>>> {
+) -> Result<grenad::Reader<BufReader<File>>>
+where
+    MF: MergeFunction,
+    crate::Error: From<MF::Error>,
+{
     let mut writer = create_writer(
         indexer.chunk_compression_type,
         indexer.chunk_compression_level,
@@ -79,6 +83,8 @@ pub fn writer_into_reader(
     grenad::Reader::new(BufReader::new(file)).map_err(Into::into)
 }
 
+/// # Safety
+/// We use memory mapping inside. So, according to the Rust community, it's unsafe.
 pub unsafe fn as_cloneable_grenad(
     reader: &grenad::Reader<BufReader<File>>,
 ) -> Result<grenad::Reader<CursorClonableMmap>> {
@@ -113,12 +119,8 @@ impl GrenadParameters {
     ///
     /// This should be called inside of a rayon thread pool,
     /// otherwise, it will take the global number of threads.
-    ///
-    /// The max memory cannot exceed a given reasonable value.
     pub fn max_memory_by_thread(&self) -> Option<usize> {
-        self.max_memory.map(|max_memory| {
-            (max_memory / rayon::current_num_threads()).min(MAX_GRENAD_SORTER_USAGE)
-        })
+        self.max_memory.map(|max_memory| (max_memory / rayon::current_num_threads()))
     }
 }
 
@@ -169,8 +171,8 @@ pub fn grenad_obkv_into_chunks<R: io::Read + io::Seek>(
 /// Write provided sorter in database using serialize_value function.
 /// merge_values function is used if an entry already exist in the database.
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::grenad")]
-pub fn write_sorter_into_database<K, V, FS, FM>(
-    sorter: Sorter<MergeFn>,
+pub fn write_sorter_into_database<K, V, FS, FM, MF>(
+    sorter: Sorter<MF>,
     database: &heed::Database<K, V>,
     wtxn: &mut heed::RwTxn<'_>,
     index_is_empty: bool,
@@ -180,6 +182,8 @@ pub fn write_sorter_into_database<K, V, FS, FM>(
 where
     FS: for<'a> Fn(&'a [u8], &'a mut Vec<u8>) -> Result<&'a [u8]>,
     FM: for<'a> Fn(&[u8], &[u8], &'a mut Vec<u8>) -> Result<Option<&'a [u8]>>,
+    MF: MergeFunction,
+    crate::Error: From<MF::Error>,
 {
     let mut buffer = Vec::new();
     let database = database.remap_types::<Bytes, Bytes>();
@@ -207,8 +211,3 @@ where
 
     Ok(())
 }
-
-/// Used when trying to merge readers, but you don't actually care about the values.
-pub fn merge_ignore_values<'a>(_key: &[u8], _values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
-    Ok(Cow::Owned(Vec::new()))
-}

crates/milli/src/update/index_documents/helpers/merge_functions.rs

@@ -3,6 +3,8 @@ use std::collections::BTreeSet;
 use std::io;
 use std::result::Result as StdResult;
 
+use either::Either;
+use grenad::MergeFunction;
 use roaring::RoaringBitmap;
 
 use crate::heed_codec::CboRoaringBitmapCodec;
@@ -10,7 +12,8 @@ use crate::update::del_add::{DelAdd, KvReaderDelAdd, KvWriterDelAdd};
 use crate::update::index_documents::transform::Operation;
 use crate::Result;
 
-pub type MergeFn = for<'a> fn(&[u8], &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>>;
+pub type EitherObkvMerge =
+    Either<ObkvsKeepLastAdditionMergeDeletions, ObkvsMergeAdditionsAndDeletions>;
 
 pub fn serialize_roaring_bitmap(bitmap: &RoaringBitmap, buffer: &mut Vec<u8>) -> io::Result<()> {
     buffer.clear();
@@ -18,35 +21,53 @@ pub fn serialize_roaring_bitmap(bitmap: &RoaringBitmap, buffer: &mut Vec<u8>) ->
     bitmap.serialize_into(buffer)
 }
 
-pub fn merge_roaring_bitmaps<'a>(_key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
-    if values.len() == 1 {
-        Ok(values[0].clone())
-    } else {
-        let merged = values
-            .iter()
-            .map(AsRef::as_ref)
-            .map(RoaringBitmap::deserialize_from)
-            .map(StdResult::unwrap)
-            .reduce(|a, b| a | b)
-            .unwrap();
-        let mut buffer = Vec::new();
-        serialize_roaring_bitmap(&merged, &mut buffer)?;
-        Ok(Cow::Owned(buffer))
+pub struct MergeRoaringBitmaps;
+
+impl MergeFunction for MergeRoaringBitmaps {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        if values.len() == 1 {
+            Ok(values[0].clone())
+        } else {
+            let merged = values
+                .iter()
+                .map(AsRef::as_ref)
+                .map(RoaringBitmap::deserialize_from)
+                .map(StdResult::unwrap)
+                .reduce(|a, b| a | b)
+                .unwrap();
+            let mut buffer = Vec::new();
+            serialize_roaring_bitmap(&merged, &mut buffer)?;
+            Ok(Cow::Owned(buffer))
+        }
     }
 }
 
-pub fn keep_first<'a>(_key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
-    Ok(values[0].clone())
+pub struct KeepFirst;
+
+impl MergeFunction for KeepFirst {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        Ok(values[0].clone())
+    }
 }
 
 /// Only the last value associated with an id is kept.
-pub fn keep_latest_obkv<'a>(_key: &[u8], obkvs: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
-    Ok(obkvs.last().unwrap().clone())
+pub struct KeepLatestObkv;
+
+impl MergeFunction for KeepLatestObkv {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], obkvs: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        Ok(obkvs.last().unwrap().clone())
+    }
 }
 
 pub fn merge_two_del_add_obkvs(
-    base: obkv::KvReaderU16<'_>,
-    update: obkv::KvReaderU16<'_>,
+    base: &obkv::KvReaderU16,
+    update: &obkv::KvReaderU16,
     merge_additions: bool,
     buffer: &mut Vec<u8>,
 ) {
@@ -66,7 +87,7 @@ pub fn merge_two_del_add_obkvs(
                     // If merge_additions is false, recreate an obkv keeping the deletions only.
                     value_buffer.clear();
                     let mut value_writer = KvWriterDelAdd::new(&mut value_buffer);
-                    let base_reader = KvReaderDelAdd::new(v);
+                    let base_reader = KvReaderDelAdd::from_slice(v);
 
                     if let Some(deletion) = base_reader.get(DelAdd::Deletion) {
                         value_writer.insert(DelAdd::Deletion, deletion).unwrap();
@@ -80,8 +101,8 @@ pub fn merge_two_del_add_obkvs(
                 // merge deletions and additions.
                 value_buffer.clear();
                 let mut value_writer = KvWriterDelAdd::new(&mut value_buffer);
-                let base_reader = KvReaderDelAdd::new(base);
-                let update_reader = KvReaderDelAdd::new(update);
+                let base_reader = KvReaderDelAdd::from_slice(base);
+                let update_reader = KvReaderDelAdd::from_slice(update);
 
                 // keep newest deletion.
                 if let Some(deletion) = update_reader
@@ -131,8 +152,8 @@ fn inner_merge_del_add_obkvs<'a>(
             break;
         }
 
-        let newest = obkv::KvReader::new(&acc);
-        let oldest = obkv::KvReader::new(&current[1..]);
+        let newest = obkv::KvReader::from_slice(&acc);
+        let oldest = obkv::KvReader::from_slice(&current[1..]);
         merge_two_del_add_obkvs(oldest, newest, merge_additions, &mut buffer);
 
         // we want the result of the merge into our accumulator.
@@ -145,65 +166,79 @@ fn inner_merge_del_add_obkvs<'a>(
 }
 
 /// Merge all the obkvs from the newest to the oldest.
-pub fn obkvs_merge_additions_and_deletions<'a>(
-    _key: &[u8],
-    obkvs: &[Cow<'a, [u8]>],
-) -> Result<Cow<'a, [u8]>> {
-    inner_merge_del_add_obkvs(obkvs, true)
+#[derive(Copy, Clone)]
+pub struct ObkvsMergeAdditionsAndDeletions;
+
+impl MergeFunction for ObkvsMergeAdditionsAndDeletions {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], obkvs: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        inner_merge_del_add_obkvs(obkvs, true)
+    }
 }
 
 /// Merge all the obkvs deletions from the newest to the oldest and keep only the newest additions.
-pub fn obkvs_keep_last_addition_merge_deletions<'a>(
-    _key: &[u8],
-    obkvs: &[Cow<'a, [u8]>],
-) -> Result<Cow<'a, [u8]>> {
-    inner_merge_del_add_obkvs(obkvs, false)
+#[derive(Copy, Clone)]
+pub struct ObkvsKeepLastAdditionMergeDeletions;
+
+impl MergeFunction for ObkvsKeepLastAdditionMergeDeletions {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], obkvs: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        inner_merge_del_add_obkvs(obkvs, false)
+    }
 }
 
 /// Do a union of all the CboRoaringBitmaps in the values.
-pub fn merge_cbo_roaring_bitmaps<'a>(
-    _key: &[u8],
-    values: &[Cow<'a, [u8]>],
-) -> Result<Cow<'a, [u8]>> {
-    if values.len() == 1 {
-        Ok(values[0].clone())
-    } else {
-        let mut vec = Vec::new();
-        CboRoaringBitmapCodec::merge_into(values, &mut vec)?;
-        Ok(Cow::from(vec))
+pub struct MergeCboRoaringBitmaps;
+
+impl MergeFunction for MergeCboRoaringBitmaps {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        if values.len() == 1 {
+            Ok(values[0].clone())
+        } else {
+            let mut vec = Vec::new();
+            CboRoaringBitmapCodec::merge_into(values, &mut vec)?;
+            Ok(Cow::from(vec))
+        }
     }
 }
 
 /// Do a union of CboRoaringBitmaps on both sides of a DelAdd obkv
 /// separately and outputs a new DelAdd with both unions.
-pub fn merge_deladd_cbo_roaring_bitmaps<'a>(
-    _key: &[u8],
-    values: &[Cow<'a, [u8]>],
-) -> Result<Cow<'a, [u8]>> {
-    if values.len() == 1 {
-        Ok(values[0].clone())
-    } else {
-        // Retrieve the bitmaps from both sides
-        let mut del_bitmaps_bytes = Vec::new();
-        let mut add_bitmaps_bytes = Vec::new();
-        for value in values {
-            let obkv = KvReaderDelAdd::new(value);
-            if let Some(bitmap_bytes) = obkv.get(DelAdd::Deletion) {
-                del_bitmaps_bytes.push(bitmap_bytes);
-            }
-            if let Some(bitmap_bytes) = obkv.get(DelAdd::Addition) {
-                add_bitmaps_bytes.push(bitmap_bytes);
-            }
-        }
+pub struct MergeDeladdCboRoaringBitmaps;
 
-        let mut output_deladd_obkv = KvWriterDelAdd::memory();
-        let mut buffer = Vec::new();
-        CboRoaringBitmapCodec::merge_into(del_bitmaps_bytes, &mut buffer)?;
-        output_deladd_obkv.insert(DelAdd::Deletion, &buffer)?;
-        buffer.clear();
-        CboRoaringBitmapCodec::merge_into(add_bitmaps_bytes, &mut buffer)?;
-        output_deladd_obkv.insert(DelAdd::Addition, &buffer)?;
-        output_deladd_obkv.into_inner().map(Cow::from).map_err(Into::into)
+impl MergeFunction for MergeDeladdCboRoaringBitmaps {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        if values.len() == 1 {
+            Ok(values[0].clone())
+        } else {
+            // Retrieve the bitmaps from both sides
+            let mut del_bitmaps_bytes = Vec::new();
+            let mut add_bitmaps_bytes = Vec::new();
+            for value in values {
+                let obkv = KvReaderDelAdd::from_slice(value);
+                if let Some(bitmap_bytes) = obkv.get(DelAdd::Deletion) {
+                    del_bitmaps_bytes.push(bitmap_bytes);
+                }
+                if let Some(bitmap_bytes) = obkv.get(DelAdd::Addition) {
+                    add_bitmaps_bytes.push(bitmap_bytes);
+                }
+            }
+
+            let mut output_deladd_obkv = KvWriterDelAdd::memory();
+            let mut buffer = Vec::new();
+            CboRoaringBitmapCodec::merge_into(del_bitmaps_bytes, &mut buffer)?;
+            output_deladd_obkv.insert(DelAdd::Deletion, &buffer)?;
+            buffer.clear();
+            CboRoaringBitmapCodec::merge_into(add_bitmaps_bytes, &mut buffer)?;
+            output_deladd_obkv.insert(DelAdd::Addition, &buffer)?;
+            output_deladd_obkv.into_inner().map(Cow::from).map_err(Into::into)
+        }
     }
 }
 
@@ -217,7 +252,7 @@ pub fn merge_deladd_cbo_roaring_bitmaps_into_cbo_roaring_bitmap<'a>(
     buffer: &'a mut Vec<u8>,
 ) -> Result<Option<&'a [u8]>> {
     Ok(CboRoaringBitmapCodec::merge_deladd_into(
-        KvReaderDelAdd::new(deladd_obkv),
+        KvReaderDelAdd::from_slice(deladd_obkv),
         previous,
         buffer,
     )?)
@@ -225,37 +260,55 @@ pub fn merge_deladd_cbo_roaring_bitmaps_into_cbo_roaring_bitmap<'a>(
 
 /// Do a union of BtreeSet on both sides of a DelAdd obkv
 /// separately and outputs a new DelAdd with both unions.
-pub fn merge_deladd_btreeset_string<'a>(
-    _key: &[u8],
-    values: &[Cow<'a, [u8]>],
-) -> Result<Cow<'a, [u8]>> {
-    if values.len() == 1 {
-        Ok(values[0].clone())
-    } else {
-        // Retrieve the bitmaps from both sides
-        let mut del_set = BTreeSet::new();
-        let mut add_set = BTreeSet::new();
-        for value in values {
-            let obkv = KvReaderDelAdd::new(value);
-            if let Some(bytes) = obkv.get(DelAdd::Deletion) {
-                let set = serde_json::from_slice::<BTreeSet<String>>(bytes).unwrap();
-                for value in set {
-                    del_set.insert(value);
-                }
-            }
-            if let Some(bytes) = obkv.get(DelAdd::Addition) {
-                let set = serde_json::from_slice::<BTreeSet<String>>(bytes).unwrap();
-                for value in set {
-                    add_set.insert(value);
-                }
-            }
-        }
+pub struct MergeDeladdBtreesetString;
 
-        let mut output_deladd_obkv = KvWriterDelAdd::memory();
-        let del = serde_json::to_vec(&del_set).unwrap();
-        output_deladd_obkv.insert(DelAdd::Deletion, &del)?;
-        let add = serde_json::to_vec(&add_set).unwrap();
-        output_deladd_obkv.insert(DelAdd::Addition, &add)?;
-        output_deladd_obkv.into_inner().map(Cow::from).map_err(Into::into)
+impl MergeFunction for MergeDeladdBtreesetString {
+    type Error = crate::Error;
+
+    fn merge<'a>(&self, _key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Cow<'a, [u8]>> {
+        if values.len() == 1 {
+            Ok(values[0].clone())
+        } else {
+            // Retrieve the bitmaps from both sides
+            let mut del_set = BTreeSet::new();
+            let mut add_set = BTreeSet::new();
+            for value in values {
+                let obkv = KvReaderDelAdd::from_slice(value);
+                if let Some(bytes) = obkv.get(DelAdd::Deletion) {
+                    let set = serde_json::from_slice::<BTreeSet<String>>(bytes).unwrap();
+                    for value in set {
+                        del_set.insert(value);
+                    }
+                }
+                if let Some(bytes) = obkv.get(DelAdd::Addition) {
+                    let set = serde_json::from_slice::<BTreeSet<String>>(bytes).unwrap();
+                    for value in set {
+                        add_set.insert(value);
+                    }
+                }
+            }
+
+            let mut output_deladd_obkv = KvWriterDelAdd::memory();
+            let del = serde_json::to_vec(&del_set).unwrap();
+            output_deladd_obkv.insert(DelAdd::Deletion, &del)?;
+            let add = serde_json::to_vec(&add_set).unwrap();
+            output_deladd_obkv.insert(DelAdd::Addition, &add)?;
+            output_deladd_obkv.into_inner().map(Cow::from).map_err(Into::into)
+        }
+    }
+}
+
+/// Used when trying to merge readers, but you don't actually care about the values.
+pub struct MergeIgnoreValues;
+
+impl MergeFunction for MergeIgnoreValues {
+    type Error = crate::Error;
+
+    fn merge<'a>(
+        &self,
+        _key: &[u8],
+        _values: &[Cow<'a, [u8]>],
+    ) -> std::result::Result<Cow<'a, [u8]>, Self::Error> {
+        Ok(Cow::Owned(Vec::new()))
     }
 }

crates/milli/src/update/index_documents/helpers/mod.rs

@@ -7,17 +7,8 @@ use std::convert::{TryFrom, TryInto};
 
 pub use clonable_mmap::{ClonableMmap, CursorClonableMmap};
 use fst::{IntoStreamer, Streamer};
-pub use grenad_helpers::{
-    as_cloneable_grenad, create_sorter, create_writer, grenad_obkv_into_chunks,
-    merge_ignore_values, sorter_into_reader, write_sorter_into_database, writer_into_reader,
-    GrenadParameters,
-};
-pub use merge_functions::{
-    keep_first, keep_latest_obkv, merge_cbo_roaring_bitmaps, merge_deladd_btreeset_string,
-    merge_deladd_cbo_roaring_bitmaps, merge_deladd_cbo_roaring_bitmaps_into_cbo_roaring_bitmap,
-    merge_roaring_bitmaps, obkvs_keep_last_addition_merge_deletions,
-    obkvs_merge_additions_and_deletions, MergeFn,
-};
+pub use grenad_helpers::*;
+pub use merge_functions::*;
 
 use crate::MAX_WORD_LENGTH;
 

crates/milli/src/update/index_documents/parallel.rs

@@ -1,86 +0,0 @@
-use heed::types::Bytes;
-use heed::{Database, RoTxn};
-use obkv::KvReaderU16;
-use roaring::RoaringBitmap;
-
-use crate::{all_obkv_to_json, DocumentId, FieldsIdsMap, Object, ObkvCodec, Result, BEU32};
-
-pub struct ImmutableObkvs<'t> {
-    ids: RoaringBitmap,
-    fields_ids_map: FieldsIdsMap,
-    slices: Vec<&'t [u8]>,
-}
-
-impl<'t> ImmutableObkvs<'t> {
-    /// Creates the structure by fetching all the OBKVs
-    /// and keeping the transaction making the pointers valid.
-    pub fn new(
-        rtxn: &'t RoTxn,
-        documents_database: Database<BEU32, ObkvCodec>,
-        fields_ids_map: FieldsIdsMap,
-        subset: RoaringBitmap,
-    ) -> heed::Result<Self> {
-        let mut slices = Vec::new();
-        let documents_database = documents_database.remap_data_type::<Bytes>();
-        for docid in &subset {
-            let slice = documents_database.get(rtxn, &docid)?.unwrap();
-            slices.push(slice);
-        }
-
-        Ok(ImmutableObkvs { ids: subset, fields_ids_map, slices })
-    }
-
-    /// Returns the OBKVs identified by the given ID.
-    pub fn obkv(&self, docid: DocumentId) -> heed::Result<Option<KvReaderU16<'t>>> {
-        match self
-            .ids
-            .rank(docid)
-            .checked_sub(1)
-            .and_then(|offset| self.slices.get(offset as usize))
-        {
-            Some(bytes) => Ok(Some(KvReaderU16::new(bytes))),
-            None => Ok(None),
-        }
-    }
-
-    /// Returns the owned rhai::Map identified by the given ID.
-    pub fn rhai_map(&self, docid: DocumentId) -> Result<Option<rhai::Map>> {
-        let obkv = match self.obkv(docid) {
-            Ok(Some(obkv)) => obkv,
-            Ok(None) => return Ok(None),
-            Err(e) => return Err(e.into()),
-        };
-
-        let all_keys = obkv.iter().map(|(k, _v)| k).collect::<Vec<_>>();
-        let map: Result<rhai::Map> = all_keys
-            .iter()
-            .copied()
-            .flat_map(|id| obkv.get(id).map(|value| (id, value)))
-            .map(|(id, value)| {
-                let name = self.fields_ids_map.name(id).ok_or(
-                    crate::error::FieldIdMapMissingEntry::FieldId {
-                        field_id: id,
-                        process: "all_obkv_to_rhaimap",
-                    },
-                )?;
-                let value = serde_json::from_slice(value)
-                    .map_err(crate::error::InternalError::SerdeJson)?;
-                Ok((name.into(), value))
-            })
-            .collect();
-
-        map.map(Some)
-    }
-
-    pub fn json_map(&self, docid: DocumentId) -> Result<Option<Object>> {
-        let obkv = match self.obkv(docid) {
-            Ok(Some(obkv)) => obkv,
-            Ok(None) => return Ok(None),
-            Err(e) => return Err(e.into()),
-        };
-
-        all_obkv_to_json(obkv, &self.fields_ids_map).map(Some)
-    }
-}
-
-unsafe impl Sync for ImmutableObkvs<'_> {}

crates/milli/src/update/index_documents/snapshots/mod.rs/filtered_placeholder_search_should_not_return_deleted_documents/word_docids.snap

@@ -1,5 +1,5 @@
 ---
-source: milli/src/update/index_documents/mod.rs
+source: crates/milli/src/update/index_documents/mod.rs
 ---
 1                [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 21, ]
 2                [21, ]
@@ -35,4 +35,3 @@ letter           [1, ]
 outdoor          [4, ]
 painting         [3, ]
 pattern          [2, 3, 9, 10, 13, 14, 16, ]
-

crates/milli/src/update/index_documents/snapshots/mod.rs/simple_documents_replace/initial/word_docids.snap

@@ -1,18 +1,15 @@
 ---
-source: milli/src/update/index_documents/mod.rs
+source: crates/milli/src/update/index_documents/mod.rs
 ---
 0                [1, ]
 1                [2, ]
 10               [1, ]
-12               [0, ]
 1344             [3, ]
 2                [0, ]
-23               [5, ]
 25               [2, ]
 3                [0, ]
-35               [5, ]
 4                [4, ]
-42               [0, 5, ]
+42               [5, ]
 456              [1, ]
 5                [0, ]
 99               [2, ]
@@ -54,4 +51,3 @@ the              [3, 4, 5, ]
 to               [5, ]
 tolkien          [3, ]
 wonderland       [2, ]
-

crates/milli/src/update/index_documents/snapshots/mod.rs/simple_documents_replace/updated/word_docids.snap

@@ -1,19 +1,16 @@
 ---
-source: milli/src/update/index_documents/mod.rs
+source: crates/milli/src/update/index_documents/mod.rs
 ---
 0                [1, ]
 1                [2, ]
 10               [1, ]
-12               [0, ]
 1344             [3, ]
 1813             [0, ]
 2                [0, ]
-23               [5, ]
 25               [2, ]
 3                [0, ]
-35               [5, ]
 4                [4, ]
-42               [0, 5, ]
+42               [5, ]
 456              [1, ]
 5                [0, ]
 99               [2, ]
@@ -54,4 +51,3 @@ the              [1, 3, 4, 5, ]
 to               [5, ]
 tolkien          [3, ]
 wonderland       [2, ]
-

crates/milli/src/update/index_documents/transform.rs

@@ -5,6 +5,7 @@ use std::collections::{BTreeMap, HashMap, HashSet};
 use std::fs::File;
 use std::io::{Read, Seek};
 
+use either::Either;
 use fxhash::FxHashMap;
 use itertools::Itertools;
 use obkv::{KvReader, KvReaderU16, KvWriter};
@@ -13,10 +14,10 @@ use serde_json::Value;
 use smartstring::SmartString;
 
 use super::helpers::{
-    create_sorter, create_writer, keep_first, obkvs_keep_last_addition_merge_deletions,
-    obkvs_merge_additions_and_deletions, sorter_into_reader, MergeFn,
+    create_sorter, sorter_into_reader, EitherObkvMerge, ObkvsKeepLastAdditionMergeDeletions,
+    ObkvsMergeAdditionsAndDeletions,
 };
-use super::{IndexDocumentsMethod, IndexerConfig};
+use super::{create_writer, IndexDocumentsMethod, IndexerConfig, KeepFirst};
 use crate::documents::{DocumentsBatchIndex, EnrichedDocument, EnrichedDocumentsBatchReader};
 use crate::error::{Error, InternalError, UserError};
 use crate::index::{db_name, main_key};
@@ -26,7 +27,7 @@ use crate::update::del_add::{
 };
 use crate::update::index_documents::GrenadParameters;
 use crate::update::settings::{InnerIndexSettings, InnerIndexSettingsDiff};
-use crate::update::{AvailableDocumentsIds, UpdateIndexingStep};
+use crate::update::{AvailableIds, UpdateIndexingStep};
 use crate::vector::parsed_vectors::{ExplicitVectors, VectorOrArrayOfVectors};
 use crate::vector::settings::WriteBackToDocuments;
 use crate::vector::ArroyWrapper;
@@ -55,13 +56,13 @@ pub struct Transform<'a, 'i> {
 
     indexer_settings: &'a IndexerConfig,
     pub index_documents_method: IndexDocumentsMethod,
-    available_documents_ids: AvailableDocumentsIds,
+    available_documents_ids: AvailableIds,
 
     // Both grenad follows the same format:
     // key | value
     // u32 | 1 byte for the Operation byte, the rest is the obkv of the document stored
-    original_sorter: grenad::Sorter<MergeFn>,
-    flattened_sorter: grenad::Sorter<MergeFn>,
+    original_sorter: grenad::Sorter<EitherObkvMerge>,
+    flattened_sorter: grenad::Sorter<EitherObkvMerge>,
 
     replaced_documents_ids: RoaringBitmap,
     new_documents_ids: RoaringBitmap,
@@ -109,11 +110,13 @@ impl<'a, 'i> Transform<'a, 'i> {
         index_documents_method: IndexDocumentsMethod,
         _autogenerate_docids: bool,
     ) -> Result<Self> {
+        use IndexDocumentsMethod::{ReplaceDocuments, UpdateDocuments};
+
         // We must choose the appropriate merge function for when two or more documents
         // with the same user id must be merged or fully replaced in the same batch.
         let merge_function = match index_documents_method {
-            IndexDocumentsMethod::ReplaceDocuments => obkvs_keep_last_addition_merge_deletions,
-            IndexDocumentsMethod::UpdateDocuments => obkvs_merge_additions_and_deletions,
+            ReplaceDocuments => Either::Left(ObkvsKeepLastAdditionMergeDeletions),
+            UpdateDocuments => Either::Right(ObkvsMergeAdditionsAndDeletions),
         };
 
         // We initialize the sorter with the user indexing settings.
@@ -124,6 +127,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             indexer_settings.chunk_compression_level,
             indexer_settings.max_nb_chunks,
             indexer_settings.max_memory.map(|mem| mem / 2),
+            true,
         );
 
         // We initialize the sorter with the user indexing settings.
@@ -134,6 +138,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             indexer_settings.chunk_compression_level,
             indexer_settings.max_nb_chunks,
             indexer_settings.max_memory.map(|mem| mem / 2),
+            true,
         );
         let documents_ids = index.documents_ids(wtxn)?;
 
@@ -141,7 +146,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             index,
             fields_ids_map: index.fields_ids_map(wtxn)?,
             indexer_settings,
-            available_documents_ids: AvailableDocumentsIds::from_documents_ids(&documents_ids),
+            available_documents_ids: AvailableIds::new(&documents_ids),
             original_sorter,
             flattened_sorter,
             index_documents_method,
@@ -279,21 +284,21 @@ impl<'a, 'i> Transform<'a, 'i> {
                     document_sorter_value_buffer.clear();
                     document_sorter_value_buffer.push(Operation::Addition as u8);
                     into_del_add_obkv(
-                        KvReaderU16::new(base_obkv),
+                        KvReaderU16::from_slice(base_obkv),
                         deladd_operation,
                         &mut document_sorter_value_buffer,
                     )?;
                     self.original_sorter
                         .insert(&document_sorter_key_buffer, &document_sorter_value_buffer)?;
-                    let base_obkv = KvReader::new(base_obkv);
+                    let base_obkv = KvReader::from_slice(base_obkv);
                     if let Some(flattened_obkv) =
-                        Self::flatten_from_fields_ids_map(&base_obkv, &mut self.fields_ids_map)?
+                        Self::flatten_from_fields_ids_map(base_obkv, &mut self.fields_ids_map)?
                     {
                         // we recreate our buffer with the flattened documents
                         document_sorter_value_buffer.clear();
                         document_sorter_value_buffer.push(Operation::Addition as u8);
                         into_del_add_obkv(
-                            KvReaderU16::new(&flattened_obkv),
+                            KvReaderU16::from_slice(&flattened_obkv),
                             deladd_operation,
                             &mut document_sorter_value_buffer,
                         )?;
@@ -312,7 +317,7 @@ impl<'a, 'i> Transform<'a, 'i> {
                 document_sorter_value_buffer.clear();
                 document_sorter_value_buffer.push(Operation::Addition as u8);
                 into_del_add_obkv(
-                    KvReaderU16::new(&obkv_buffer),
+                    KvReaderU16::from_slice(&obkv_buffer),
                     DelAddOperation::Addition,
                     &mut document_sorter_value_buffer,
                 )?;
@@ -320,14 +325,14 @@ impl<'a, 'i> Transform<'a, 'i> {
                 self.original_sorter
                     .insert(&document_sorter_key_buffer, &document_sorter_value_buffer)?;
 
-                let flattened_obkv = KvReader::new(&obkv_buffer);
+                let flattened_obkv = KvReader::from_slice(&obkv_buffer);
                 if let Some(obkv) =
-                    Self::flatten_from_fields_ids_map(&flattened_obkv, &mut self.fields_ids_map)?
+                    Self::flatten_from_fields_ids_map(flattened_obkv, &mut self.fields_ids_map)?
                 {
                     document_sorter_value_buffer.clear();
                     document_sorter_value_buffer.push(Operation::Addition as u8);
                     into_del_add_obkv(
-                        KvReaderU16::new(&obkv),
+                        KvReaderU16::from_slice(&obkv),
                         DelAddOperation::Addition,
                         &mut document_sorter_value_buffer,
                     )?
@@ -357,194 +362,6 @@ impl<'a, 'i> Transform<'a, 'i> {
         Ok(documents_count)
     }
 
-    /// The counter part of `read_documents` that removes documents either from the transform or the database.
-    /// It can be called before, after or in between two calls of the `read_documents`.
-    ///
-    /// It needs to update all the internal datastructure in the transform.
-    /// - If the document is coming from the database -> it's marked as a to_delete document
-    /// - If the document to remove was inserted by the `read_documents` method before AND was present in the db,
-    ///   it's marked as `to_delete` + added into the grenad to ensure we don't reinsert it.
-    /// - If the document to remove was inserted by the `read_documents` method before but was NOT present in the db,
-    ///   it's added into the grenad to ensure we don't insert it + removed from the list of new documents ids.
-    /// - If the document to remove was not present in either the db or the transform we do nothing.
-    #[tracing::instrument(level = "trace", skip_all, target = "indexing::documents")]
-    pub fn remove_documents<FA>(
-        &mut self,
-        mut to_remove: Vec<String>,
-        wtxn: &mut heed::RwTxn<'_>,
-        should_abort: FA,
-    ) -> Result<usize>
-    where
-        FA: Fn() -> bool + Sync,
-    {
-        // there may be duplicates in the documents to remove.
-        to_remove.sort_unstable();
-        to_remove.dedup();
-
-        let external_documents_ids = self.index.external_documents_ids();
-
-        let mut documents_deleted = 0;
-        let mut document_sorter_value_buffer = Vec::new();
-        let mut document_sorter_key_buffer = Vec::new();
-        for to_remove in to_remove {
-            if should_abort() {
-                return Err(Error::InternalError(InternalError::AbortedIndexation));
-            }
-
-            // Check if the document has been added in the current indexing process.
-            let deleted_from_current =
-                match self.new_external_documents_ids_builder.entry((*to_remove).into()) {
-                    // if the document was added in a previous iteration of the transform we make it as deleted in the sorters.
-                    HEntry::Occupied(entry) => {
-                        let docid = *entry.get() as u32;
-                        // Key is the concatenation of the internal docid and the external one.
-                        document_sorter_key_buffer.clear();
-                        document_sorter_key_buffer.extend_from_slice(&docid.to_be_bytes());
-                        document_sorter_key_buffer.extend_from_slice(to_remove.as_bytes());
-                        document_sorter_value_buffer.clear();
-                        document_sorter_value_buffer.push(Operation::Deletion as u8);
-                        obkv::KvWriterU16::new(&mut document_sorter_value_buffer).finish().unwrap();
-                        self.original_sorter
-                            .insert(&document_sorter_key_buffer, &document_sorter_value_buffer)?;
-                        self.flattened_sorter
-                            .insert(docid.to_be_bytes(), &document_sorter_value_buffer)?;
-
-                        // we must NOT update the list of replaced_documents_ids
-                        // Either:
-                        // 1. It's already in it and there is nothing to do
-                        // 2. It wasn't in it because the document was created by a previous batch and since
-                        //    we're removing it there is nothing to do.
-                        self.new_documents_ids.remove(docid);
-                        entry.remove_entry();
-                        true
-                    }
-                    HEntry::Vacant(_) => false,
-                };
-
-            // If the document was already in the db we mark it as a `to_delete` document.
-            // Then we push the document in sorters in deletion mode.
-            let deleted_from_db = match external_documents_ids.get(wtxn, &to_remove)? {
-                Some(docid) => {
-                    self.remove_document_from_db(
-                        docid,
-                        to_remove,
-                        wtxn,
-                        &mut document_sorter_key_buffer,
-                        &mut document_sorter_value_buffer,
-                    )?;
-                    true
-                }
-                None => false,
-            };
-
-            // increase counter only if the document existed somewhere before.
-            if deleted_from_current || deleted_from_db {
-                documents_deleted += 1;
-            }
-        }
-
-        Ok(documents_deleted)
-    }
-
-    /// Removes documents from db using their internal document ids.
-    ///
-    /// # Warning
-    ///
-    /// This function is dangerous and will only work correctly if:
-    ///
-    /// - All the passed ids currently exist in the database
-    /// - No batching using the standards `remove_documents` and `add_documents` took place
-    ///
-    /// TODO: make it impossible to call `remove_documents` or `add_documents` on an instance that calls this function.
-    #[tracing::instrument(level = "trace", skip_all, target = "indexing::details")]
-    pub fn remove_documents_from_db_no_batch<FA>(
-        &mut self,
-        to_remove: &RoaringBitmap,
-        wtxn: &mut heed::RwTxn<'_>,
-        should_abort: FA,
-    ) -> Result<usize>
-    where
-        FA: Fn() -> bool + Sync,
-    {
-        let mut documents_deleted = 0;
-        let mut document_sorter_value_buffer = Vec::new();
-        let mut document_sorter_key_buffer = Vec::new();
-        let external_ids = self.index.external_id_of(wtxn, to_remove.iter())?;
-
-        for (internal_docid, external_docid) in to_remove.iter().zip(external_ids) {
-            let external_docid = external_docid?;
-            if should_abort() {
-                return Err(Error::InternalError(InternalError::AbortedIndexation));
-            }
-            self.remove_document_from_db(
-                internal_docid,
-                external_docid,
-                wtxn,
-                &mut document_sorter_key_buffer,
-                &mut document_sorter_value_buffer,
-            )?;
-
-            documents_deleted += 1;
-        }
-
-        Ok(documents_deleted)
-    }
-
-    fn remove_document_from_db(
-        &mut self,
-        internal_docid: u32,
-        external_docid: String,
-        txn: &heed::RoTxn<'_>,
-        document_sorter_key_buffer: &mut Vec<u8>,
-        document_sorter_value_buffer: &mut Vec<u8>,
-    ) -> Result<()> {
-        self.replaced_documents_ids.insert(internal_docid);
-
-        // fetch the obkv document
-        let original_key = internal_docid;
-        let base_obkv = self
-            .index
-            .documents
-            .remap_data_type::<heed::types::Bytes>()
-            .get(txn, &original_key)?
-            .ok_or(InternalError::DatabaseMissingEntry {
-                db_name: db_name::DOCUMENTS,
-                key: None,
-            })?;
-
-        // Key is the concatenation of the internal docid and the external one.
-        document_sorter_key_buffer.clear();
-        document_sorter_key_buffer.extend_from_slice(&internal_docid.to_be_bytes());
-        document_sorter_key_buffer.extend_from_slice(external_docid.as_bytes());
-        // push it as to delete in the original_sorter
-        document_sorter_value_buffer.clear();
-        document_sorter_value_buffer.push(Operation::Deletion as u8);
-        into_del_add_obkv(
-            KvReaderU16::new(base_obkv),
-            DelAddOperation::Deletion,
-            document_sorter_value_buffer,
-        )?;
-        self.original_sorter.insert(&document_sorter_key_buffer, &document_sorter_value_buffer)?;
-
-        // flatten it and push it as to delete in the flattened_sorter
-        let flattened_obkv = KvReader::new(base_obkv);
-        if let Some(obkv) =
-            Self::flatten_from_fields_ids_map(&flattened_obkv, &mut self.fields_ids_map)?
-        {
-            // we recreate our buffer with the flattened documents
-            document_sorter_value_buffer.clear();
-            document_sorter_value_buffer.push(Operation::Deletion as u8);
-            into_del_add_obkv(
-                KvReaderU16::new(&obkv),
-                DelAddOperation::Deletion,
-                document_sorter_value_buffer,
-            )?;
-        }
-        self.flattened_sorter
-            .insert(internal_docid.to_be_bytes(), &document_sorter_value_buffer)?;
-        Ok(())
-    }
-
     // Flatten a document from the fields ids map contained in self and insert the new
     // created fields. Returns `None` if the document doesn't need to be flattened.
     #[tracing::instrument(
@@ -553,7 +370,7 @@ impl<'a, 'i> Transform<'a, 'i> {
         target = "indexing::transform"
     )]
     fn flatten_from_fields_ids_map(
-        obkv: &KvReader<'_, FieldId>,
+        obkv: &KvReader<FieldId>,
         fields_ids_map: &mut FieldsIdsMap,
     ) -> Result<Option<Vec<u8>>> {
         if obkv
@@ -721,10 +538,10 @@ impl<'a, 'i> Transform<'a, 'i> {
                 total_documents: self.documents_count,
             });
 
-            for (key, value) in KvReader::new(val) {
-                let reader = KvReaderDelAdd::new(value);
+            for (key, value) in KvReader::from_slice(val) {
+                let reader = KvReaderDelAdd::from_slice(value);
                 match (reader.get(DelAdd::Deletion), reader.get(DelAdd::Addition)) {
-                    (None, None) => {}
+                    (None, None) => (),
                     (None, Some(_)) => {
                         // New field
                         let name = self.fields_ids_map.name(key).ok_or(
@@ -803,7 +620,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             fst_new_external_documents_ids_builder.insert(key, value)
         })?;
 
-        let old_inner_settings = InnerIndexSettings::from_index(self.index, wtxn)?;
+        let old_inner_settings = InnerIndexSettings::from_index(self.index, wtxn, None)?;
         let fields_ids_map = self.fields_ids_map;
         let primary_key_id = self.index.primary_key(wtxn)?.and_then(|name| fields_ids_map.id(name));
         let mut new_inner_settings = old_inner_settings.clone();
@@ -838,7 +655,7 @@ impl<'a, 'i> Transform<'a, 'i> {
     /// then fill the provided buffers with delta documents using KvWritterDelAdd.
     #[allow(clippy::too_many_arguments)] // need the vectors + fid, feel free to create a struct xo xo
     fn rebind_existing_document(
-        old_obkv: KvReader<'_, FieldId>,
+        old_obkv: &KvReader<FieldId>,
         settings_diff: &InnerIndexSettingsDiff,
         modified_faceted_fields: &HashSet<String>,
         mut injected_vectors: serde_json::Map<String, serde_json::Value>,
@@ -926,7 +743,7 @@ impl<'a, 'i> Transform<'a, 'i> {
         }
 
         let data = obkv_writer.into_inner()?;
-        let obkv = KvReader::<FieldId>::new(&data);
+        let obkv = KvReader::<FieldId>::from_slice(&data);
 
         if let Some(original_obkv_buffer) = original_obkv_buffer {
             original_obkv_buffer.clear();
@@ -936,8 +753,8 @@ impl<'a, 'i> Transform<'a, 'i> {
         if let Some(flattened_obkv_buffer) = flattened_obkv_buffer {
             // take the non-flattened version if flatten_from_fields_ids_map returns None.
             let mut fields_ids_map = settings_diff.new.fields_ids_map.clone();
-            let flattened = Self::flatten_from_fields_ids_map(&obkv, &mut fields_ids_map)?;
-            let flattened = flattened.as_deref().map_or(obkv, KvReader::new);
+            let flattened = Self::flatten_from_fields_ids_map(obkv, &mut fields_ids_map)?;
+            let flattened = flattened.as_deref().map_or(obkv, KvReader::from_slice);
 
             flattened_obkv_buffer.clear();
             into_del_add_obkv_conditional_operation(flattened, flattened_obkv_buffer, |id| {
@@ -980,11 +797,12 @@ impl<'a, 'i> Transform<'a, 'i> {
         let mut original_sorter = if settings_diff.reindex_vectors() {
             Some(create_sorter(
                 grenad::SortAlgorithm::Stable,
-                keep_first,
+                KeepFirst,
                 self.indexer_settings.chunk_compression_type,
                 self.indexer_settings.chunk_compression_level,
                 self.indexer_settings.max_nb_chunks,
                 self.indexer_settings.max_memory.map(|mem| mem / 2),
+                true,
             ))
         } else {
             None
@@ -1019,11 +837,12 @@ impl<'a, 'i> Transform<'a, 'i> {
             if settings_diff.reindex_searchable() || settings_diff.reindex_facets() {
                 Some(create_sorter(
                     grenad::SortAlgorithm::Stable,
-                    keep_first,
+                    KeepFirst,
                     self.indexer_settings.chunk_compression_type,
                     self.indexer_settings.chunk_compression_level,
                     self.indexer_settings.max_nb_chunks,
                     self.indexer_settings.max_memory.map(|mem| mem / 2),
+                    true,
                 ))
             } else {
                 None
@@ -1137,6 +956,9 @@ fn drop_and_reuse<U, T>(mut vec: Vec<U>) -> Vec<T> {
 
 #[cfg(test)]
 mod test {
+    use grenad::MergeFunction;
+    use obkv::KvReaderU16;
+
     use super::*;
 
     #[test]
@@ -1148,21 +970,21 @@ mod test {
         kv_writer.insert(0_u8, [0]).unwrap();
         let buffer = kv_writer.into_inner().unwrap();
         into_del_add_obkv(
-            KvReaderU16::new(&buffer),
+            KvReaderU16::from_slice(&buffer),
             DelAddOperation::Addition,
             &mut additive_doc_0,
         )
         .unwrap();
         additive_doc_0.insert(0, Operation::Addition as u8);
         into_del_add_obkv(
-            KvReaderU16::new(&buffer),
+            KvReaderU16::from_slice(&buffer),
             DelAddOperation::Deletion,
             &mut deletive_doc_0,
         )
         .unwrap();
         deletive_doc_0.insert(0, Operation::Deletion as u8);
         into_del_add_obkv(
-            KvReaderU16::new(&buffer),
+            KvReaderU16::from_slice(&buffer),
             DelAddOperation::DeletionAndAddition,
             &mut del_add_doc_0,
         )
@@ -1174,7 +996,7 @@ mod test {
         kv_writer.insert(1_u8, [1]).unwrap();
         let buffer = kv_writer.into_inner().unwrap();
         into_del_add_obkv(
-            KvReaderU16::new(&buffer),
+            KvReaderU16::from_slice(&buffer),
             DelAddOperation::Addition,
             &mut additive_doc_1,
         )
@@ -1187,32 +1009,39 @@ mod test {
         kv_writer.insert(1_u8, [1]).unwrap();
         let buffer = kv_writer.into_inner().unwrap();
         into_del_add_obkv(
-            KvReaderU16::new(&buffer),
+            KvReaderU16::from_slice(&buffer),
             DelAddOperation::Addition,
             &mut additive_doc_0_1,
         )
         .unwrap();
         additive_doc_0_1.insert(0, Operation::Addition as u8);
 
-        let ret = obkvs_merge_additions_and_deletions(&[], &[Cow::from(additive_doc_0.as_slice())])
-            .unwrap();
+        let ret = MergeFunction::merge(
+            &ObkvsMergeAdditionsAndDeletions,
+            &[],
+            &[Cow::from(additive_doc_0.as_slice())],
+        )
+        .unwrap();
         assert_eq!(*ret, additive_doc_0);
 
-        let ret = obkvs_merge_additions_and_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsMergeAdditionsAndDeletions,
             &[],
             &[Cow::from(deletive_doc_0.as_slice()), Cow::from(additive_doc_0.as_slice())],
         )
         .unwrap();
         assert_eq!(*ret, del_add_doc_0);
 
-        let ret = obkvs_merge_additions_and_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsMergeAdditionsAndDeletions,
             &[],
             &[Cow::from(additive_doc_0.as_slice()), Cow::from(deletive_doc_0.as_slice())],
         )
         .unwrap();
         assert_eq!(*ret, deletive_doc_0);
 
-        let ret = obkvs_merge_additions_and_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsMergeAdditionsAndDeletions,
             &[],
             &[
                 Cow::from(additive_doc_1.as_slice()),
@@ -1223,21 +1052,24 @@ mod test {
         .unwrap();
         assert_eq!(*ret, del_add_doc_0);
 
-        let ret = obkvs_merge_additions_and_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsMergeAdditionsAndDeletions,
             &[],
             &[Cow::from(additive_doc_1.as_slice()), Cow::from(additive_doc_0.as_slice())],
         )
         .unwrap();
         assert_eq!(*ret, additive_doc_0_1);
 
-        let ret = obkvs_keep_last_addition_merge_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsKeepLastAdditionMergeDeletions,
             &[],
             &[Cow::from(additive_doc_1.as_slice()), Cow::from(additive_doc_0.as_slice())],
         )
         .unwrap();
         assert_eq!(*ret, additive_doc_0);
 
-        let ret = obkvs_keep_last_addition_merge_deletions(
+        let ret = MergeFunction::merge(
+            &ObkvsKeepLastAdditionMergeDeletions,
             &[],
             &[
                 Cow::from(deletive_doc_0.as_slice()),

crates/milli/src/update/index_documents/typed_chunk.rs

@@ -4,18 +4,17 @@ use std::fs::File;
 use std::io::{self, BufReader};
 
 use bytemuck::allocation::pod_collect_to_vec;
-use grenad::{Merger, MergerBuilder};
+use grenad::{MergeFunction, Merger, MergerBuilder};
 use heed::types::Bytes;
 use heed::{BytesDecode, RwTxn};
 use obkv::{KvReader, KvWriter};
 use roaring::RoaringBitmap;
 
 use super::helpers::{
-    self, keep_first, merge_deladd_btreeset_string, merge_deladd_cbo_roaring_bitmaps,
-    merge_deladd_cbo_roaring_bitmaps_into_cbo_roaring_bitmap, merge_ignore_values, valid_lmdb_key,
-    CursorClonableMmap,
+    self, merge_deladd_cbo_roaring_bitmaps_into_cbo_roaring_bitmap, valid_lmdb_key,
+    CursorClonableMmap, KeepFirst, MergeDeladdBtreesetString, MergeDeladdCboRoaringBitmaps,
+    MergeIgnoreValues,
 };
-use super::MergeFn;
 use crate::external_documents_ids::{DocumentOperation, DocumentOperationKind};
 use crate::facet::FacetType;
 use crate::index::db_name::DOCUMENTS;
@@ -24,7 +23,7 @@ use crate::proximity::MAX_DISTANCE;
 use crate::update::del_add::{deladd_serialize_add_side, DelAdd, KvReaderDelAdd};
 use crate::update::facet::FacetsUpdate;
 use crate::update::index_documents::helpers::{
-    as_cloneable_grenad, keep_latest_obkv, try_split_array_at,
+    as_cloneable_grenad, try_split_array_at, KeepLatestObkv,
 };
 use crate::update::settings::InnerIndexSettingsDiff;
 use crate::vector::ArroyWrapper;
@@ -141,7 +140,7 @@ pub(crate) fn write_typed_chunk_into_index(
             let vectors_fid =
                 fields_ids_map.id(crate::vector::parsed_vectors::RESERVED_VECTORS_FIELD_NAME);
 
-            let mut builder = MergerBuilder::new(keep_latest_obkv as MergeFn);
+            let mut builder = MergerBuilder::new(KeepLatestObkv);
             for typed_chunk in typed_chunks {
                 let TypedChunk::Documents(chunk) = typed_chunk else {
                     unreachable!();
@@ -163,7 +162,7 @@ pub(crate) fn write_typed_chunk_into_index(
             let mut vectors_buffer = Vec::new();
             while let Some((key, reader)) = iter.next()? {
                 let mut writer: KvWriter<_, FieldId> = KvWriter::memory();
-                let reader: KvReader<'_, FieldId> = KvReader::new(reader);
+                let reader: &KvReader<FieldId> = reader.into();
 
                 let (document_id_bytes, external_id_bytes) = try_split_array_at(key)
                     .ok_or(SerializationError::Decoding { db_name: Some(DOCUMENTS) })?;
@@ -171,7 +170,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 let external_id = std::str::from_utf8(external_id_bytes)?;
 
                 for (field_id, value) in reader.iter() {
-                    let del_add_reader = KvReaderDelAdd::new(value);
+                    let del_add_reader = KvReaderDelAdd::from_slice(value);
 
                     if let Some(addition) = del_add_reader.get(DelAdd::Addition) {
                         let addition = if vectors_fid == Some(field_id) {
@@ -235,7 +234,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_word_count_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdWordCountDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -258,13 +257,10 @@ pub(crate) fn write_typed_chunk_into_index(
             let span = tracing::trace_span!(target: "indexing::write_db", "word_docids");
             let _entered = span.enter();
 
-            let mut word_docids_builder =
-                MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
-            let mut exact_word_docids_builder =
-                MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
-            let mut word_fid_docids_builder =
-                MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
-            let mut fst_merger_builder = MergerBuilder::new(merge_ignore_values as MergeFn);
+            let mut word_docids_builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
+            let mut exact_word_docids_builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
+            let mut word_fid_docids_builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
+            let mut fst_merger_builder = MergerBuilder::new(MergeIgnoreValues);
             for typed_chunk in typed_chunks {
                 let TypedChunk::WordDocids {
                     word_docids_reader,
@@ -329,7 +325,7 @@ pub(crate) fn write_typed_chunk_into_index(
             let span = tracing::trace_span!(target: "indexing::write_db", "word_position_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::WordPositionDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -353,7 +349,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db","field_id_facet_number_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             let mut data_size = 0;
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdFacetNumberDocids(facet_id_number_docids) = typed_chunk
@@ -375,10 +371,9 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_facet_string_docids");
             let _entered = span.enter();
 
-            let mut facet_id_string_builder =
-                MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut facet_id_string_builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             let mut normalized_facet_id_string_builder =
-                MergerBuilder::new(merge_deladd_btreeset_string as MergeFn);
+                MergerBuilder::new(MergeDeladdBtreesetString);
             let mut data_size = 0;
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdFacetStringDocids((
@@ -412,7 +407,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_facet_exists_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdFacetExistsDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -436,7 +431,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_facet_is_null_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdFacetIsNullDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -459,7 +454,7 @@ pub(crate) fn write_typed_chunk_into_index(
             let span = tracing::trace_span!(target: "indexing::write_db", "field_id_facet_is_empty_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdFacetIsEmptyDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -483,7 +478,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "word_pair_proximity_docids");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(merge_deladd_cbo_roaring_bitmaps as MergeFn);
+            let mut builder = MergerBuilder::new(MergeDeladdCboRoaringBitmaps);
             for typed_chunk in typed_chunks {
                 let TypedChunk::WordPairProximityDocids(chunk) = typed_chunk else {
                     unreachable!();
@@ -516,7 +511,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_docid_facet_numbers");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(keep_first as MergeFn);
+            let mut builder = MergerBuilder::new(KeepFirst);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdDocidFacetNumbers(chunk) = typed_chunk else {
                     unreachable!();
@@ -530,7 +525,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 index.field_id_docid_facet_f64s.remap_types::<Bytes, Bytes>();
             let mut iter = merger.into_stream_merger_iter()?;
             while let Some((key, value)) = iter.next()? {
-                let reader = KvReaderDelAdd::new(value);
+                let reader = KvReaderDelAdd::from_slice(value);
                 if valid_lmdb_key(key) {
                     match (reader.get(DelAdd::Deletion), reader.get(DelAdd::Addition)) {
                         (None, None) => {}
@@ -550,7 +545,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 tracing::trace_span!(target: "indexing::write_db", "field_id_docid_facet_strings");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(keep_first as MergeFn);
+            let mut builder = MergerBuilder::new(KeepFirst);
             for typed_chunk in typed_chunks {
                 let TypedChunk::FieldIdDocidFacetStrings(chunk) = typed_chunk else {
                     unreachable!();
@@ -564,7 +559,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 index.field_id_docid_facet_strings.remap_types::<Bytes, Bytes>();
             let mut iter = merger.into_stream_merger_iter()?;
             while let Some((key, value)) = iter.next()? {
-                let reader = KvReaderDelAdd::new(value);
+                let reader = KvReaderDelAdd::from_slice(value);
                 if valid_lmdb_key(key) {
                     match (reader.get(DelAdd::Deletion), reader.get(DelAdd::Addition)) {
                         (None, None) => {}
@@ -583,7 +578,7 @@ pub(crate) fn write_typed_chunk_into_index(
             let span = tracing::trace_span!(target: "indexing::write_db", "geo_points");
             let _entered = span.enter();
 
-            let mut builder = MergerBuilder::new(keep_first as MergeFn);
+            let mut builder = MergerBuilder::new(KeepFirst);
             for typed_chunk in typed_chunks {
                 let TypedChunk::GeoPoints(chunk) = typed_chunk else {
                     unreachable!();
@@ -601,7 +596,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 // convert the key back to a u32 (4 bytes)
                 let docid = key.try_into().map(DocumentId::from_be_bytes).unwrap();
 
-                let deladd_obkv = KvReaderDelAdd::new(value);
+                let deladd_obkv = KvReaderDelAdd::from_slice(value);
                 if let Some(value) = deladd_obkv.get(DelAdd::Deletion) {
                     let geopoint = extract_geo_point(value, docid);
                     rtree.remove(&geopoint);
@@ -620,9 +615,9 @@ pub(crate) fn write_typed_chunk_into_index(
             let span = tracing::trace_span!(target: "indexing::write_db", "vector_points");
             let _entered = span.enter();
 
-            let mut remove_vectors_builder = MergerBuilder::new(keep_first as MergeFn);
-            let mut manual_vectors_builder = MergerBuilder::new(keep_first as MergeFn);
-            let mut embeddings_builder = MergerBuilder::new(keep_first as MergeFn);
+            let mut remove_vectors_builder = MergerBuilder::new(KeepFirst);
+            let mut manual_vectors_builder = MergerBuilder::new(KeepFirst);
+            let mut embeddings_builder = MergerBuilder::new(KeepFirst);
             let mut add_to_user_provided = RoaringBitmap::new();
             let mut remove_from_user_provided = RoaringBitmap::new();
             let mut params = None;
@@ -719,7 +714,7 @@ pub(crate) fn write_typed_chunk_into_index(
                 let (left, _index) = try_split_array_at(key).unwrap();
                 let docid = DocumentId::from_be_bytes(left);
 
-                let vector_deladd_obkv = KvReaderDelAdd::new(value);
+                let vector_deladd_obkv = KvReaderDelAdd::from_slice(value);
                 if let Some(value) = vector_deladd_obkv.get(DelAdd::Deletion) {
                     let vector: Vec<f32> = pod_collect_to_vec(value);
 
@@ -742,7 +737,7 @@ pub(crate) fn write_typed_chunk_into_index(
 }
 
 /// Converts the latitude and longitude back to an xyz GeoPoint.
-fn extract_geo_point(value: &[u8], docid: DocumentId) -> GeoPoint {
+pub fn extract_geo_point(value: &[u8], docid: DocumentId) -> GeoPoint {
     let (lat, tail) = helpers::try_split_array_at::<u8, 8>(value).unwrap();
     let (lng, _) = helpers::try_split_array_at::<u8, 8>(tail).unwrap();
     let point = [f64::from_ne_bytes(lat), f64::from_ne_bytes(lng)];
@@ -750,9 +745,13 @@ fn extract_geo_point(value: &[u8], docid: DocumentId) -> GeoPoint {
     GeoPoint::new(xyz_point, (docid, point))
 }
 
-fn merge_word_docids_reader_into_fst(
-    merger: Merger<CursorClonableMmap, MergeFn>,
-) -> Result<fst::Set<Vec<u8>>> {
+fn merge_word_docids_reader_into_fst<MF>(
+    merger: Merger<CursorClonableMmap, MF>,
+) -> Result<fst::Set<Vec<u8>>>
+where
+    MF: MergeFunction,
+    crate::Error: From<MF::Error>,
+{
     let mut iter = merger.into_stream_merger_iter()?;
     let mut builder = fst::SetBuilder::memory();
 
@@ -766,8 +765,8 @@ fn merge_word_docids_reader_into_fst(
 /// Write provided entries in database using serialize_value function.
 /// merge_values function is used if an entry already exist in the database.
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::write_db")]
-fn write_entries_into_database<R, K, V, FS, FM>(
-    merger: Merger<R, MergeFn>,
+fn write_entries_into_database<R, K, V, FS, FM, MF>(
+    merger: Merger<R, MF>,
     database: &heed::Database<K, V>,
     wtxn: &mut RwTxn<'_>,
     serialize_value: FS,
@@ -777,6 +776,8 @@ where
     R: io::Read + io::Seek,
     FS: for<'a> Fn(&'a [u8], &'a mut Vec<u8>) -> Result<&'a [u8]>,
     FM: for<'a> Fn(&[u8], &[u8], &'a mut Vec<u8>) -> Result<Option<&'a [u8]>>,
+    MF: MergeFunction,
+    crate::Error: From<MF::Error>,
 {
     let mut buffer = Vec::new();
     let database = database.remap_types::<Bytes, Bytes>();
@@ -803,20 +804,22 @@ where
 /// Akin to the `write_entries_into_database` function but specialized
 /// for the case when we only index additional searchable fields only.
 #[tracing::instrument(level = "trace", skip_all, target = "indexing::write_db")]
-fn write_proximity_entries_into_database_additional_searchables<R>(
-    merger: Merger<R, MergeFn>,
+fn write_proximity_entries_into_database_additional_searchables<R, MF>(
+    merger: Merger<R, MF>,
     database: &heed::Database<U8StrStrCodec, CboRoaringBitmapCodec>,
     wtxn: &mut RwTxn<'_>,
 ) -> Result<()>
 where
     R: io::Read + io::Seek,
+    MF: MergeFunction,
+    crate::Error: From<MF::Error>,
 {
     let mut iter = merger.into_stream_merger_iter()?;
     while let Some((key, value)) = iter.next()? {
         if valid_lmdb_key(key) {
             let (proximity_to_insert, word1, word2) =
                 U8StrStrCodec::bytes_decode(key).map_err(heed::Error::Decoding)?;
-            let data_to_insert = match KvReaderDelAdd::new(value).get(DelAdd::Addition) {
+            let data_to_insert = match KvReaderDelAdd::from_slice(value).get(DelAdd::Addition) {
                 Some(value) => {
                     CboRoaringBitmapCodec::bytes_decode(value).map_err(heed::Error::Decoding)?
                 }

crates/milli/src/update/indexer_config.rs

@@ -1,5 +1,6 @@
 use grenad::CompressionType;
 
+use super::GrenadParameters;
 use crate::thread_pool_no_abort::ThreadPoolNoAbort;
 
 #[derive(Debug)]
@@ -15,6 +16,17 @@ pub struct IndexerConfig {
     pub skip_index_budget: bool,
 }
 
+impl IndexerConfig {
+    pub fn grenad_parameters(&self) -> GrenadParameters {
+        GrenadParameters {
+            chunk_compression_type: self.chunk_compression_type,
+            chunk_compression_level: self.chunk_compression_level,
+            max_memory: self.max_memory,
+            max_nb_chunks: self.max_nb_chunks,
+        }
+    }
+}
+
 impl Default for IndexerConfig {
     fn default() -> Self {
         Self {

crates/milli/src/update/mod.rs

@@ -1,11 +1,9 @@
-pub use self::available_documents_ids::AvailableDocumentsIds;
+pub use self::available_ids::AvailableIds;
 pub use self::clear_documents::ClearDocuments;
+pub use self::concurrent_available_ids::ConcurrentAvailableIds;
 pub use self::facet::bulk::FacetsUpdateBulk;
 pub use self::facet::incremental::FacetsUpdateIncrementalInner;
-pub use self::index_documents::{
-    merge_cbo_roaring_bitmaps, merge_roaring_bitmaps, DocumentAdditionResult, DocumentId,
-    IndexDocuments, IndexDocumentsConfig, IndexDocumentsMethod, MergeFn,
-};
+pub use self::index_documents::*;
 pub use self::indexer_config::IndexerConfig;
 pub use self::settings::{validate_embedding_settings, Setting, Settings};
 pub use self::update_step::UpdateIndexingStep;
@@ -13,13 +11,15 @@ pub use self::word_prefix_docids::WordPrefixDocids;
 pub use self::words_prefix_integer_docids::WordPrefixIntegerDocids;
 pub use self::words_prefixes_fst::WordsPrefixesFst;
 
-mod available_documents_ids;
+mod available_ids;
 mod clear_documents;
+mod concurrent_available_ids;
 pub(crate) mod del_add;
 pub(crate) mod facet;
 mod index_documents;
 mod indexer_config;
-mod settings;
+pub mod new;
+pub(crate) mod settings;
 mod update_step;
 mod word_prefix_docids;
 mod words_prefix_integer_docids;

crates/milli/src/update/new/channel.rs

@@ -0,0 +1,516 @@
+use std::marker::PhantomData;
+use std::sync::atomic::{AtomicUsize, Ordering};
+
+use crossbeam_channel::{IntoIter, Receiver, SendError, Sender};
+use heed::types::Bytes;
+use heed::BytesDecode;
+use memmap2::Mmap;
+use roaring::RoaringBitmap;
+
+use super::extract::FacetKind;
+use super::StdResult;
+use crate::heed_codec::facet::{FieldDocIdFacetF64Codec, FieldDocIdFacetStringCodec};
+use crate::index::main_key::{GEO_FACETED_DOCUMENTS_IDS_KEY, GEO_RTREE_KEY};
+use crate::index::IndexEmbeddingConfig;
+use crate::update::new::KvReaderFieldId;
+use crate::vector::Embedding;
+use crate::{DocumentId, Index};
+
+/// The capacity of the channel is currently in number of messages.
+pub fn extractor_writer_channel(cap: usize) -> (ExtractorSender, WriterReceiver) {
+    let (sender, receiver) = crossbeam_channel::bounded(cap);
+    (
+        ExtractorSender {
+            sender,
+            send_count: Default::default(),
+            writer_contentious_count: Default::default(),
+            extractor_contentious_count: Default::default(),
+        },
+        WriterReceiver(receiver),
+    )
+}
+
+pub enum KeyValueEntry {
+    Small { key_length: usize, data: Box<[u8]> },
+    Large { key_entry: KeyEntry, data: Mmap },
+}
+
+impl KeyValueEntry {
+    pub fn from_small_key_value(key: &[u8], value: &[u8]) -> Self {
+        let mut data = Vec::with_capacity(key.len() + value.len());
+        data.extend_from_slice(key);
+        data.extend_from_slice(value);
+        KeyValueEntry::Small { key_length: key.len(), data: data.into_boxed_slice() }
+    }
+
+    fn from_large_key_value(key: &[u8], value: Mmap) -> Self {
+        KeyValueEntry::Large { key_entry: KeyEntry::from_key(key), data: value }
+    }
+
+    pub fn key(&self) -> &[u8] {
+        match self {
+            KeyValueEntry::Small { key_length, data } => &data[..*key_length],
+            KeyValueEntry::Large { key_entry, data: _ } => key_entry.entry(),
+        }
+    }
+
+    pub fn value(&self) -> &[u8] {
+        match self {
+            KeyValueEntry::Small { key_length, data } => &data[*key_length..],
+            KeyValueEntry::Large { key_entry: _, data } => &data[..],
+        }
+    }
+}
+
+pub struct KeyEntry {
+    data: Box<[u8]>,
+}
+
+impl KeyEntry {
+    pub fn from_key(key: &[u8]) -> Self {
+        KeyEntry { data: key.to_vec().into_boxed_slice() }
+    }
+
+    pub fn entry(&self) -> &[u8] {
+        self.data.as_ref()
+    }
+}
+
+pub enum EntryOperation {
+    Delete(KeyEntry),
+    Write(KeyValueEntry),
+}
+
+pub enum WriterOperation {
+    DbOperation(DbOperation),
+    ArroyOperation(ArroyOperation),
+}
+
+pub enum ArroyOperation {
+    /// TODO: call when deleting regular documents
+    DeleteVectors {
+        docid: DocumentId,
+    },
+    SetVectors {
+        docid: DocumentId,
+        embedder_id: u8,
+        embeddings: Vec<Embedding>,
+    },
+    SetVector {
+        docid: DocumentId,
+        embedder_id: u8,
+        embedding: Embedding,
+    },
+    Finish {
+        configs: Vec<IndexEmbeddingConfig>,
+    },
+}
+
+pub struct DbOperation {
+    database: Database,
+    entry: EntryOperation,
+}
+
+#[derive(Debug)]
+pub enum Database {
+    Main,
+    Documents,
+    ExternalDocumentsIds,
+    ExactWordDocids,
+    FidWordCountDocids,
+    WordDocids,
+    WordFidDocids,
+    WordPairProximityDocids,
+    WordPositionDocids,
+    FacetIdIsNullDocids,
+    FacetIdIsEmptyDocids,
+    FacetIdExistsDocids,
+    FacetIdF64NumberDocids,
+    FacetIdStringDocids,
+    FieldIdDocidFacetStrings,
+    FieldIdDocidFacetF64s,
+}
+
+impl Database {
+    pub fn database(&self, index: &Index) -> heed::Database<Bytes, Bytes> {
+        match self {
+            Database::Main => index.main.remap_types(),
+            Database::Documents => index.documents.remap_types(),
+            Database::ExternalDocumentsIds => index.external_documents_ids.remap_types(),
+            Database::ExactWordDocids => index.exact_word_docids.remap_types(),
+            Database::WordDocids => index.word_docids.remap_types(),
+            Database::WordFidDocids => index.word_fid_docids.remap_types(),
+            Database::WordPositionDocids => index.word_position_docids.remap_types(),
+            Database::FidWordCountDocids => index.field_id_word_count_docids.remap_types(),
+            Database::WordPairProximityDocids => index.word_pair_proximity_docids.remap_types(),
+            Database::FacetIdIsNullDocids => index.facet_id_is_null_docids.remap_types(),
+            Database::FacetIdIsEmptyDocids => index.facet_id_is_empty_docids.remap_types(),
+            Database::FacetIdExistsDocids => index.facet_id_exists_docids.remap_types(),
+            Database::FacetIdF64NumberDocids => index.facet_id_f64_docids.remap_types(),
+            Database::FacetIdStringDocids => index.facet_id_string_docids.remap_types(),
+            Database::FieldIdDocidFacetStrings => index.field_id_docid_facet_strings.remap_types(),
+            Database::FieldIdDocidFacetF64s => index.field_id_docid_facet_f64s.remap_types(),
+        }
+    }
+}
+
+impl From<FacetKind> for Database {
+    fn from(value: FacetKind) -> Self {
+        match value {
+            FacetKind::Number => Database::FacetIdF64NumberDocids,
+            FacetKind::String => Database::FacetIdStringDocids,
+            FacetKind::Null => Database::FacetIdIsNullDocids,
+            FacetKind::Empty => Database::FacetIdIsEmptyDocids,
+            FacetKind::Exists => Database::FacetIdExistsDocids,
+        }
+    }
+}
+
+impl DbOperation {
+    pub fn database(&self, index: &Index) -> heed::Database<Bytes, Bytes> {
+        self.database.database(index)
+    }
+
+    pub fn entry(self) -> EntryOperation {
+        self.entry
+    }
+}
+
+pub struct WriterReceiver(Receiver<WriterOperation>);
+
+impl IntoIterator for WriterReceiver {
+    type Item = WriterOperation;
+    type IntoIter = IntoIter<Self::Item>;
+
+    fn into_iter(self) -> Self::IntoIter {
+        self.0.into_iter()
+    }
+}
+
+pub struct ExtractorSender {
+    sender: Sender<WriterOperation>,
+    /// The number of message we sent in total in the channel.
+    send_count: AtomicUsize,
+    /// The number of times we sent something in a channel that was full.
+    writer_contentious_count: AtomicUsize,
+    /// The number of times we sent something in a channel that was empty.
+    extractor_contentious_count: AtomicUsize,
+}
+
+impl Drop for ExtractorSender {
+    fn drop(&mut self) {
+        let send_count = *self.send_count.get_mut();
+        let writer_contentious_count = *self.writer_contentious_count.get_mut();
+        let extractor_contentious_count = *self.extractor_contentious_count.get_mut();
+        tracing::debug!(
+            "Extractor channel stats: {send_count} sends, \
+            {writer_contentious_count} writer contentions ({}%), \
+            {extractor_contentious_count} extractor contentions ({}%)",
+            (writer_contentious_count as f32 / send_count as f32) * 100.0,
+            (extractor_contentious_count as f32 / send_count as f32) * 100.0
+        )
+    }
+}
+
+impl ExtractorSender {
+    pub fn docids<D: DatabaseType>(&self) -> WordDocidsSender<'_, D> {
+        WordDocidsSender { sender: self, _marker: PhantomData }
+    }
+
+    pub fn facet_docids(&self) -> FacetDocidsSender<'_> {
+        FacetDocidsSender { sender: self }
+    }
+
+    pub fn field_id_docid_facet_sender(&self) -> FieldIdDocidFacetSender<'_> {
+        FieldIdDocidFacetSender(self)
+    }
+
+    pub fn documents(&self) -> DocumentsSender<'_> {
+        DocumentsSender(self)
+    }
+
+    pub fn embeddings(&self) -> EmbeddingSender<'_> {
+        EmbeddingSender(&self.sender)
+    }
+
+    pub fn geo(&self) -> GeoSender<'_> {
+        GeoSender(&self.sender)
+    }
+
+    fn send_delete_vector(&self, docid: DocumentId) -> StdResult<(), SendError<()>> {
+        match self
+            .sender
+            .send(WriterOperation::ArroyOperation(ArroyOperation::DeleteVectors { docid }))
+        {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+
+    fn send_db_operation(&self, op: DbOperation) -> StdResult<(), SendError<()>> {
+        if self.sender.is_full() {
+            self.writer_contentious_count.fetch_add(1, Ordering::SeqCst);
+        }
+        if self.sender.is_empty() {
+            self.extractor_contentious_count.fetch_add(1, Ordering::SeqCst);
+        }
+
+        self.send_count.fetch_add(1, Ordering::SeqCst);
+        match self.sender.send(WriterOperation::DbOperation(op)) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+}
+
+pub enum ExactWordDocids {}
+pub enum FidWordCountDocids {}
+pub enum WordDocids {}
+pub enum WordFidDocids {}
+pub enum WordPairProximityDocids {}
+pub enum WordPositionDocids {}
+
+pub trait DatabaseType {
+    const DATABASE: Database;
+}
+
+impl DatabaseType for ExactWordDocids {
+    const DATABASE: Database = Database::ExactWordDocids;
+}
+
+impl DatabaseType for FidWordCountDocids {
+    const DATABASE: Database = Database::FidWordCountDocids;
+}
+
+impl DatabaseType for WordDocids {
+    const DATABASE: Database = Database::WordDocids;
+}
+
+impl DatabaseType for WordFidDocids {
+    const DATABASE: Database = Database::WordFidDocids;
+}
+
+impl DatabaseType for WordPairProximityDocids {
+    const DATABASE: Database = Database::WordPairProximityDocids;
+}
+
+impl DatabaseType for WordPositionDocids {
+    const DATABASE: Database = Database::WordPositionDocids;
+}
+
+pub trait DocidsSender {
+    fn write(&self, key: &[u8], value: &[u8]) -> StdResult<(), SendError<()>>;
+    fn delete(&self, key: &[u8]) -> StdResult<(), SendError<()>>;
+}
+
+pub struct WordDocidsSender<'a, D> {
+    sender: &'a ExtractorSender,
+    _marker: PhantomData<D>,
+}
+
+impl<D: DatabaseType> DocidsSender for WordDocidsSender<'_, D> {
+    fn write(&self, key: &[u8], value: &[u8]) -> StdResult<(), SendError<()>> {
+        let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(key, value));
+        match self.sender.send_db_operation(DbOperation { database: D::DATABASE, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+
+    fn delete(&self, key: &[u8]) -> StdResult<(), SendError<()>> {
+        let entry = EntryOperation::Delete(KeyEntry::from_key(key));
+        match self.sender.send_db_operation(DbOperation { database: D::DATABASE, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+}
+
+pub struct FacetDocidsSender<'a> {
+    sender: &'a ExtractorSender,
+}
+
+impl DocidsSender for FacetDocidsSender<'_> {
+    fn write(&self, key: &[u8], value: &[u8]) -> StdResult<(), SendError<()>> {
+        let (facet_kind, key) = FacetKind::extract_from_key(key);
+        let database = Database::from(facet_kind);
+        // let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(key, value));
+        let entry = match facet_kind {
+            // skip level group size
+            FacetKind::String | FacetKind::Number => {
+                // add facet group size
+                let value = [&[1], value].concat();
+                EntryOperation::Write(KeyValueEntry::from_small_key_value(key, &value))
+            }
+            _ => EntryOperation::Write(KeyValueEntry::from_small_key_value(key, value)),
+        };
+        match self.sender.send_db_operation(DbOperation { database, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+
+    fn delete(&self, key: &[u8]) -> StdResult<(), SendError<()>> {
+        let (facet_kind, key) = FacetKind::extract_from_key(key);
+        let database = Database::from(facet_kind);
+        let entry = EntryOperation::Delete(KeyEntry::from_key(key));
+        match self.sender.send_db_operation(DbOperation { database, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+}
+
+pub struct FieldIdDocidFacetSender<'a>(&'a ExtractorSender);
+
+impl FieldIdDocidFacetSender<'_> {
+    pub fn write_facet_string(&self, key: &[u8], value: &[u8]) -> StdResult<(), SendError<()>> {
+        debug_assert!(FieldDocIdFacetStringCodec::bytes_decode(key).is_ok());
+        let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(key, value));
+        self.0
+            .send_db_operation(DbOperation { database: Database::FieldIdDocidFacetStrings, entry })
+    }
+
+    pub fn write_facet_f64(&self, key: &[u8]) -> StdResult<(), SendError<()>> {
+        debug_assert!(FieldDocIdFacetF64Codec::bytes_decode(key).is_ok());
+        let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(key, &[]));
+        self.0.send_db_operation(DbOperation { database: Database::FieldIdDocidFacetF64s, entry })
+    }
+
+    pub fn delete_facet_string(&self, key: &[u8]) -> StdResult<(), SendError<()>> {
+        debug_assert!(FieldDocIdFacetStringCodec::bytes_decode(key).is_ok());
+        let entry = EntryOperation::Delete(KeyEntry::from_key(key));
+        self.0
+            .send_db_operation(DbOperation { database: Database::FieldIdDocidFacetStrings, entry })
+    }
+
+    pub fn delete_facet_f64(&self, key: &[u8]) -> StdResult<(), SendError<()>> {
+        debug_assert!(FieldDocIdFacetF64Codec::bytes_decode(key).is_ok());
+        let entry = EntryOperation::Delete(KeyEntry::from_key(key));
+        self.0.send_db_operation(DbOperation { database: Database::FieldIdDocidFacetF64s, entry })
+    }
+}
+
+pub struct DocumentsSender<'a>(&'a ExtractorSender);
+
+impl DocumentsSender<'_> {
+    /// TODO do that efficiently
+    pub fn uncompressed(
+        &self,
+        docid: DocumentId,
+        external_id: String,
+        document: &KvReaderFieldId,
+    ) -> StdResult<(), SendError<()>> {
+        let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(
+            &docid.to_be_bytes(),
+            document.as_bytes(),
+        ));
+        match self.0.send_db_operation(DbOperation { database: Database::Documents, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }?;
+
+        let entry = EntryOperation::Write(KeyValueEntry::from_small_key_value(
+            external_id.as_bytes(),
+            &docid.to_be_bytes(),
+        ));
+        match self
+            .0
+            .send_db_operation(DbOperation { database: Database::ExternalDocumentsIds, entry })
+        {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+
+    pub fn delete(&self, docid: DocumentId, external_id: String) -> StdResult<(), SendError<()>> {
+        let entry = EntryOperation::Delete(KeyEntry::from_key(&docid.to_be_bytes()));
+        match self.0.send_db_operation(DbOperation { database: Database::Documents, entry }) {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }?;
+
+        self.0.send_delete_vector(docid)?;
+
+        let entry = EntryOperation::Delete(KeyEntry::from_key(external_id.as_bytes()));
+        match self
+            .0
+            .send_db_operation(DbOperation { database: Database::ExternalDocumentsIds, entry })
+        {
+            Ok(()) => Ok(()),
+            Err(SendError(_)) => Err(SendError(())),
+        }
+    }
+}
+
+pub struct EmbeddingSender<'a>(&'a Sender<WriterOperation>);
+
+impl EmbeddingSender<'_> {
+    pub fn set_vectors(
+        &self,
+        docid: DocumentId,
+        embedder_id: u8,
+        embeddings: Vec<Embedding>,
+    ) -> StdResult<(), SendError<()>> {
+        self.0
+            .send(WriterOperation::ArroyOperation(ArroyOperation::SetVectors {
+                docid,
+                embedder_id,
+                embeddings,
+            }))
+            .map_err(|_| SendError(()))
+    }
+
+    pub fn set_vector(
+        &self,
+        docid: DocumentId,
+        embedder_id: u8,
+        embedding: Embedding,
+    ) -> StdResult<(), SendError<()>> {
+        self.0
+            .send(WriterOperation::ArroyOperation(ArroyOperation::SetVector {
+                docid,
+                embedder_id,
+                embedding,
+            }))
+            .map_err(|_| SendError(()))
+    }
+
+    /// Marks all embedders as "to be built"
+    pub fn finish(self, configs: Vec<IndexEmbeddingConfig>) -> StdResult<(), SendError<()>> {
+        self.0
+            .send(WriterOperation::ArroyOperation(ArroyOperation::Finish { configs }))
+            .map_err(|_| SendError(()))
+    }
+}
+
+pub struct GeoSender<'a>(&'a Sender<WriterOperation>);
+
+impl GeoSender<'_> {
+    pub fn set_rtree(&self, value: Mmap) -> StdResult<(), SendError<()>> {
+        self.0
+            .send(WriterOperation::DbOperation(DbOperation {
+                database: Database::Main,
+                entry: EntryOperation::Write(KeyValueEntry::from_large_key_value(
+                    GEO_RTREE_KEY.as_bytes(),
+                    value,
+                )),
+            }))
+            .map_err(|_| SendError(()))
+    }
+
+    pub fn set_geo_faceted(&self, bitmap: &RoaringBitmap) -> StdResult<(), SendError<()>> {
+        let mut buffer = Vec::new();
+        bitmap.serialize_into(&mut buffer).unwrap();
+
+        self.0
+            .send(WriterOperation::DbOperation(DbOperation {
+                database: Database::Main,
+                entry: EntryOperation::Write(KeyValueEntry::from_small_key_value(
+                    GEO_FACETED_DOCUMENTS_IDS_KEY.as_bytes(),
+                    &buffer,
+                )),
+            }))
+            .map_err(|_| SendError(()))
+    }
+}

crates/milli/src/update/new/document.rs

@@ -0,0 +1,425 @@
+use std::collections::{BTreeMap, BTreeSet};
+
+use heed::RoTxn;
+use raw_collections::RawMap;
+use serde_json::value::RawValue;
+
+use super::vector_document::VectorDocument;
+use super::{KvReaderFieldId, KvWriterFieldId};
+use crate::documents::FieldIdMapper;
+use crate::vector::parsed_vectors::RESERVED_VECTORS_FIELD_NAME;
+use crate::{DocumentId, GlobalFieldsIdsMap, Index, InternalError, Result, UserError};
+
+/// A view into a document that can represent either the current version from the DB,
+/// the update data from payload or other means, or the merged updated version.
+///
+/// The 'doc lifetime is meant to live sufficiently for the document to be handled by the extractors.
+pub trait Document<'doc> {
+    /// Iterate over all **top-level** fields of the document, returning their name and raw JSON value.
+    ///
+    /// - The returned values *may* contain nested fields.
+    /// - The `_vectors` and `_geo` fields are **ignored** by this method, meaning  they are **not returned** by this method.
+    fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'doc str, &'doc RawValue)>>;
+
+    fn len(&self) -> usize;
+
+    fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+
+    /// Get the **top-level** with the specified name, if exists.
+    ///
+    /// - The `_vectors` and `_geo` fields are **ignored** by this method, meaning e.g. `top_level_field("_vectors")` will return `Ok(None)`
+    fn top_level_field(&self, k: &str) -> Result<Option<&'doc RawValue>>;
+
+    /// Returns the unparsed value of the `_vectors` field from the document data.
+    ///
+    /// This field alone is insufficient to retrieve vectors, as they may be stored in a dedicated location in the database.
+    /// Use a [`super::vector_document::VectorDocument`] to access the vector.
+    ///
+    /// This method is meant as a convenience for implementors of [`super::vector_document::VectorDocument`].
+    fn vectors_field(&self) -> Result<Option<&'doc RawValue>>;
+
+    /// Returns the unparsed value of the `_geo` field from the document data.
+    ///
+    /// This field alone is insufficient to retrieve geo data, as they may be stored in a dedicated location in the database.
+    /// Use a [`super::geo_document::GeoDocument`] to access the vector.
+    ///
+    /// This method is meant as a convenience for implementors of [`super::geo_document::GeoDocument`].
+    fn geo_field(&self) -> Result<Option<&'doc RawValue>>;
+}
+
+#[derive(Debug)]
+pub struct DocumentFromDb<'t, Mapper: FieldIdMapper>
+where
+    Mapper: FieldIdMapper,
+{
+    fields_ids_map: &'t Mapper,
+    content: &'t KvReaderFieldId,
+}
+
+impl<'t, Mapper: FieldIdMapper> Clone for DocumentFromDb<'t, Mapper> {
+    #[inline]
+    fn clone(&self) -> Self {
+        *self
+    }
+}
+impl<'t, Mapper: FieldIdMapper> Copy for DocumentFromDb<'t, Mapper> {}
+
+impl<'t, Mapper: FieldIdMapper> Document<'t> for DocumentFromDb<'t, Mapper> {
+    fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'t str, &'t RawValue)>> {
+        let mut it = self.content.iter();
+
+        std::iter::from_fn(move || loop {
+            let (fid, value) = it.next()?;
+            let name = match self.fields_ids_map.name(fid).ok_or(
+                InternalError::FieldIdMapMissingEntry(crate::FieldIdMapMissingEntry::FieldId {
+                    field_id: fid,
+                    process: "getting current document",
+                }),
+            ) {
+                Ok(name) => name,
+                Err(error) => return Some(Err(error.into())),
+            };
+
+            if name == RESERVED_VECTORS_FIELD_NAME || name == "_geo" {
+                continue;
+            }
+
+            let res = (|| {
+                let value =
+                    serde_json::from_slice(value).map_err(crate::InternalError::SerdeJson)?;
+
+                Ok((name, value))
+            })();
+
+            return Some(res);
+        })
+    }
+
+    fn vectors_field(&self) -> Result<Option<&'t RawValue>> {
+        self.field(RESERVED_VECTORS_FIELD_NAME)
+    }
+
+    fn geo_field(&self) -> Result<Option<&'t RawValue>> {
+        self.field("_geo")
+    }
+
+    fn len(&self) -> usize {
+        self.content.iter().count()
+    }
+
+    fn top_level_field(&self, k: &str) -> Result<Option<&'t RawValue>> {
+        if k == RESERVED_VECTORS_FIELD_NAME || k == "_geo" {
+            return Ok(None);
+        }
+        self.field(k)
+    }
+}
+
+impl<'t, Mapper: FieldIdMapper> DocumentFromDb<'t, Mapper> {
+    pub fn new(
+        docid: DocumentId,
+        rtxn: &'t RoTxn,
+        index: &'t Index,
+        db_fields_ids_map: &'t Mapper,
+    ) -> Result<Option<Self>> {
+        index.documents.get(rtxn, &docid).map_err(crate::Error::from).map(|reader| {
+            reader.map(|reader| Self { fields_ids_map: db_fields_ids_map, content: reader })
+        })
+    }
+
+    pub fn field(&self, name: &str) -> Result<Option<&'t RawValue>> {
+        let Some(fid) = self.fields_ids_map.id(name) else {
+            return Ok(None);
+        };
+        let Some(value) = self.content.get(fid) else { return Ok(None) };
+        Ok(Some(serde_json::from_slice(value).map_err(InternalError::SerdeJson)?))
+    }
+}
+
+#[derive(Debug)]
+pub struct DocumentFromVersions<'a, 'doc> {
+    versions: &'a Versions<'doc>,
+}
+
+impl<'a, 'doc> DocumentFromVersions<'a, 'doc> {
+    pub fn new(versions: &'a Versions<'doc>) -> Self {
+        Self { versions }
+    }
+}
+
+impl<'a, 'doc> Document<'doc> for DocumentFromVersions<'a, 'doc> {
+    fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'doc str, &'doc RawValue)>> {
+        self.versions.iter_top_level_fields().map(Ok)
+    }
+
+    fn vectors_field(&self) -> Result<Option<&'doc RawValue>> {
+        Ok(self.versions.vectors_field())
+    }
+
+    fn geo_field(&self) -> Result<Option<&'doc RawValue>> {
+        Ok(self.versions.geo_field())
+    }
+
+    fn len(&self) -> usize {
+        self.versions.len()
+    }
+
+    fn top_level_field(&self, k: &str) -> Result<Option<&'doc RawValue>> {
+        Ok(self.versions.top_level_field(k))
+    }
+}
+
+#[derive(Debug)]
+pub struct MergedDocument<'a, 'doc, 't, Mapper: FieldIdMapper> {
+    new_doc: DocumentFromVersions<'a, 'doc>,
+    db: Option<DocumentFromDb<'t, Mapper>>,
+}
+
+impl<'a, 'doc, 't, Mapper: FieldIdMapper> MergedDocument<'a, 'doc, 't, Mapper> {
+    pub fn with_db(
+        docid: DocumentId,
+        rtxn: &'t RoTxn,
+        index: &'t Index,
+        db_fields_ids_map: &'t Mapper,
+        new_doc: DocumentFromVersions<'a, 'doc>,
+    ) -> Result<Self> {
+        let db = DocumentFromDb::new(docid, rtxn, index, db_fields_ids_map)?;
+        Ok(Self { new_doc, db })
+    }
+
+    pub fn without_db(new_doc: DocumentFromVersions<'a, 'doc>) -> Self {
+        Self { new_doc, db: None }
+    }
+}
+
+impl<'d, 'doc: 'd, 't: 'd, Mapper: FieldIdMapper> Document<'d>
+    for MergedDocument<'d, 'doc, 't, Mapper>
+{
+    fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'d str, &'d RawValue)>> {
+        let mut new_doc_it = self.new_doc.iter_top_level_fields();
+        let mut db_it = self.db.iter().flat_map(|db| db.iter_top_level_fields());
+        let mut seen_fields = BTreeSet::new();
+
+        std::iter::from_fn(move || {
+            if let Some(next) = new_doc_it.next() {
+                if let Ok((name, _)) = next {
+                    seen_fields.insert(name);
+                }
+                return Some(next);
+            }
+            loop {
+                match db_it.next()? {
+                    Ok((name, value)) => {
+                        if seen_fields.contains(name) {
+                            continue;
+                        }
+                        return Some(Ok((name, value)));
+                    }
+                    Err(err) => return Some(Err(err)),
+                }
+            }
+        })
+    }
+
+    fn vectors_field(&self) -> Result<Option<&'d RawValue>> {
+        if let Some(vectors) = self.new_doc.vectors_field()? {
+            return Ok(Some(vectors));
+        }
+
+        let Some(db) = self.db else { return Ok(None) };
+
+        db.vectors_field()
+    }
+
+    fn geo_field(&self) -> Result<Option<&'d RawValue>> {
+        if let Some(geo) = self.new_doc.geo_field()? {
+            return Ok(Some(geo));
+        }
+
+        let Some(db) = self.db else { return Ok(None) };
+
+        db.geo_field()
+    }
+
+    fn len(&self) -> usize {
+        self.iter_top_level_fields().count()
+    }
+
+    fn top_level_field(&self, k: &str) -> Result<Option<&'d RawValue>> {
+        if let Some(f) = self.new_doc.top_level_field(k)? {
+            return Ok(Some(f));
+        }
+        if let Some(db) = self.db {
+            return db.field(k);
+        }
+        Ok(None)
+    }
+}
+
+impl<'doc, D> Document<'doc> for &D
+where
+    D: Document<'doc>,
+{
+    fn iter_top_level_fields(&self) -> impl Iterator<Item = Result<(&'doc str, &'doc RawValue)>> {
+        D::iter_top_level_fields(self)
+    }
+
+    fn vectors_field(&self) -> Result<Option<&'doc RawValue>> {
+        D::vectors_field(self)
+    }
+
+    fn geo_field(&self) -> Result<Option<&'doc RawValue>> {
+        D::geo_field(self)
+    }
+
+    fn len(&self) -> usize {
+        D::len(self)
+    }
+
+    fn top_level_field(&self, k: &str) -> Result<Option<&'doc RawValue>> {
+        D::top_level_field(self, k)
+    }
+}
+
+/// Turn this document into an obkv, whose fields are indexed by the provided `FieldIdMapper`.
+///
+/// The produced obkv is suitable for storing into the documents DB, meaning:
+///
+/// - It contains the contains of `_vectors` that are not configured as an embedder
+/// - It contains all the top-level fields of the document, with their raw JSON value as value.
+///
+/// # Panics
+///
+/// - If the document contains a top-level field that is not present in `fields_ids_map`.
+///
+pub fn write_to_obkv<'s, 'a, 'map, 'buffer>(
+    document: &'s impl Document<'s>,
+    vector_document: Option<&'s impl VectorDocument<'s>>,
+    fields_ids_map: &'a mut GlobalFieldsIdsMap<'map>,
+    mut document_buffer: &'a mut bumpalo::collections::Vec<'buffer, u8>,
+) -> Result<&'a KvReaderFieldId>
+where
+    's: 'a,
+{
+    // will be used in 'inject_vectors
+    let vectors_value: Box<RawValue>;
+
+    document_buffer.clear();
+    let mut unordered_field_buffer = Vec::new();
+    unordered_field_buffer.clear();
+
+    let mut writer = KvWriterFieldId::new(&mut document_buffer);
+
+    for res in document.iter_top_level_fields() {
+        let (field_name, value) = res?;
+        let field_id =
+            fields_ids_map.id_or_insert(field_name).ok_or(UserError::AttributeLimitReached)?;
+        unordered_field_buffer.push((field_id, value));
+    }
+
+    'inject_vectors: {
+        let Some(vector_document) = vector_document else { break 'inject_vectors };
+
+        let mut vectors = BTreeMap::new();
+        for res in vector_document.iter_vectors() {
+            let (name, entry) = res?;
+            if entry.has_configured_embedder {
+                continue; // we don't write vectors with configured embedder in documents
+            }
+            vectors.insert(
+                name,
+                if entry.implicit {
+                    serde_json::json!(entry.embeddings)
+                } else {
+                    serde_json::json!({
+                        "regenerate": entry.regenerate,
+                        // TODO: consider optimizing the shape of embedders here to store an array of f32 rather than a JSON object
+                        "embeddings": entry.embeddings,
+                    })
+                },
+            );
+        }
+
+        if vectors.is_empty() {
+            break 'inject_vectors;
+        }
+
+        let vectors_fid = fields_ids_map
+            .id_or_insert(RESERVED_VECTORS_FIELD_NAME)
+            .ok_or(UserError::AttributeLimitReached)?;
+
+        vectors_value = serde_json::value::to_raw_value(&vectors).unwrap();
+        unordered_field_buffer.push((vectors_fid, &vectors_value));
+    }
+
+    if let Some(geo_value) = document.geo_field()? {
+        let fid = fields_ids_map.id_or_insert("_geo").ok_or(UserError::AttributeLimitReached)?;
+        fields_ids_map.id_or_insert("_geo.lat").ok_or(UserError::AttributeLimitReached)?;
+        fields_ids_map.id_or_insert("_geo.lng").ok_or(UserError::AttributeLimitReached)?;
+        unordered_field_buffer.push((fid, geo_value));
+    }
+
+    unordered_field_buffer.sort_by_key(|(fid, _)| *fid);
+    for (fid, value) in unordered_field_buffer.iter() {
+        writer.insert(*fid, value.get().as_bytes()).unwrap();
+    }
+
+    writer.finish().unwrap();
+    Ok(KvReaderFieldId::from_slice(document_buffer))
+}
+
+pub type Entry<'doc> = (&'doc str, &'doc RawValue);
+
+#[derive(Debug)]
+pub struct Versions<'doc> {
+    data: RawMap<'doc>,
+}
+
+impl<'doc> Versions<'doc> {
+    pub fn multiple(
+        mut versions: impl Iterator<Item = Result<RawMap<'doc>>>,
+    ) -> Result<Option<Self>> {
+        let Some(data) = versions.next() else { return Ok(None) };
+        let mut data = data?;
+        for future_version in versions {
+            let future_version = future_version?;
+            for (field, value) in future_version {
+                data.insert(field, value);
+            }
+        }
+        Ok(Some(Self::single(data)))
+    }
+
+    pub fn single(version: RawMap<'doc>) -> Self {
+        Self { data: version }
+    }
+
+    pub fn iter_top_level_fields(&self) -> impl Iterator<Item = (&'doc str, &'doc RawValue)> + '_ {
+        self.data.iter().filter(|(k, _)| *k != RESERVED_VECTORS_FIELD_NAME && *k != "_geo")
+    }
+
+    pub fn vectors_field(&self) -> Option<&'doc RawValue> {
+        self.data.get(RESERVED_VECTORS_FIELD_NAME)
+    }
+
+    pub fn geo_field(&self) -> Option<&'doc RawValue> {
+        self.data.get("_geo")
+    }
+
+    pub fn len(&self) -> usize {
+        self.data.len()
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.data.is_empty()
+    }
+
+    pub fn top_level_field(&self, k: &str) -> Option<&'doc RawValue> {
+        if k == RESERVED_VECTORS_FIELD_NAME || k == "_geo" {
+            return None;
+        }
+        self.data.get(k)
+    }
+}

crates/milli/src/update/new/document_change.rs

@@ -0,0 +1,203 @@
+use bumpalo::Bump;
+use heed::RoTxn;
+
+use super::document::{DocumentFromDb, DocumentFromVersions, MergedDocument, Versions};
+use super::vector_document::{
+    MergedVectorDocument, VectorDocumentFromDb, VectorDocumentFromVersions,
+};
+use crate::documents::FieldIdMapper;
+use crate::vector::EmbeddingConfigs;
+use crate::{DocumentId, Index, Result};
+
+pub enum DocumentChange<'doc> {
+    Deletion(Deletion<'doc>),
+    Update(Update<'doc>),
+    Insertion(Insertion<'doc>),
+}
+
+pub struct Deletion<'doc> {
+    docid: DocumentId,
+    external_document_id: &'doc str,
+}
+
+pub struct Update<'doc> {
+    docid: DocumentId,
+    external_document_id: &'doc str,
+    new: Versions<'doc>,
+    has_deletion: bool,
+}
+
+pub struct Insertion<'doc> {
+    docid: DocumentId,
+    external_document_id: &'doc str,
+    new: Versions<'doc>,
+}
+
+impl<'doc> DocumentChange<'doc> {
+    pub fn docid(&self) -> DocumentId {
+        match &self {
+            Self::Deletion(inner) => inner.docid(),
+            Self::Update(inner) => inner.docid(),
+            Self::Insertion(inner) => inner.docid(),
+        }
+    }
+
+    pub fn external_docid(&self) -> &'doc str {
+        match self {
+            DocumentChange::Deletion(deletion) => deletion.external_document_id(),
+            DocumentChange::Update(update) => update.external_document_id(),
+            DocumentChange::Insertion(insertion) => insertion.external_document_id(),
+        }
+    }
+}
+
+impl<'doc> Deletion<'doc> {
+    pub fn create(docid: DocumentId, external_document_id: &'doc str) -> Self {
+        Self { docid, external_document_id }
+    }
+
+    pub fn docid(&self) -> DocumentId {
+        self.docid
+    }
+
+    pub fn external_document_id(&self) -> &'doc str {
+        self.external_document_id
+    }
+
+    pub fn current<'a, Mapper: FieldIdMapper>(
+        &self,
+        rtxn: &'a RoTxn,
+        index: &'a Index,
+        mapper: &'a Mapper,
+    ) -> Result<DocumentFromDb<'a, Mapper>> {
+        Ok(DocumentFromDb::new(self.docid, rtxn, index, mapper)?.ok_or(
+            crate::error::UserError::UnknownInternalDocumentId { document_id: self.docid },
+        )?)
+    }
+}
+
+impl<'doc> Insertion<'doc> {
+    pub fn create(docid: DocumentId, external_document_id: &'doc str, new: Versions<'doc>) -> Self {
+        Insertion { docid, external_document_id, new }
+    }
+
+    pub fn docid(&self) -> DocumentId {
+        self.docid
+    }
+
+    pub fn external_document_id(&self) -> &'doc str {
+        self.external_document_id
+    }
+    pub fn inserted(&self) -> DocumentFromVersions<'_, 'doc> {
+        DocumentFromVersions::new(&self.new)
+    }
+
+    pub fn inserted_vectors(
+        &self,
+        doc_alloc: &'doc Bump,
+        embedders: &'doc EmbeddingConfigs,
+    ) -> Result<Option<VectorDocumentFromVersions<'doc>>> {
+        VectorDocumentFromVersions::new(self.external_document_id, &self.new, doc_alloc, embedders)
+    }
+}
+
+impl<'doc> Update<'doc> {
+    pub fn create(
+        docid: DocumentId,
+        external_document_id: &'doc str,
+        new: Versions<'doc>,
+        has_deletion: bool,
+    ) -> Self {
+        Update { docid, new, external_document_id, has_deletion }
+    }
+
+    pub fn docid(&self) -> DocumentId {
+        self.docid
+    }
+
+    pub fn external_document_id(&self) -> &'doc str {
+        self.external_document_id
+    }
+    pub fn current<'a, Mapper: FieldIdMapper>(
+        &self,
+        rtxn: &'a RoTxn,
+        index: &'a Index,
+        mapper: &'a Mapper,
+    ) -> Result<DocumentFromDb<'a, Mapper>> {
+        Ok(DocumentFromDb::new(self.docid, rtxn, index, mapper)?.ok_or(
+            crate::error::UserError::UnknownInternalDocumentId { document_id: self.docid },
+        )?)
+    }
+
+    pub fn current_vectors<'a, Mapper: FieldIdMapper>(
+        &self,
+        rtxn: &'a RoTxn,
+        index: &'a Index,
+        mapper: &'a Mapper,
+        doc_alloc: &'a Bump,
+    ) -> Result<VectorDocumentFromDb<'a>> {
+        Ok(VectorDocumentFromDb::new(self.docid, index, rtxn, mapper, doc_alloc)?.ok_or(
+            crate::error::UserError::UnknownInternalDocumentId { document_id: self.docid },
+        )?)
+    }
+
+    pub fn updated(&self) -> DocumentFromVersions<'_, 'doc> {
+        DocumentFromVersions::new(&self.new)
+    }
+
+    pub fn merged<'t, Mapper: FieldIdMapper>(
+        &self,
+        rtxn: &'t RoTxn,
+        index: &'t Index,
+        mapper: &'t Mapper,
+    ) -> Result<MergedDocument<'_, 'doc, 't, Mapper>> {
+        if self.has_deletion {
+            Ok(MergedDocument::without_db(DocumentFromVersions::new(&self.new)))
+        } else {
+            MergedDocument::with_db(
+                self.docid,
+                rtxn,
+                index,
+                mapper,
+                DocumentFromVersions::new(&self.new),
+            )
+        }
+    }
+
+    pub fn updated_vectors(
+        &self,
+        doc_alloc: &'doc Bump,
+        embedders: &'doc EmbeddingConfigs,
+    ) -> Result<Option<VectorDocumentFromVersions<'doc>>> {
+        VectorDocumentFromVersions::new(self.external_document_id, &self.new, doc_alloc, embedders)
+    }
+
+    pub fn merged_vectors<Mapper: FieldIdMapper>(
+        &self,
+        rtxn: &'doc RoTxn,
+        index: &'doc Index,
+        mapper: &'doc Mapper,
+        doc_alloc: &'doc Bump,
+        embedders: &'doc EmbeddingConfigs,
+    ) -> Result<Option<MergedVectorDocument<'doc>>> {
+        if self.has_deletion {
+            MergedVectorDocument::without_db(
+                self.external_document_id,
+                &self.new,
+                doc_alloc,
+                embedders,
+            )
+        } else {
+            MergedVectorDocument::with_db(
+                self.docid,
+                self.external_document_id,
+                index,
+                rtxn,
+                mapper,
+                &self.new,
+                doc_alloc,
+                embedders,
+            )
+        }
+    }
+}

crates/milli/src/update/new/extract/cache.rs

@@ -0,0 +1,722 @@
+//! # How the Merge Algorithm works
+//!
+//! Each extractor create #Threads caches and balances the entries
+//! based on the hash of the keys. To do that we can use the
+//! hashbrown::hash_map::RawEntryBuilderMut::from_key_hashed_nocheck.
+//! This way we can compute the hash on our own, decide on the cache to
+//! target, and insert it into the right HashMap.
+//!
+//! #Thread  -> caches
+//! t1       -> [t1c1, t1c2, t1c3]
+//! t2       -> [t2c1, t2c2, t2c3]
+//! t3       -> [t3c1, t3c2, t3c3]
+//!
+//! When the extractors are done filling the caches, we want to merge
+//! the content of all the caches. We do a transpose and each thread is
+//! assigned the associated cache. By doing that we know that every key
+//! is put in a known cache and will collide with keys in the other
+//! caches of the other threads.
+//!
+//! #Thread  -> caches
+//! t1       -> [t1c1, t2c1, t3c1]
+//! t2       -> [t1c2, t2c2, t3c2]
+//! t3       -> [t1c3, t2c3, t3c3]
+//!
+//! When we encountered a miss in the other caches we must still try
+//! to find it in the spilled entries. This is the reason why we use
+//! a grenad sorter/reader so that we can seek "efficiently" for a key.
+//!
+//! ## More Detailled Algorithm
+//!
+//! Each sub-cache has an in-memory HashMap and some spilled
+//! lexicographically ordered entries on disk (grenad). We first iterate
+//! over the spilled entries of all the caches at once by using a merge
+//! join algorithm. This algorithm will merge the entries by using its
+//! merge function.
+//!
+//! Everytime a merged entry is emited by the merge join algorithm we also
+//! fetch the value from the other in-memory caches (HashMaps) to finish
+//! the merge. Everytime we retrieve an entry from the in-memory caches
+//! we mark them with a tombstone for later.
+//!
+//! Once we are done with the spilled entries we iterate over the in-memory
+//! HashMaps. We iterate over the first one, retrieve the content from the
+//! other onces and mark them with a tombstone again. We also make sure
+//! to ignore the dead (tombstoned) ones.
+//!
+//! ## Memory Control
+//!
+//! We can detect that there are no more memory available when the
+//! bump allocator reaches a threshold. When this is the case we
+//! freeze the cache. There is one bump allocator by thread and the
+//! memory must be well balanced as we manage one type of extraction
+//! at a time with well-balanced documents.
+//!
+//! It means that the unknown new keys added to the
+//! cache are directly spilled to disk: basically a key followed by a
+//! del/add bitmap. For the known keys we can keep modifying them in
+//! the materialized version in the cache: update the del/add bitmaps.
+//!
+//! For now we can use a grenad sorter for spilling even thought I think
+//! it's not the most efficient way (too many files open, sorting entries).
+
+use std::cmp::Ordering;
+use std::collections::binary_heap::PeekMut;
+use std::collections::BinaryHeap;
+use std::fs::File;
+use std::hash::BuildHasher;
+use std::io::BufReader;
+use std::{io, iter, mem};
+
+use bumpalo::Bump;
+use grenad::ReaderCursor;
+use hashbrown::hash_map::RawEntryMut;
+use hashbrown::HashMap;
+use raw_collections::bbbul::{BitPacker, BitPacker4x};
+use raw_collections::map::FrozenMap;
+use raw_collections::{Bbbul, FrozenBbbul};
+use roaring::RoaringBitmap;
+use rustc_hash::FxBuildHasher;
+
+use crate::update::del_add::{DelAdd, KvWriterDelAdd};
+use crate::update::new::thread_local::MostlySend;
+use crate::update::new::KvReaderDelAdd;
+use crate::update::MergeDeladdCboRoaringBitmaps;
+use crate::{CboRoaringBitmapCodec, Result};
+
+/// A cache that stores bytes keys associated to CboDelAddRoaringBitmaps.
+///
+/// Internally balances the content over `N` buckets for future merging.
+pub struct BalancedCaches<'extractor> {
+    hasher: FxBuildHasher,
+    alloc: &'extractor Bump,
+    max_memory: Option<usize>,
+    caches: InnerCaches<'extractor>,
+}
+
+enum InnerCaches<'extractor> {
+    Normal(NormalCaches<'extractor>),
+    Spilling(SpillingCaches<'extractor>),
+}
+
+impl<'extractor> BalancedCaches<'extractor> {
+    pub fn new_in(buckets: usize, max_memory: Option<usize>, alloc: &'extractor Bump) -> Self {
+        Self {
+            hasher: FxBuildHasher,
+            max_memory,
+            caches: InnerCaches::Normal(NormalCaches {
+                caches: iter::repeat_with(|| HashMap::with_hasher_in(FxBuildHasher, alloc))
+                    .take(buckets)
+                    .collect(),
+            }),
+            alloc,
+        }
+    }
+
+    fn buckets(&self) -> usize {
+        match &self.caches {
+            InnerCaches::Normal(caches) => caches.caches.len(),
+            InnerCaches::Spilling(caches) => caches.caches.len(),
+        }
+    }
+
+    pub fn insert_del_u32(&mut self, key: &[u8], n: u32) -> Result<()> {
+        if self.max_memory.map_or(false, |mm| self.alloc.allocated_bytes() >= mm) {
+            self.start_spilling()?;
+        }
+
+        let buckets = self.buckets();
+        match &mut self.caches {
+            InnerCaches::Normal(normal) => {
+                normal.insert_del_u32(&self.hasher, self.alloc, buckets, key, n);
+                Ok(())
+            }
+            InnerCaches::Spilling(spilling) => {
+                spilling.insert_del_u32(&self.hasher, self.alloc, buckets, key, n)
+            }
+        }
+    }
+
+    pub fn insert_add_u32(&mut self, key: &[u8], n: u32) -> Result<()> {
+        if self.max_memory.map_or(false, |mm| self.alloc.allocated_bytes() >= mm) {
+            self.start_spilling()?;
+        }
+
+        let buckets = self.buckets();
+        match &mut self.caches {
+            InnerCaches::Normal(normal) => {
+                normal.insert_add_u32(&self.hasher, self.alloc, buckets, key, n);
+                Ok(())
+            }
+            InnerCaches::Spilling(spilling) => {
+                spilling.insert_add_u32(&self.hasher, self.alloc, buckets, key, n)
+            }
+        }
+    }
+
+    /// Make sure the cache is no longer allocating data
+    /// and writes every new and unknow entry to disk.
+    fn start_spilling(&mut self) -> Result<()> {
+        let BalancedCaches { hasher: _, alloc, max_memory: _, caches } = self;
+
+        if let InnerCaches::Normal(normal_caches) = caches {
+            tracing::trace!(
+                "We are spilling after we allocated {} bytes on thread #{}",
+                alloc.allocated_bytes(),
+                rayon::current_thread_index().unwrap_or(0)
+            );
+
+            let allocated: usize = normal_caches.caches.iter().map(|m| m.allocation_size()).sum();
+            tracing::trace!("The last allocated HashMap took {allocated} bytes");
+
+            let dummy = NormalCaches { caches: Vec::new() };
+            let NormalCaches { caches: cache_maps } = mem::replace(normal_caches, dummy);
+            *caches = InnerCaches::Spilling(SpillingCaches::from_cache_maps(cache_maps));
+        }
+
+        Ok(())
+    }
+
+    pub fn freeze(&mut self) -> Result<Vec<FrozenCache<'_, 'extractor>>> {
+        match &mut self.caches {
+            InnerCaches::Normal(NormalCaches { caches }) => caches
+                .iter_mut()
+                .enumerate()
+                .map(|(bucket, map)| {
+                    // safety: we are transmuting the Bbbul into a FrozenBbbul
+                    //         that are the same size.
+                    let map = unsafe {
+                        std::mem::transmute::<
+                            &mut HashMap<
+                                &[u8],
+                                DelAddBbbul<BitPacker4x>, // from this
+                                FxBuildHasher,
+                                &Bump,
+                            >,
+                            &mut HashMap<
+                                &[u8],
+                                FrozenDelAddBbbul<BitPacker4x>, // to that
+                                FxBuildHasher,
+                                &Bump,
+                            >,
+                        >(map)
+                    };
+                    Ok(FrozenCache { bucket, cache: FrozenMap::new(map), spilled: Vec::new() })
+                })
+                .collect(),
+            InnerCaches::Spilling(SpillingCaches { caches, spilled_entries, .. }) => caches
+                .iter_mut()
+                .zip(mem::take(spilled_entries))
+                .enumerate()
+                .map(|(bucket, (map, sorter))| {
+                    let spilled = sorter
+                        .into_reader_cursors()?
+                        .into_iter()
+                        .map(ReaderCursor::into_inner)
+                        .map(BufReader::new)
+                        .map(|bufreader| grenad::Reader::new(bufreader).map_err(Into::into))
+                        .collect::<Result<_>>()?;
+                    // safety: we are transmuting the Bbbul into a FrozenBbbul
+                    //         that are the same size.
+                    let map = unsafe {
+                        std::mem::transmute::<
+                            &mut HashMap<
+                                &[u8],
+                                DelAddBbbul<BitPacker4x>, // from this
+                                FxBuildHasher,
+                                &Bump,
+                            >,
+                            &mut HashMap<
+                                &[u8],
+                                FrozenDelAddBbbul<BitPacker4x>, // to that
+                                FxBuildHasher,
+                                &Bump,
+                            >,
+                        >(map)
+                    };
+                    Ok(FrozenCache { bucket, cache: FrozenMap::new(map), spilled })
+                })
+                .collect(),
+        }
+    }
+}
+
+unsafe impl MostlySend for BalancedCaches<'_> {}
+
+struct NormalCaches<'extractor> {
+    caches: Vec<
+        HashMap<
+            &'extractor [u8],
+            DelAddBbbul<'extractor, BitPacker4x>,
+            FxBuildHasher,
+            &'extractor Bump,
+        >,
+    >,
+}
+
+impl<'extractor> NormalCaches<'extractor> {
+    pub fn insert_del_u32(
+        &mut self,
+        hasher: &FxBuildHasher,
+        alloc: &'extractor Bump,
+        buckets: usize,
+        key: &[u8],
+        n: u32,
+    ) {
+        let hash = hasher.hash_one(key);
+        let bucket = compute_bucket_from_hash(buckets, hash);
+
+        match self.caches[bucket].raw_entry_mut().from_hash(hash, |&k| k == key) {
+            RawEntryMut::Occupied(mut entry) => {
+                entry.get_mut().del.get_or_insert_with(|| Bbbul::new_in(alloc)).insert(n);
+            }
+            RawEntryMut::Vacant(entry) => {
+                entry.insert_hashed_nocheck(
+                    hash,
+                    alloc.alloc_slice_copy(key),
+                    DelAddBbbul::new_del_u32_in(n, alloc),
+                );
+            }
+        }
+    }
+
+    pub fn insert_add_u32(
+        &mut self,
+        hasher: &FxBuildHasher,
+        alloc: &'extractor Bump,
+        buckets: usize,
+        key: &[u8],
+        n: u32,
+    ) {
+        let hash = hasher.hash_one(key);
+        let bucket = compute_bucket_from_hash(buckets, hash);
+        match self.caches[bucket].raw_entry_mut().from_hash(hash, |&k| k == key) {
+            RawEntryMut::Occupied(mut entry) => {
+                entry.get_mut().add.get_or_insert_with(|| Bbbul::new_in(alloc)).insert(n);
+            }
+            RawEntryMut::Vacant(entry) => {
+                entry.insert_hashed_nocheck(
+                    hash,
+                    alloc.alloc_slice_copy(key),
+                    DelAddBbbul::new_add_u32_in(n, alloc),
+                );
+            }
+        }
+    }
+}
+
+struct SpillingCaches<'extractor> {
+    caches: Vec<
+        HashMap<
+            &'extractor [u8],
+            DelAddBbbul<'extractor, BitPacker4x>,
+            FxBuildHasher,
+            &'extractor Bump,
+        >,
+    >,
+    spilled_entries: Vec<grenad::Sorter<MergeDeladdCboRoaringBitmaps>>,
+    deladd_buffer: Vec<u8>,
+    cbo_buffer: Vec<u8>,
+}
+
+impl<'extractor> SpillingCaches<'extractor> {
+    fn from_cache_maps(
+        caches: Vec<
+            HashMap<
+                &'extractor [u8],
+                DelAddBbbul<'extractor, BitPacker4x>,
+                FxBuildHasher,
+                &'extractor Bump,
+            >,
+        >,
+    ) -> SpillingCaches<'extractor> {
+        SpillingCaches {
+            spilled_entries: iter::repeat_with(|| {
+                let mut builder = grenad::SorterBuilder::new(MergeDeladdCboRoaringBitmaps);
+                builder.dump_threshold(0);
+                builder.allow_realloc(false);
+                builder.build()
+            })
+            .take(caches.len())
+            .collect(),
+            caches,
+            deladd_buffer: Vec::new(),
+            cbo_buffer: Vec::new(),
+        }
+    }
+
+    pub fn insert_del_u32(
+        &mut self,
+        hasher: &FxBuildHasher,
+        alloc: &'extractor Bump,
+        buckets: usize,
+        key: &[u8],
+        n: u32,
+    ) -> Result<()> {
+        let hash = hasher.hash_one(key);
+        let bucket = compute_bucket_from_hash(buckets, hash);
+        match self.caches[bucket].raw_entry_mut().from_hash(hash, |&k| k == key) {
+            RawEntryMut::Occupied(mut entry) => {
+                entry.get_mut().del.get_or_insert_with(|| Bbbul::new_in(alloc)).insert(n);
+                Ok(())
+            }
+            RawEntryMut::Vacant(_entry) => spill_entry_to_sorter(
+                &mut self.spilled_entries[bucket],
+                &mut self.deladd_buffer,
+                &mut self.cbo_buffer,
+                key,
+                DelAddRoaringBitmap::new_del_u32(n),
+            ),
+        }
+    }
+
+    pub fn insert_add_u32(
+        &mut self,
+        hasher: &FxBuildHasher,
+        alloc: &'extractor Bump,
+        buckets: usize,
+        key: &[u8],
+        n: u32,
+    ) -> Result<()> {
+        let hash = hasher.hash_one(key);
+        let bucket = compute_bucket_from_hash(buckets, hash);
+        match self.caches[bucket].raw_entry_mut().from_hash(hash, |&k| k == key) {
+            RawEntryMut::Occupied(mut entry) => {
+                entry.get_mut().add.get_or_insert_with(|| Bbbul::new_in(alloc)).insert(n);
+                Ok(())
+            }
+            RawEntryMut::Vacant(_entry) => spill_entry_to_sorter(
+                &mut self.spilled_entries[bucket],
+                &mut self.deladd_buffer,
+                &mut self.cbo_buffer,
+                key,
+                DelAddRoaringBitmap::new_add_u32(n),
+            ),
+        }
+    }
+}
+
+#[inline]
+fn compute_bucket_from_hash(buckets: usize, hash: u64) -> usize {
+    hash as usize % buckets
+}
+
+fn spill_entry_to_sorter(
+    spilled_entries: &mut grenad::Sorter<MergeDeladdCboRoaringBitmaps>,
+    deladd_buffer: &mut Vec<u8>,
+    cbo_buffer: &mut Vec<u8>,
+    key: &[u8],
+    deladd: DelAddRoaringBitmap,
+) -> Result<()> {
+    deladd_buffer.clear();
+    let mut value_writer = KvWriterDelAdd::new(deladd_buffer);
+
+    match deladd {
+        DelAddRoaringBitmap { del: Some(del), add: None } => {
+            cbo_buffer.clear();
+            CboRoaringBitmapCodec::serialize_into(&del, cbo_buffer);
+            value_writer.insert(DelAdd::Deletion, &cbo_buffer)?;
+        }
+        DelAddRoaringBitmap { del: None, add: Some(add) } => {
+            cbo_buffer.clear();
+            CboRoaringBitmapCodec::serialize_into(&add, cbo_buffer);
+            value_writer.insert(DelAdd::Addition, &cbo_buffer)?;
+        }
+        DelAddRoaringBitmap { del: Some(del), add: Some(add) } => {
+            cbo_buffer.clear();
+            CboRoaringBitmapCodec::serialize_into(&del, cbo_buffer);
+            value_writer.insert(DelAdd::Deletion, &cbo_buffer)?;
+
+            cbo_buffer.clear();
+            CboRoaringBitmapCodec::serialize_into(&add, cbo_buffer);
+            value_writer.insert(DelAdd::Addition, &cbo_buffer)?;
+        }
+        DelAddRoaringBitmap { del: None, add: None } => return Ok(()),
+    }
+
+    let bytes = value_writer.into_inner().unwrap();
+    spilled_entries.insert(key, bytes).map_err(Into::into)
+}
+
+pub struct FrozenCache<'a, 'extractor> {
+    bucket: usize,
+    cache: FrozenMap<
+        'a,
+        'extractor,
+        &'extractor [u8],
+        FrozenDelAddBbbul<'extractor, BitPacker4x>,
+        FxBuildHasher,
+    >,
+    spilled: Vec<grenad::Reader<BufReader<File>>>,
+}
+
+pub fn transpose_and_freeze_caches<'a, 'extractor>(
+    caches: &'a mut [BalancedCaches<'extractor>],
+) -> Result<Vec<Vec<FrozenCache<'a, 'extractor>>>> {
+    let width = caches.first().map(BalancedCaches::buckets).unwrap_or(0);
+    let mut bucket_caches: Vec<_> = iter::repeat_with(Vec::new).take(width).collect();
+
+    for thread_cache in caches {
+        for frozen in thread_cache.freeze()? {
+            bucket_caches[frozen.bucket].push(frozen);
+        }
+    }
+
+    Ok(bucket_caches)
+}
+
+/// Merges the caches that must be all associated to the same bucket.
+///
+/// # Panics
+///
+/// - If the bucket IDs in these frozen caches are not exactly the same.
+pub fn merge_caches<F>(frozen: Vec<FrozenCache>, mut f: F) -> Result<()>
+where
+    F: for<'a> FnMut(&'a [u8], DelAddRoaringBitmap) -> Result<()>,
+{
+    let mut maps = Vec::new();
+    let mut readers = Vec::new();
+    let mut current_bucket = None;
+    for FrozenCache { bucket, cache, ref mut spilled } in frozen {
+        assert_eq!(*current_bucket.get_or_insert(bucket), bucket);
+        maps.push(cache);
+        readers.append(spilled);
+    }
+
+    // First manage the spilled entries by looking into the HashMaps,
+    // merge them and mark them as dummy.
+    let mut heap = BinaryHeap::new();
+    for (source_index, source) in readers.into_iter().enumerate() {
+        let mut cursor = source.into_cursor()?;
+        if cursor.move_on_next()?.is_some() {
+            heap.push(Entry { cursor, source_index });
+        }
+    }
+
+    loop {
+        let mut first_entry = match heap.pop() {
+            Some(entry) => entry,
+            None => break,
+        };
+
+        let (first_key, first_value) = match first_entry.cursor.current() {
+            Some((key, value)) => (key, value),
+            None => break,
+        };
+
+        let mut output = DelAddRoaringBitmap::from_bytes(first_value)?;
+        while let Some(mut entry) = heap.peek_mut() {
+            if let Some((key, _value)) = entry.cursor.current() {
+                if first_key == key {
+                    let new = DelAddRoaringBitmap::from_bytes(first_value)?;
+                    output = output.merge(new);
+                    // When we are done we the current value of this entry move make
+                    // it move forward and let the heap reorganize itself (on drop)
+                    if entry.cursor.move_on_next()?.is_none() {
+                        PeekMut::pop(entry);
+                    }
+                } else {
+                    break;
+                }
+            }
+        }
+
+        // Once we merged all of the spilled bitmaps we must also
+        // fetch the entries from the non-spilled entries (the HashMaps).
+        for (map_index, map) in maps.iter_mut().enumerate() {
+            if first_entry.source_index != map_index {
+                if let Some(new) = map.get_mut(first_key) {
+                    output.union_and_clear_bbbul(new);
+                }
+            }
+        }
+
+        // We send the merged entry outside.
+        (f)(first_key, output)?;
+
+        // Don't forget to put the first entry back into the heap.
+        if first_entry.cursor.move_on_next()?.is_some() {
+            heap.push(first_entry)
+        }
+    }
+
+    // Then manage the content on the HashMap entries that weren't taken (mem::take).
+    while let Some(mut map) = maps.pop() {
+        for (key, bbbul) in map.iter_mut() {
+            // Make sure we don't try to work with entries already managed by the spilled
+            if bbbul.is_empty() {
+                continue;
+            }
+
+            let mut output = DelAddRoaringBitmap::empty();
+            output.union_and_clear_bbbul(bbbul);
+
+            for rhs in maps.iter_mut() {
+                if let Some(new) = rhs.get_mut(key) {
+                    output.union_and_clear_bbbul(new);
+                }
+            }
+
+            // We send the merged entry outside.
+            (f)(key, output)?;
+        }
+    }
+
+    Ok(())
+}
+
+struct Entry<R> {
+    cursor: ReaderCursor<R>,
+    source_index: usize,
+}
+
+impl<R> Ord for Entry<R> {
+    fn cmp(&self, other: &Entry<R>) -> Ordering {
+        let skey = self.cursor.current().map(|(k, _)| k);
+        let okey = other.cursor.current().map(|(k, _)| k);
+        skey.cmp(&okey).then(self.source_index.cmp(&other.source_index)).reverse()
+    }
+}
+
+impl<R> Eq for Entry<R> {}
+
+impl<R> PartialEq for Entry<R> {
+    fn eq(&self, other: &Entry<R>) -> bool {
+        self.cmp(other) == Ordering::Equal
+    }
+}
+
+impl<R> PartialOrd for Entry<R> {
+    fn partial_cmp(&self, other: &Entry<R>) -> Option<Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+pub struct DelAddBbbul<'bump, B> {
+    pub del: Option<Bbbul<'bump, B>>,
+    pub add: Option<Bbbul<'bump, B>>,
+}
+
+impl<'bump, B: BitPacker> DelAddBbbul<'bump, B> {
+    pub fn new_del_u32_in(n: u32, bump: &'bump Bump) -> Self {
+        let mut bbbul = Bbbul::new_in(bump);
+        bbbul.insert(n);
+        DelAddBbbul { del: Some(bbbul), add: None }
+    }
+
+    pub fn new_add_u32_in(n: u32, bump: &'bump Bump) -> Self {
+        let mut bbbul = Bbbul::new_in(bump);
+        bbbul.insert(n);
+        DelAddBbbul { del: None, add: Some(bbbul) }
+    }
+}
+
+pub struct FrozenDelAddBbbul<'bump, B> {
+    pub del: Option<FrozenBbbul<'bump, B>>,
+    pub add: Option<FrozenBbbul<'bump, B>>,
+}
+
+impl<'bump, B> FrozenDelAddBbbul<'bump, B> {
+    fn is_empty(&self) -> bool {
+        self.del.is_none() && self.add.is_none()
+    }
+}
+
+#[derive(Debug, Default, Clone)]
+pub struct DelAddRoaringBitmap {
+    pub del: Option<RoaringBitmap>,
+    pub add: Option<RoaringBitmap>,
+}
+
+impl DelAddRoaringBitmap {
+    fn from_bytes(bytes: &[u8]) -> io::Result<DelAddRoaringBitmap> {
+        let reader = KvReaderDelAdd::from_slice(bytes);
+
+        let del = match reader.get(DelAdd::Deletion) {
+            Some(bytes) => CboRoaringBitmapCodec::deserialize_from(bytes).map(Some)?,
+            None => None,
+        };
+
+        let add = match reader.get(DelAdd::Addition) {
+            Some(bytes) => CboRoaringBitmapCodec::deserialize_from(bytes).map(Some)?,
+            None => None,
+        };
+
+        Ok(DelAddRoaringBitmap { del, add })
+    }
+
+    pub fn empty() -> DelAddRoaringBitmap {
+        DelAddRoaringBitmap { del: None, add: None }
+    }
+
+    pub fn insert_del_u32(&mut self, n: u32) {
+        self.del.get_or_insert_with(RoaringBitmap::new).insert(n);
+    }
+
+    pub fn insert_add_u32(&mut self, n: u32) {
+        self.add.get_or_insert_with(RoaringBitmap::new).insert(n);
+    }
+
+    pub fn new_del_u32(n: u32) -> Self {
+        DelAddRoaringBitmap { del: Some(RoaringBitmap::from([n])), add: None }
+    }
+
+    pub fn new_add_u32(n: u32) -> Self {
+        DelAddRoaringBitmap { del: None, add: Some(RoaringBitmap::from([n])) }
+    }
+
+    pub fn union_and_clear_bbbul<B: BitPacker>(&mut self, bbbul: &mut FrozenDelAddBbbul<'_, B>) {
+        let FrozenDelAddBbbul { del, add } = bbbul;
+
+        if let Some(ref mut bbbul) = del.take() {
+            let del = self.del.get_or_insert_with(RoaringBitmap::new);
+            let mut iter = bbbul.iter_and_clear();
+            while let Some(block) = iter.next_block() {
+                let iter = block.iter().copied();
+                let block = RoaringBitmap::from_sorted_iter(iter).unwrap();
+                *del |= block;
+            }
+        }
+
+        if let Some(ref mut bbbul) = add.take() {
+            let add = self.add.get_or_insert_with(RoaringBitmap::new);
+            let mut iter = bbbul.iter_and_clear();
+            while let Some(block) = iter.next_block() {
+                let iter = block.iter().copied();
+                let block = RoaringBitmap::from_sorted_iter(iter).unwrap();
+                *add |= block;
+            }
+        }
+    }
+
+    pub fn merge(self, rhs: DelAddRoaringBitmap) -> DelAddRoaringBitmap {
+        let DelAddRoaringBitmap { del, add } = self;
+        let DelAddRoaringBitmap { del: ndel, add: nadd } = rhs;
+
+        let del = match (del, ndel) {
+            (None, None) => None,
+            (None, Some(del)) | (Some(del), None) => Some(del),
+            (Some(del), Some(ndel)) => Some(del | ndel),
+        };
+
+        let add = match (add, nadd) {
+            (None, None) => None,
+            (None, Some(add)) | (Some(add), None) => Some(add),
+            (Some(add), Some(nadd)) => Some(add | nadd),
+        };
+
+        DelAddRoaringBitmap { del, add }
+    }
+
+    pub fn apply_to(&self, documents_ids: &mut RoaringBitmap) {
+        let DelAddRoaringBitmap { del, add } = self;
+
+        if let Some(del) = del {
+            *documents_ids -= del;
+        }
+
+        if let Some(add) = add {
+            *documents_ids |= add;
+        }
+    }
+}

crates/milli/src/update/new/extract/documents.rs

@@ -0,0 +1,150 @@
+use std::cell::RefCell;
+
+use bumpalo::Bump;
+use hashbrown::HashMap;
+
+use super::DelAddRoaringBitmap;
+use crate::update::new::channel::DocumentsSender;
+use crate::update::new::document::{write_to_obkv, Document as _};
+use crate::update::new::indexer::document_changes::{DocumentChangeContext, Extractor};
+use crate::update::new::ref_cell_ext::RefCellExt as _;
+use crate::update::new::thread_local::FullySend;
+use crate::update::new::DocumentChange;
+use crate::vector::EmbeddingConfigs;
+use crate::Result;
+pub struct DocumentsExtractor<'a> {
+    document_sender: &'a DocumentsSender<'a>,
+    embedders: &'a EmbeddingConfigs,
+}
+
+impl<'a> DocumentsExtractor<'a> {
+    pub fn new(document_sender: &'a DocumentsSender<'a>, embedders: &'a EmbeddingConfigs) -> Self {
+        Self { document_sender, embedders }
+    }
+}
+
+#[derive(Default)]
+pub struct DocumentExtractorData {
+    pub docids_delta: DelAddRoaringBitmap,
+    pub field_distribution_delta: HashMap<String, i64>,
+}
+
+impl<'a, 'extractor> Extractor<'extractor> for DocumentsExtractor<'a> {
+    type Data = FullySend<RefCell<DocumentExtractorData>>;
+
+    fn init_data(&self, _extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
+        Ok(FullySend(Default::default()))
+    }
+
+    fn process<'doc>(
+        &self,
+        changes: impl Iterator<Item = Result<DocumentChange<'doc>>>,
+        context: &DocumentChangeContext<Self::Data>,
+    ) -> Result<()> {
+        let mut document_buffer = bumpalo::collections::Vec::new_in(&context.doc_alloc);
+        let mut document_extractor_data = context.data.0.borrow_mut_or_yield();
+
+        for change in changes {
+            let change = change?;
+            // **WARNING**: the exclusive borrow on `new_fields_ids_map` needs to be taken **inside** of the `for change in changes` loop
+            // Otherwise, `BorrowMutError` will occur for document changes that also need the new_fields_ids_map (e.g.: UpdateByFunction)
+            let mut new_fields_ids_map = context.new_fields_ids_map.borrow_mut_or_yield();
+
+            let external_docid = change.external_docid().to_owned();
+
+            // document but we need to create a function that collects and compresses documents.
+            match change {
+                DocumentChange::Deletion(deletion) => {
+                    let docid = deletion.docid();
+                    let content = deletion.current(
+                        &context.rtxn,
+                        context.index,
+                        &context.db_fields_ids_map,
+                    )?;
+                    let geo_iter =
+                        content.geo_field().transpose().map(|res| res.map(|rv| ("_geo", rv)));
+                    for res in content.iter_top_level_fields().chain(geo_iter) {
+                        let (f, _) = res?;
+                        let entry = document_extractor_data
+                            .field_distribution_delta
+                            .entry_ref(f)
+                            .or_default();
+                        *entry -= 1;
+                    }
+                    document_extractor_data.docids_delta.insert_del_u32(docid);
+                    self.document_sender.delete(docid, external_docid).unwrap();
+                }
+                DocumentChange::Update(update) => {
+                    let docid = update.docid();
+                    let content =
+                        update.current(&context.rtxn, context.index, &context.db_fields_ids_map)?;
+                    let geo_iter =
+                        content.geo_field().transpose().map(|res| res.map(|rv| ("_geo", rv)));
+                    for res in content.iter_top_level_fields().chain(geo_iter) {
+                        let (f, _) = res?;
+                        let entry = document_extractor_data
+                            .field_distribution_delta
+                            .entry_ref(f)
+                            .or_default();
+                        *entry -= 1;
+                    }
+                    let content = update.updated();
+                    let geo_iter =
+                        content.geo_field().transpose().map(|res| res.map(|rv| ("_geo", rv)));
+                    for res in content.iter_top_level_fields().chain(geo_iter) {
+                        let (f, _) = res?;
+                        let entry = document_extractor_data
+                            .field_distribution_delta
+                            .entry_ref(f)
+                            .or_default();
+                        *entry += 1;
+                    }
+
+                    let content =
+                        update.merged(&context.rtxn, context.index, &context.db_fields_ids_map)?;
+                    let vector_content = update.merged_vectors(
+                        &context.rtxn,
+                        context.index,
+                        &context.db_fields_ids_map,
+                        &context.doc_alloc,
+                        self.embedders,
+                    )?;
+                    let content = write_to_obkv(
+                        &content,
+                        vector_content.as_ref(),
+                        &mut new_fields_ids_map,
+                        &mut document_buffer,
+                    )?;
+                    self.document_sender.uncompressed(docid, external_docid, content).unwrap();
+                }
+                DocumentChange::Insertion(insertion) => {
+                    let docid = insertion.docid();
+                    let content = insertion.inserted();
+                    let geo_iter =
+                        content.geo_field().transpose().map(|res| res.map(|rv| ("_geo", rv)));
+                    for res in content.iter_top_level_fields().chain(geo_iter) {
+                        let (f, _) = res?;
+                        let entry = document_extractor_data
+                            .field_distribution_delta
+                            .entry_ref(f)
+                            .or_default();
+                        *entry += 1;
+                    }
+                    let inserted_vectors =
+                        insertion.inserted_vectors(&context.doc_alloc, self.embedders)?;
+                    let content = write_to_obkv(
+                        &content,
+                        inserted_vectors.as_ref(),
+                        &mut new_fields_ids_map,
+                        &mut document_buffer,
+                    )?;
+                    document_extractor_data.docids_delta.insert_add_u32(docid);
+                    self.document_sender.uncompressed(docid, external_docid, content).unwrap();
+                    // extracted_dictionary_sender.send(self, dictionary: &[u8]);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}

crates/milli/src/update/new/extract/faceted/extract_facets.rs

@@ -0,0 +1,418 @@
+use std::cell::RefCell;
+use std::collections::HashSet;
+use std::ops::DerefMut as _;
+
+use bumpalo::collections::Vec as BVec;
+use bumpalo::Bump;
+use hashbrown::HashMap;
+use heed::RoTxn;
+use serde_json::Value;
+
+use super::super::cache::BalancedCaches;
+use super::facet_document::extract_document_facets;
+use super::FacetKind;
+use crate::heed_codec::facet::OrderedF64Codec;
+use crate::update::del_add::DelAdd;
+use crate::update::new::channel::FieldIdDocidFacetSender;
+use crate::update::new::extract::perm_json_p;
+use crate::update::new::indexer::document_changes::{
+    extract, DocumentChangeContext, DocumentChanges, Extractor, IndexingContext, Progress,
+};
+use crate::update::new::ref_cell_ext::RefCellExt as _;
+use crate::update::new::steps::Step;
+use crate::update::new::thread_local::{FullySend, ThreadLocal};
+use crate::update::new::DocumentChange;
+use crate::update::GrenadParameters;
+use crate::{DocumentId, FieldId, Index, Result, MAX_FACET_VALUE_LENGTH};
+
+pub struct FacetedExtractorData<'a> {
+    attributes_to_extract: &'a [&'a str],
+    sender: &'a FieldIdDocidFacetSender<'a>,
+    grenad_parameters: GrenadParameters,
+    buckets: usize,
+}
+
+impl<'a, 'extractor> Extractor<'extractor> for FacetedExtractorData<'a> {
+    type Data = RefCell<BalancedCaches<'extractor>>;
+
+    fn init_data(&self, extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
+        Ok(RefCell::new(BalancedCaches::new_in(
+            self.buckets,
+            self.grenad_parameters.max_memory_by_thread(),
+            extractor_alloc,
+        )))
+    }
+
+    fn process<'doc>(
+        &self,
+        changes: impl Iterator<Item = Result<DocumentChange<'doc>>>,
+        context: &DocumentChangeContext<Self::Data>,
+    ) -> Result<()> {
+        for change in changes {
+            let change = change?;
+            FacetedDocidsExtractor::extract_document_change(
+                context,
+                self.attributes_to_extract,
+                change,
+                self.sender,
+            )?
+        }
+        Ok(())
+    }
+}
+
+pub struct FacetedDocidsExtractor;
+
+impl FacetedDocidsExtractor {
+    fn extract_document_change(
+        context: &DocumentChangeContext<RefCell<BalancedCaches>>,
+        attributes_to_extract: &[&str],
+        document_change: DocumentChange,
+        sender: &FieldIdDocidFacetSender,
+    ) -> Result<()> {
+        let index = &context.index;
+        let rtxn = &context.rtxn;
+        let mut new_fields_ids_map = context.new_fields_ids_map.borrow_mut_or_yield();
+        let mut cached_sorter = context.data.borrow_mut_or_yield();
+        let mut del_add_facet_value = DelAddFacetValue::new(&context.doc_alloc);
+        let docid = document_change.docid();
+        let res = match document_change {
+            DocumentChange::Deletion(inner) => extract_document_facets(
+                attributes_to_extract,
+                inner.current(rtxn, index, context.db_fields_ids_map)?,
+                inner.external_document_id(),
+                new_fields_ids_map.deref_mut(),
+                &mut |fid, depth, value| {
+                    Self::facet_fn_with_options(
+                        &context.doc_alloc,
+                        cached_sorter.deref_mut(),
+                        BalancedCaches::insert_del_u32,
+                        &mut del_add_facet_value,
+                        DelAddFacetValue::insert_del,
+                        docid,
+                        fid,
+                        depth,
+                        value,
+                    )
+                },
+            ),
+            DocumentChange::Update(inner) => {
+                extract_document_facets(
+                    attributes_to_extract,
+                    inner.current(rtxn, index, context.db_fields_ids_map)?,
+                    inner.external_document_id(),
+                    new_fields_ids_map.deref_mut(),
+                    &mut |fid, depth, value| {
+                        Self::facet_fn_with_options(
+                            &context.doc_alloc,
+                            cached_sorter.deref_mut(),
+                            BalancedCaches::insert_del_u32,
+                            &mut del_add_facet_value,
+                            DelAddFacetValue::insert_del,
+                            docid,
+                            fid,
+                            depth,
+                            value,
+                        )
+                    },
+                )?;
+
+                extract_document_facets(
+                    attributes_to_extract,
+                    inner.merged(rtxn, index, context.db_fields_ids_map)?,
+                    inner.external_document_id(),
+                    new_fields_ids_map.deref_mut(),
+                    &mut |fid, depth, value| {
+                        Self::facet_fn_with_options(
+                            &context.doc_alloc,
+                            cached_sorter.deref_mut(),
+                            BalancedCaches::insert_add_u32,
+                            &mut del_add_facet_value,
+                            DelAddFacetValue::insert_add,
+                            docid,
+                            fid,
+                            depth,
+                            value,
+                        )
+                    },
+                )
+            }
+            DocumentChange::Insertion(inner) => extract_document_facets(
+                attributes_to_extract,
+                inner.inserted(),
+                inner.external_document_id(),
+                new_fields_ids_map.deref_mut(),
+                &mut |fid, depth, value| {
+                    Self::facet_fn_with_options(
+                        &context.doc_alloc,
+                        cached_sorter.deref_mut(),
+                        BalancedCaches::insert_add_u32,
+                        &mut del_add_facet_value,
+                        DelAddFacetValue::insert_add,
+                        docid,
+                        fid,
+                        depth,
+                        value,
+                    )
+                },
+            ),
+        };
+
+        del_add_facet_value.send_data(docid, sender, &context.doc_alloc).unwrap();
+        res
+    }
+
+    #[allow(clippy::too_many_arguments)]
+    fn facet_fn_with_options<'extractor, 'doc>(
+        doc_alloc: &'doc Bump,
+        cached_sorter: &mut BalancedCaches<'extractor>,
+        cache_fn: impl Fn(&mut BalancedCaches<'extractor>, &[u8], u32) -> Result<()>,
+        del_add_facet_value: &mut DelAddFacetValue<'doc>,
+        facet_fn: impl Fn(&mut DelAddFacetValue<'doc>, FieldId, BVec<'doc, u8>, FacetKind),
+        docid: DocumentId,
+        fid: FieldId,
+        depth: perm_json_p::Depth,
+        value: &Value,
+    ) -> Result<()> {
+        let mut buffer = BVec::new_in(doc_alloc);
+        // Exists
+        // key: fid
+        buffer.push(FacetKind::Exists as u8);
+        buffer.extend_from_slice(&fid.to_be_bytes());
+        cache_fn(cached_sorter, &buffer, docid)?;
+
+        match value {
+            // Number
+            // key: fid - level - orderedf64 - originalf64
+            Value::Number(number) => {
+                let mut ordered = [0u8; 16];
+                if number
+                    .as_f64()
+                    .and_then(|n| OrderedF64Codec::serialize_into(n, &mut ordered).ok())
+                    .is_some()
+                {
+                    let mut number = BVec::with_capacity_in(16, doc_alloc);
+                    number.extend_from_slice(&ordered);
+                    facet_fn(del_add_facet_value, fid, number, FacetKind::Number);
+
+                    buffer.clear();
+                    buffer.push(FacetKind::Number as u8);
+                    buffer.extend_from_slice(&fid.to_be_bytes());
+                    buffer.push(0); // level 0
+                    buffer.extend_from_slice(&ordered);
+                    cache_fn(cached_sorter, &buffer, docid)
+                } else {
+                    Ok(())
+                }
+            }
+            // String
+            // key: fid - level - truncated_string
+            Value::String(s) if !s.is_empty() => {
+                let mut string = BVec::new_in(doc_alloc);
+                string.extend_from_slice(s.as_bytes());
+                facet_fn(del_add_facet_value, fid, string, FacetKind::String);
+
+                let normalized = crate::normalize_facet(s);
+                let truncated = truncate_str(&normalized);
+                buffer.clear();
+                buffer.push(FacetKind::String as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                buffer.push(0); // level 0
+                buffer.extend_from_slice(truncated.as_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            // Bool is handled as a string
+            Value::Bool(b) => {
+                let b = if *b { "true" } else { "false" };
+                let mut string = BVec::new_in(doc_alloc);
+                string.extend_from_slice(b.as_bytes());
+                facet_fn(del_add_facet_value, fid, string, FacetKind::String);
+
+                buffer.clear();
+                buffer.push(FacetKind::String as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                buffer.push(0); // level 0
+                buffer.extend_from_slice(b.as_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            // Null
+            // key: fid
+            Value::Null if depth == perm_json_p::Depth::OnBaseKey => {
+                buffer.clear();
+                buffer.push(FacetKind::Null as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            // Empty
+            // key: fid
+            Value::Array(a) if a.is_empty() && depth == perm_json_p::Depth::OnBaseKey => {
+                buffer.clear();
+                buffer.push(FacetKind::Empty as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            Value::String(_) if depth == perm_json_p::Depth::OnBaseKey => {
+                buffer.clear();
+                buffer.push(FacetKind::Empty as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            Value::Object(o) if o.is_empty() && depth == perm_json_p::Depth::OnBaseKey => {
+                buffer.clear();
+                buffer.push(FacetKind::Empty as u8);
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                cache_fn(cached_sorter, &buffer, docid)
+            }
+            // Otherwise, do nothing
+            _ => Ok(()),
+        }
+    }
+
+    fn attributes_to_extract<'a>(rtxn: &'a RoTxn, index: &'a Index) -> Result<HashSet<String>> {
+        index.user_defined_faceted_fields(rtxn)
+    }
+}
+
+struct DelAddFacetValue<'doc> {
+    strings: HashMap<(FieldId, BVec<'doc, u8>), DelAdd, hashbrown::DefaultHashBuilder, &'doc Bump>,
+    f64s: HashMap<(FieldId, BVec<'doc, u8>), DelAdd, hashbrown::DefaultHashBuilder, &'doc Bump>,
+}
+
+impl<'doc> DelAddFacetValue<'doc> {
+    fn new(doc_alloc: &'doc Bump) -> Self {
+        Self { strings: HashMap::new_in(doc_alloc), f64s: HashMap::new_in(doc_alloc) }
+    }
+
+    fn insert_add(&mut self, fid: FieldId, value: BVec<'doc, u8>, kind: FacetKind) {
+        let cache = match kind {
+            FacetKind::String => &mut self.strings,
+            FacetKind::Number => &mut self.f64s,
+            _ => return,
+        };
+
+        let key = (fid, value);
+        if let Some(DelAdd::Deletion) = cache.get(&key) {
+            cache.remove(&key);
+        } else {
+            cache.insert(key, DelAdd::Addition);
+        }
+    }
+
+    fn insert_del(&mut self, fid: FieldId, value: BVec<'doc, u8>, kind: FacetKind) {
+        let cache = match kind {
+            FacetKind::String => &mut self.strings,
+            FacetKind::Number => &mut self.f64s,
+            _ => return,
+        };
+
+        let key = (fid, value);
+        if let Some(DelAdd::Addition) = cache.get(&key) {
+            cache.remove(&key);
+        } else {
+            cache.insert(key, DelAdd::Deletion);
+        }
+    }
+
+    fn send_data(
+        self,
+        docid: DocumentId,
+        sender: &FieldIdDocidFacetSender,
+        doc_alloc: &Bump,
+    ) -> std::result::Result<(), crossbeam_channel::SendError<()>> {
+        let mut buffer = bumpalo::collections::Vec::new_in(doc_alloc);
+        for ((fid, value), deladd) in self.strings {
+            if let Ok(s) = std::str::from_utf8(&value) {
+                buffer.clear();
+                buffer.extend_from_slice(&fid.to_be_bytes());
+                buffer.extend_from_slice(&docid.to_be_bytes());
+                let normalized = crate::normalize_facet(s);
+                let truncated = truncate_str(&normalized);
+                buffer.extend_from_slice(truncated.as_bytes());
+                match deladd {
+                    DelAdd::Deletion => sender.delete_facet_string(&buffer)?,
+                    DelAdd::Addition => sender.write_facet_string(&buffer, &value)?,
+                }
+            }
+        }
+
+        for ((fid, value), deladd) in self.f64s {
+            buffer.clear();
+            buffer.extend_from_slice(&fid.to_be_bytes());
+            buffer.extend_from_slice(&docid.to_be_bytes());
+            buffer.extend_from_slice(&value);
+            match deladd {
+                DelAdd::Deletion => sender.delete_facet_f64(&buffer)?,
+                DelAdd::Addition => sender.write_facet_f64(&buffer)?,
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// Truncates a string to the biggest valid LMDB key size.
+fn truncate_str(s: &str) -> &str {
+    let index = s
+        .char_indices()
+        .map(|(idx, _)| idx)
+        .chain(std::iter::once(s.len()))
+        .take_while(|idx| idx <= &MAX_FACET_VALUE_LENGTH)
+        .last();
+
+    &s[..index.unwrap_or(0)]
+}
+
+impl FacetedDocidsExtractor {
+    #[tracing::instrument(level = "trace", skip_all, target = "indexing::extract::faceted")]
+    pub fn run_extraction<
+        'pl,
+        'fid,
+        'indexer,
+        'index,
+        'extractor,
+        DC: DocumentChanges<'pl>,
+        MSP,
+        SP,
+    >(
+        grenad_parameters: GrenadParameters,
+        document_changes: &DC,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP, SP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        sender: &FieldIdDocidFacetSender,
+        step: Step,
+    ) -> Result<Vec<BalancedCaches<'extractor>>>
+    where
+        MSP: Fn() -> bool + Sync,
+        SP: Fn(Progress) + Sync,
+    {
+        let index = indexing_context.index;
+        let rtxn = index.read_txn()?;
+        let attributes_to_extract = Self::attributes_to_extract(&rtxn, index)?;
+        let attributes_to_extract: Vec<_> =
+            attributes_to_extract.iter().map(|s| s.as_ref()).collect();
+        let datastore = ThreadLocal::new();
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::extract", "docids_extraction");
+            let _entered = span.enter();
+
+            let extractor = FacetedExtractorData {
+                attributes_to_extract: &attributes_to_extract,
+                grenad_parameters,
+                buckets: rayon::current_num_threads(),
+                sender,
+            };
+            extract(
+                document_changes,
+                &extractor,
+                indexing_context,
+                extractor_allocs,
+                &datastore,
+                step,
+            )?;
+        }
+
+        Ok(datastore.into_iter().map(RefCell::into_inner).collect())
+    }
+}

crates/milli/src/update/new/extract/faceted/facet_document.rs

@@ -0,0 +1,87 @@
+use serde_json::Value;
+
+use crate::update::new::document::Document;
+use crate::update::new::extract::geo::extract_geo_coordinates;
+use crate::update::new::extract::perm_json_p;
+use crate::{FieldId, GlobalFieldsIdsMap, InternalError, Result, UserError};
+
+pub fn extract_document_facets<'doc>(
+    attributes_to_extract: &[&str],
+    document: impl Document<'doc>,
+    external_document_id: &str,
+    field_id_map: &mut GlobalFieldsIdsMap,
+    facet_fn: &mut impl FnMut(FieldId, perm_json_p::Depth, &Value) -> Result<()>,
+) -> Result<()> {
+    for res in document.iter_top_level_fields() {
+        let (field_name, value) = res?;
+
+        let mut tokenize_field =
+            |name: &str, depth: perm_json_p::Depth, value: &Value| match field_id_map
+                .id_or_insert(name)
+            {
+                Some(field_id) => facet_fn(field_id, depth, value),
+                None => Err(UserError::AttributeLimitReached.into()),
+            };
+
+        // if the current field is searchable or contains a searchable attribute
+        let selection = perm_json_p::select_field(field_name, Some(attributes_to_extract), &[]);
+        if selection != perm_json_p::Selection::Skip {
+            // parse json.
+            match serde_json::value::to_value(value).map_err(InternalError::SerdeJson)? {
+                Value::Object(object) => {
+                    perm_json_p::seek_leaf_values_in_object(
+                        &object,
+                        Some(attributes_to_extract),
+                        &[], // skip no attributes
+                        field_name,
+                        perm_json_p::Depth::OnBaseKey,
+                        &mut tokenize_field,
+                    )?;
+
+                    if selection == perm_json_p::Selection::Select {
+                        tokenize_field(
+                            field_name,
+                            perm_json_p::Depth::OnBaseKey,
+                            &Value::Object(object),
+                        )?;
+                    }
+                }
+                Value::Array(array) => {
+                    perm_json_p::seek_leaf_values_in_array(
+                        &array,
+                        Some(attributes_to_extract),
+                        &[], // skip no attributes
+                        field_name,
+                        perm_json_p::Depth::OnBaseKey,
+                        &mut tokenize_field,
+                    )?;
+
+                    if selection == perm_json_p::Selection::Select {
+                        tokenize_field(
+                            field_name,
+                            perm_json_p::Depth::OnBaseKey,
+                            &Value::Array(array),
+                        )?;
+                    }
+                }
+                value => tokenize_field(field_name, perm_json_p::Depth::OnBaseKey, &value)?,
+            }
+        }
+    }
+
+    if attributes_to_extract.contains(&"_geo") {
+        if let Some(geo_value) = document.geo_field()? {
+            if let Some([lat, lng]) = extract_geo_coordinates(external_document_id, geo_value)? {
+                let (lat_fid, lng_fid) = field_id_map
+                    .id_or_insert("_geo.lat")
+                    .zip(field_id_map.id_or_insert("_geo.lng"))
+                    .ok_or(UserError::AttributeLimitReached)?;
+
+                facet_fn(lat_fid, perm_json_p::Depth::OnBaseKey, &lat.into())?;
+                facet_fn(lng_fid, perm_json_p::Depth::OnBaseKey, &lng.into())?;
+            }
+        }
+    }
+
+    Ok(())
+}

crates/milli/src/update/new/extract/faceted/mod.rs

@@ -0,0 +1,33 @@
+mod extract_facets;
+mod facet_document;
+
+pub use extract_facets::FacetedDocidsExtractor;
+
+#[repr(u8)]
+#[derive(Debug, Clone, Copy)]
+pub enum FacetKind {
+    Number = 0,
+    String = 1,
+    Null = 2,
+    Empty = 3,
+    Exists,
+}
+
+impl From<u8> for FacetKind {
+    fn from(value: u8) -> Self {
+        match value {
+            0 => Self::Number,
+            1 => Self::String,
+            2 => Self::Null,
+            3 => Self::Empty,
+            4 => Self::Exists,
+            _ => unreachable!(),
+        }
+    }
+}
+
+impl FacetKind {
+    pub fn extract_from_key(key: &[u8]) -> (FacetKind, &[u8]) {
+        (FacetKind::from(key[0]), &key[1..])
+    }
+}

crates/milli/src/update/new/extract/geo/mod.rs

@@ -0,0 +1,324 @@
+use std::cell::RefCell;
+use std::fs::File;
+use std::io::{self, BufReader, BufWriter, ErrorKind, Read, Write as _};
+use std::{iter, mem, result};
+
+use bumpalo::Bump;
+use bytemuck::{bytes_of, pod_read_unaligned, Pod, Zeroable};
+use heed::RoTxn;
+use serde_json::value::RawValue;
+use serde_json::Value;
+
+use crate::error::GeoError;
+use crate::update::new::document::Document;
+use crate::update::new::indexer::document_changes::{DocumentChangeContext, Extractor};
+use crate::update::new::ref_cell_ext::RefCellExt as _;
+use crate::update::new::thread_local::MostlySend;
+use crate::update::new::DocumentChange;
+use crate::update::GrenadParameters;
+use crate::{lat_lng_to_xyz, DocumentId, GeoPoint, Index, InternalError, Result};
+
+pub struct GeoExtractor {
+    grenad_parameters: GrenadParameters,
+}
+
+impl GeoExtractor {
+    pub fn new(
+        rtxn: &RoTxn,
+        index: &Index,
+        grenad_parameters: GrenadParameters,
+    ) -> Result<Option<Self>> {
+        let is_sortable = index.sortable_fields(rtxn)?.contains("_geo");
+        let is_filterable = index.filterable_fields(rtxn)?.contains("_geo");
+        if is_sortable || is_filterable {
+            Ok(Some(GeoExtractor { grenad_parameters }))
+        } else {
+            Ok(None)
+        }
+    }
+}
+
+#[derive(Pod, Zeroable, Copy, Clone)]
+#[repr(C, packed)]
+pub struct ExtractedGeoPoint {
+    pub docid: DocumentId,
+    pub lat_lng: [f64; 2],
+}
+
+impl From<ExtractedGeoPoint> for GeoPoint {
+    /// Converts the latitude and longitude back to an xyz GeoPoint.
+    fn from(value: ExtractedGeoPoint) -> Self {
+        let [lat, lng] = value.lat_lng;
+        let point = [lat, lng];
+        let xyz_point = lat_lng_to_xyz(&point);
+        GeoPoint::new(xyz_point, (value.docid, point))
+    }
+}
+
+pub struct GeoExtractorData<'extractor> {
+    /// The set of documents ids that were removed. If a document sees its geo
+    /// point being updated, we first put it in the deleted and then in the inserted.
+    removed: bumpalo::collections::Vec<'extractor, ExtractedGeoPoint>,
+    inserted: bumpalo::collections::Vec<'extractor, ExtractedGeoPoint>,
+    /// TODO Do the doc
+    spilled_removed: Option<BufWriter<File>>,
+    /// TODO Do the doc
+    spilled_inserted: Option<BufWriter<File>>,
+}
+
+impl<'extractor> GeoExtractorData<'extractor> {
+    pub fn freeze(self) -> Result<FrozenGeoExtractorData<'extractor>> {
+        let GeoExtractorData { removed, inserted, spilled_removed, spilled_inserted } = self;
+
+        Ok(FrozenGeoExtractorData {
+            removed: removed.into_bump_slice(),
+            inserted: inserted.into_bump_slice(),
+            spilled_removed: spilled_removed
+                .map(|bw| bw.into_inner().map(BufReader::new).map_err(|iie| iie.into_error()))
+                .transpose()?,
+            spilled_inserted: spilled_inserted
+                .map(|bw| bw.into_inner().map(BufReader::new).map_err(|iie| iie.into_error()))
+                .transpose()?,
+        })
+    }
+}
+
+unsafe impl MostlySend for GeoExtractorData<'_> {}
+
+pub struct FrozenGeoExtractorData<'extractor> {
+    pub removed: &'extractor [ExtractedGeoPoint],
+    pub inserted: &'extractor [ExtractedGeoPoint],
+    pub spilled_removed: Option<BufReader<File>>,
+    pub spilled_inserted: Option<BufReader<File>>,
+}
+
+impl<'extractor> FrozenGeoExtractorData<'extractor> {
+    pub fn iter_and_clear_removed(
+        &mut self,
+    ) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
+        mem::take(&mut self.removed)
+            .iter()
+            .copied()
+            .map(Ok)
+            .chain(iterator_over_spilled_geopoints(&mut self.spilled_removed))
+    }
+
+    pub fn iter_and_clear_inserted(
+        &mut self,
+    ) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
+        mem::take(&mut self.inserted)
+            .iter()
+            .copied()
+            .map(Ok)
+            .chain(iterator_over_spilled_geopoints(&mut self.spilled_inserted))
+    }
+}
+
+fn iterator_over_spilled_geopoints(
+    spilled: &mut Option<BufReader<File>>,
+) -> impl IntoIterator<Item = io::Result<ExtractedGeoPoint>> + '_ {
+    let mut spilled = spilled.take();
+    iter::from_fn(move || match &mut spilled {
+        Some(file) => {
+            let geopoint_bytes = &mut [0u8; mem::size_of::<ExtractedGeoPoint>()];
+            match file.read_exact(geopoint_bytes) {
+                Ok(()) => Some(Ok(pod_read_unaligned(geopoint_bytes))),
+                Err(e) if e.kind() == ErrorKind::UnexpectedEof => None,
+                Err(e) => Some(Err(e)),
+            }
+        }
+        None => None,
+    })
+}
+
+impl<'extractor> Extractor<'extractor> for GeoExtractor {
+    type Data = RefCell<GeoExtractorData<'extractor>>;
+
+    fn init_data<'doc>(&'doc self, extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
+        Ok(RefCell::new(GeoExtractorData {
+            removed: bumpalo::collections::Vec::new_in(extractor_alloc),
+            // inserted: Uell::new_in(extractor_alloc),
+            inserted: bumpalo::collections::Vec::new_in(extractor_alloc),
+            spilled_inserted: None,
+            spilled_removed: None,
+        }))
+    }
+
+    fn process<'doc>(
+        &'doc self,
+        changes: impl Iterator<Item = Result<DocumentChange<'doc>>>,
+        context: &'doc DocumentChangeContext<Self::Data>,
+    ) -> Result<()> {
+        let rtxn = &context.rtxn;
+        let index = context.index;
+        let max_memory = self.grenad_parameters.max_memory_by_thread();
+        let db_fields_ids_map = context.db_fields_ids_map;
+        let mut data_ref = context.data.borrow_mut_or_yield();
+
+        for change in changes {
+            if max_memory.map_or(false, |mm| context.extractor_alloc.allocated_bytes() >= mm) {
+                // We must spill as we allocated too much memory
+                data_ref.spilled_removed = tempfile::tempfile().map(BufWriter::new).map(Some)?;
+                data_ref.spilled_inserted = tempfile::tempfile().map(BufWriter::new).map(Some)?;
+            }
+
+            match change? {
+                DocumentChange::Deletion(deletion) => {
+                    let docid = deletion.docid();
+                    let external_id = deletion.external_document_id();
+                    let current = deletion.current(rtxn, index, db_fields_ids_map)?;
+                    let current_geo = current
+                        .geo_field()?
+                        .map(|geo| extract_geo_coordinates(external_id, geo))
+                        .transpose()?;
+
+                    if let Some(lat_lng) = current_geo.flatten() {
+                        let geopoint = ExtractedGeoPoint { docid, lat_lng };
+                        match &mut data_ref.spilled_removed {
+                            Some(file) => file.write_all(bytes_of(&geopoint))?,
+                            None => data_ref.removed.push(geopoint),
+                        }
+                    }
+                }
+                DocumentChange::Update(update) => {
+                    let current = update.current(rtxn, index, db_fields_ids_map)?;
+                    let external_id = update.external_document_id();
+                    let docid = update.docid();
+
+                    let current_geo = current
+                        .geo_field()?
+                        .map(|geo| extract_geo_coordinates(external_id, geo))
+                        .transpose()?;
+
+                    let updated_geo = update
+                        .updated()
+                        .geo_field()?
+                        .map(|geo| extract_geo_coordinates(external_id, geo))
+                        .transpose()?;
+
+                    if current_geo != updated_geo {
+                        // If the current and new geo points are different it means that
+                        // we need to replace the current by the new point and therefore
+                        // delete the current point from the RTree.
+                        if let Some(lat_lng) = current_geo.flatten() {
+                            let geopoint = ExtractedGeoPoint { docid, lat_lng };
+                            match &mut data_ref.spilled_removed {
+                                Some(file) => file.write_all(bytes_of(&geopoint))?,
+                                None => data_ref.removed.push(geopoint),
+                            }
+                        }
+
+                        if let Some(lat_lng) = updated_geo.flatten() {
+                            let geopoint = ExtractedGeoPoint { docid, lat_lng };
+                            match &mut data_ref.spilled_inserted {
+                                Some(file) => file.write_all(bytes_of(&geopoint))?,
+                                None => data_ref.inserted.push(geopoint),
+                            }
+                        }
+                    }
+                }
+                DocumentChange::Insertion(insertion) => {
+                    let external_id = insertion.external_document_id();
+                    let docid = insertion.docid();
+
+                    let inserted_geo = insertion
+                        .inserted()
+                        .geo_field()?
+                        .map(|geo| extract_geo_coordinates(external_id, geo))
+                        .transpose()?;
+
+                    if let Some(lat_lng) = inserted_geo.flatten() {
+                        let geopoint = ExtractedGeoPoint { docid, lat_lng };
+                        match &mut data_ref.spilled_inserted {
+                            Some(file) => file.write_all(bytes_of(&geopoint))?,
+                            None => data_ref.inserted.push(geopoint),
+                        }
+                    }
+                }
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// Extracts and validate the latitude and latitude from a document geo field.
+///
+/// It can be of the form `{ "lat": 0.0, "lng": "1.0" }`.
+pub fn extract_geo_coordinates(
+    external_id: &str,
+    raw_value: &RawValue,
+) -> Result<Option<[f64; 2]>> {
+    let mut geo = match serde_json::from_str(raw_value.get()).map_err(InternalError::SerdeJson)? {
+        Value::Null => return Ok(None),
+        Value::Object(map) => map,
+        value => {
+            return Err(
+                GeoError::NotAnObject { document_id: Value::from(external_id), value }.into()
+            )
+        }
+    };
+
+    let [lat, lng] = match (geo.remove("lat"), geo.remove("lng")) {
+        (Some(lat), Some(lng)) => {
+            if geo.is_empty() {
+                [lat, lng]
+            } else {
+                return Err(GeoError::UnexpectedExtraFields {
+                    document_id: Value::from(external_id),
+                    value: Value::from(geo),
+                }
+                .into());
+            }
+        }
+        (Some(_), None) => {
+            return Err(GeoError::MissingLongitude { document_id: Value::from(external_id) }.into())
+        }
+        (None, Some(_)) => {
+            return Err(GeoError::MissingLatitude { document_id: Value::from(external_id) }.into())
+        }
+        (None, None) => {
+            return Err(GeoError::MissingLatitudeAndLongitude {
+                document_id: Value::from(external_id),
+            }
+            .into())
+        }
+    };
+
+    match (extract_finite_float_from_value(lat), extract_finite_float_from_value(lng)) {
+        (Ok(lat), Ok(lng)) => Ok(Some([lat, lng])),
+        (Ok(_), Err(value)) => {
+            Err(GeoError::BadLongitude { document_id: Value::from(external_id), value }.into())
+        }
+        (Err(value), Ok(_)) => {
+            Err(GeoError::BadLatitude { document_id: Value::from(external_id), value }.into())
+        }
+        (Err(lat), Err(lng)) => Err(GeoError::BadLatitudeAndLongitude {
+            document_id: Value::from(external_id),
+            lat,
+            lng,
+        }
+        .into()),
+    }
+}
+
+/// Extracts and validate that a serde JSON Value is actually a finite f64.
+pub fn extract_finite_float_from_value(value: Value) -> result::Result<f64, Value> {
+    let number = match value {
+        Value::Number(ref n) => match n.as_f64() {
+            Some(number) => number,
+            None => return Err(value),
+        },
+        Value::String(ref s) => match s.parse::<f64>() {
+            Ok(number) => number,
+            Err(_) => return Err(value),
+        },
+        value => return Err(value),
+    };
+
+    if number.is_finite() {
+        Ok(number)
+    } else {
+        Err(value)
+    }
+}

crates/milli/src/update/new/extract/mod.rs

@@ -0,0 +1,201 @@
+mod cache;
+mod documents;
+mod faceted;
+mod geo;
+mod searchable;
+mod vectors;
+
+use bumpalo::Bump;
+pub use cache::{merge_caches, transpose_and_freeze_caches, BalancedCaches, DelAddRoaringBitmap};
+pub use documents::*;
+pub use faceted::*;
+pub use geo::*;
+pub use searchable::*;
+pub use vectors::EmbeddingExtractor;
+
+use super::indexer::document_changes::{DocumentChanges, IndexingContext, Progress};
+use super::steps::Step;
+use super::thread_local::{FullySend, ThreadLocal};
+use crate::update::GrenadParameters;
+use crate::Result;
+
+pub trait DocidsExtractor {
+    fn run_extraction<'pl, 'fid, 'indexer, 'index, 'extractor, DC: DocumentChanges<'pl>, MSP, SP>(
+        grenad_parameters: GrenadParameters,
+        document_changes: &DC,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP, SP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: Step,
+    ) -> Result<Vec<BalancedCaches<'extractor>>>
+    where
+        MSP: Fn() -> bool + Sync,
+        SP: Fn(Progress) + Sync;
+}
+
+/// TODO move in permissive json pointer
+pub mod perm_json_p {
+    use serde_json::{Map, Value};
+
+    use crate::Result;
+    const SPLIT_SYMBOL: char = '.';
+
+    /// Returns `true` if the `selector` match the `key`.
+    ///
+    /// ```text
+    /// Example:
+    /// `animaux`           match `animaux`
+    /// `animaux.chien`     match `animaux`
+    /// `animaux.chien`     match `animaux`
+    /// `animaux.chien.nom` match `animaux`
+    /// `animaux.chien.nom` match `animaux.chien`
+    /// -----------------------------------------
+    /// `animaux`    doesn't match `animaux.chien`
+    /// `animaux.`   doesn't match `animaux`
+    /// `animaux.ch` doesn't match `animaux.chien`
+    /// `animau`     doesn't match `animaux`
+    /// ```
+    pub fn contained_in(selector: &str, key: &str) -> bool {
+        selector.starts_with(key)
+            && selector[key.len()..].chars().next().map(|c| c == SPLIT_SYMBOL).unwrap_or(true)
+    }
+
+    #[derive(Debug, Clone, Copy, PartialEq, Eq)]
+    pub enum Depth {
+        /// The perm json ptr is currently on the field of an object
+        OnBaseKey,
+        /// The perm json ptr is currently inside of an array
+        InsideArray,
+    }
+
+    pub fn seek_leaf_values_in_object(
+        value: &Map<String, Value>,
+        selectors: Option<&[&str]>,
+        skip_selectors: &[&str],
+        base_key: &str,
+        base_depth: Depth,
+        seeker: &mut impl FnMut(&str, Depth, &Value) -> Result<()>,
+    ) -> Result<()> {
+        if value.is_empty() {
+            seeker(base_key, base_depth, &Value::Object(Map::with_capacity(0)))?;
+        }
+
+        for (key, value) in value.iter() {
+            let base_key = if base_key.is_empty() {
+                key.to_string()
+            } else {
+                format!("{}{}{}", base_key, SPLIT_SYMBOL, key)
+            };
+
+            // here if the user only specified `doggo` we need to iterate in all the fields of `doggo`
+            // so we check the contained_in on both side
+            let selection = select_field(&base_key, selectors, skip_selectors);
+            if selection != Selection::Skip {
+                match value {
+                    Value::Object(object) => {
+                        if selection == Selection::Select {
+                            seeker(&base_key, Depth::OnBaseKey, value)?;
+                        }
+
+                        seek_leaf_values_in_object(
+                            object,
+                            selectors,
+                            skip_selectors,
+                            &base_key,
+                            Depth::OnBaseKey,
+                            seeker,
+                        )
+                    }
+                    Value::Array(array) => {
+                        if selection == Selection::Select {
+                            seeker(&base_key, Depth::OnBaseKey, value)?;
+                        }
+
+                        seek_leaf_values_in_array(
+                            array,
+                            selectors,
+                            skip_selectors,
+                            &base_key,
+                            Depth::OnBaseKey,
+                            seeker,
+                        )
+                    }
+                    value => seeker(&base_key, Depth::OnBaseKey, value),
+                }?;
+            }
+        }
+
+        Ok(())
+    }
+
+    pub fn seek_leaf_values_in_array(
+        values: &[Value],
+        selectors: Option<&[&str]>,
+        skip_selectors: &[&str],
+        base_key: &str,
+        base_depth: Depth,
+        seeker: &mut impl FnMut(&str, Depth, &Value) -> Result<()>,
+    ) -> Result<()> {
+        if values.is_empty() {
+            seeker(base_key, base_depth, &Value::Array(vec![]))?;
+        }
+
+        for value in values {
+            match value {
+                Value::Object(object) => seek_leaf_values_in_object(
+                    object,
+                    selectors,
+                    skip_selectors,
+                    base_key,
+                    Depth::InsideArray,
+                    seeker,
+                ),
+                Value::Array(array) => seek_leaf_values_in_array(
+                    array,
+                    selectors,
+                    skip_selectors,
+                    base_key,
+                    Depth::InsideArray,
+                    seeker,
+                ),
+                value => seeker(base_key, Depth::InsideArray, value),
+            }?;
+        }
+
+        Ok(())
+    }
+
+    pub fn select_field(
+        field_name: &str,
+        selectors: Option<&[&str]>,
+        skip_selectors: &[&str],
+    ) -> Selection {
+        if skip_selectors.iter().any(|skip_selector| {
+            contained_in(skip_selector, field_name) || contained_in(field_name, skip_selector)
+        }) {
+            Selection::Skip
+        } else if let Some(selectors) = selectors {
+            let mut selection = Selection::Skip;
+            for selector in selectors {
+                if contained_in(field_name, selector) {
+                    selection = Selection::Select;
+                    break;
+                } else if contained_in(selector, field_name) {
+                    selection = Selection::Parent;
+                }
+            }
+            selection
+        } else {
+            Selection::Select
+        }
+    }
+
+    #[derive(Debug, Clone, Copy, PartialEq, Eq)]
+    pub enum Selection {
+        /// The field is a parent of the of a nested field that must be selected
+        Parent,
+        /// The field must be selected
+        Select,
+        /// The field must be skipped
+        Skip,
+    }
+}

crates/milli/src/update/new/extract/searchable/extract_word_docids.rs

@@ -0,0 +1,421 @@
+use std::cell::RefCell;
+use std::collections::HashMap;
+use std::mem::size_of;
+use std::ops::DerefMut as _;
+
+use bumpalo::collections::vec::Vec as BumpVec;
+use bumpalo::Bump;
+use heed::RoTxn;
+
+use super::tokenize_document::{tokenizer_builder, DocumentTokenizer};
+use crate::update::new::extract::cache::BalancedCaches;
+use crate::update::new::extract::perm_json_p::contained_in;
+use crate::update::new::indexer::document_changes::{
+    extract, DocumentChangeContext, DocumentChanges, Extractor, IndexingContext, Progress,
+};
+use crate::update::new::ref_cell_ext::RefCellExt as _;
+use crate::update::new::steps::Step;
+use crate::update::new::thread_local::{FullySend, MostlySend, ThreadLocal};
+use crate::update::new::DocumentChange;
+use crate::update::GrenadParameters;
+use crate::{bucketed_position, DocumentId, FieldId, Index, Result, MAX_POSITION_PER_ATTRIBUTE};
+
+const MAX_COUNTED_WORDS: usize = 30;
+
+pub struct WordDocidsBalancedCaches<'extractor> {
+    word_fid_docids: BalancedCaches<'extractor>,
+    word_docids: BalancedCaches<'extractor>,
+    exact_word_docids: BalancedCaches<'extractor>,
+    word_position_docids: BalancedCaches<'extractor>,
+    fid_word_count_docids: BalancedCaches<'extractor>,
+    fid_word_count: HashMap<FieldId, (usize, usize)>,
+    current_docid: Option<DocumentId>,
+}
+
+unsafe impl<'extractor> MostlySend for WordDocidsBalancedCaches<'extractor> {}
+
+impl<'extractor> WordDocidsBalancedCaches<'extractor> {
+    /// TODO Make sure to give the same max_memory to all of them, without splitting it
+    pub fn new_in(buckets: usize, max_memory: Option<usize>, alloc: &'extractor Bump) -> Self {
+        Self {
+            word_fid_docids: BalancedCaches::new_in(buckets, max_memory, alloc),
+            word_docids: BalancedCaches::new_in(buckets, max_memory, alloc),
+            exact_word_docids: BalancedCaches::new_in(buckets, max_memory, alloc),
+            word_position_docids: BalancedCaches::new_in(buckets, max_memory, alloc),
+            fid_word_count_docids: BalancedCaches::new_in(buckets, max_memory, alloc),
+            fid_word_count: HashMap::new(),
+            current_docid: None,
+        }
+    }
+
+    fn insert_add_u32(
+        &mut self,
+        field_id: FieldId,
+        position: u16,
+        word: &str,
+        exact: bool,
+        docid: u32,
+        bump: &Bump,
+    ) -> Result<()> {
+        let word_bytes = word.as_bytes();
+        if exact {
+            self.exact_word_docids.insert_add_u32(word_bytes, docid)?;
+        } else {
+            self.word_docids.insert_add_u32(word_bytes, docid)?;
+        }
+
+        let buffer_size = word_bytes.len() + 1 + size_of::<FieldId>();
+        let mut buffer = BumpVec::with_capacity_in(buffer_size, bump);
+
+        buffer.clear();
+        buffer.extend_from_slice(word_bytes);
+        buffer.push(0);
+        buffer.extend_from_slice(&field_id.to_be_bytes());
+        self.word_fid_docids.insert_add_u32(&buffer, docid)?;
+
+        let position = bucketed_position(position);
+        buffer.clear();
+        buffer.extend_from_slice(word_bytes);
+        buffer.push(0);
+        buffer.extend_from_slice(&position.to_be_bytes());
+        self.word_position_docids.insert_add_u32(&buffer, docid)?;
+
+        if self.current_docid.map_or(false, |id| docid != id) {
+            self.flush_fid_word_count(&mut buffer)?;
+        }
+
+        self.fid_word_count
+            .entry(field_id)
+            .and_modify(|(_current_count, new_count)| *new_count += 1)
+            .or_insert((0, 1));
+        self.current_docid = Some(docid);
+
+        Ok(())
+    }
+
+    fn insert_del_u32(
+        &mut self,
+        field_id: FieldId,
+        position: u16,
+        word: &str,
+        exact: bool,
+        docid: u32,
+        bump: &Bump,
+    ) -> Result<()> {
+        let word_bytes = word.as_bytes();
+        if exact {
+            self.exact_word_docids.insert_del_u32(word_bytes, docid)?;
+        } else {
+            self.word_docids.insert_del_u32(word_bytes, docid)?;
+        }
+
+        let buffer_size = word_bytes.len() + 1 + size_of::<FieldId>();
+        let mut buffer = BumpVec::with_capacity_in(buffer_size, bump);
+
+        buffer.clear();
+        buffer.extend_from_slice(word_bytes);
+        buffer.push(0);
+        buffer.extend_from_slice(&field_id.to_be_bytes());
+        self.word_fid_docids.insert_del_u32(&buffer, docid)?;
+
+        let position = bucketed_position(position);
+        buffer.clear();
+        buffer.extend_from_slice(word_bytes);
+        buffer.push(0);
+        buffer.extend_from_slice(&position.to_be_bytes());
+        self.word_position_docids.insert_del_u32(&buffer, docid)?;
+
+        if self.current_docid.map_or(false, |id| docid != id) {
+            self.flush_fid_word_count(&mut buffer)?;
+        }
+
+        self.fid_word_count
+            .entry(field_id)
+            .and_modify(|(current_count, _new_count)| *current_count += 1)
+            .or_insert((1, 0));
+
+        self.current_docid = Some(docid);
+
+        Ok(())
+    }
+
+    fn flush_fid_word_count(&mut self, buffer: &mut BumpVec<u8>) -> Result<()> {
+        for (fid, (current_count, new_count)) in self.fid_word_count.drain() {
+            if current_count != new_count {
+                if current_count <= MAX_COUNTED_WORDS {
+                    buffer.clear();
+                    buffer.extend_from_slice(&fid.to_be_bytes());
+                    buffer.push(current_count as u8);
+                    self.fid_word_count_docids
+                        .insert_del_u32(buffer, self.current_docid.unwrap())?;
+                }
+                if new_count <= MAX_COUNTED_WORDS {
+                    buffer.clear();
+                    buffer.extend_from_slice(&fid.to_be_bytes());
+                    buffer.push(new_count as u8);
+                    self.fid_word_count_docids
+                        .insert_add_u32(buffer, self.current_docid.unwrap())?;
+                }
+            }
+        }
+
+        Ok(())
+    }
+}
+
+pub struct WordDocidsCaches<'extractor> {
+    pub word_docids: Vec<BalancedCaches<'extractor>>,
+    pub word_fid_docids: Vec<BalancedCaches<'extractor>>,
+    pub exact_word_docids: Vec<BalancedCaches<'extractor>>,
+    pub word_position_docids: Vec<BalancedCaches<'extractor>>,
+    pub fid_word_count_docids: Vec<BalancedCaches<'extractor>>,
+}
+
+impl<'extractor> WordDocidsCaches<'extractor> {
+    fn new() -> Self {
+        Self {
+            word_docids: Vec::new(),
+            word_fid_docids: Vec::new(),
+            exact_word_docids: Vec::new(),
+            word_position_docids: Vec::new(),
+            fid_word_count_docids: Vec::new(),
+        }
+    }
+
+    fn push(&mut self, other: WordDocidsBalancedCaches<'extractor>) -> Result<()> {
+        let WordDocidsBalancedCaches {
+            word_docids,
+            word_fid_docids,
+            exact_word_docids,
+            word_position_docids,
+            fid_word_count_docids,
+            fid_word_count: _,
+            current_docid: _,
+        } = other;
+
+        self.word_docids.push(word_docids);
+        self.word_fid_docids.push(word_fid_docids);
+        self.exact_word_docids.push(exact_word_docids);
+        self.word_position_docids.push(word_position_docids);
+        self.fid_word_count_docids.push(fid_word_count_docids);
+
+        Ok(())
+    }
+}
+
+pub struct WordDocidsExtractorData<'a> {
+    tokenizer: &'a DocumentTokenizer<'a>,
+    grenad_parameters: GrenadParameters,
+    buckets: usize,
+}
+
+impl<'a, 'extractor> Extractor<'extractor> for WordDocidsExtractorData<'a> {
+    type Data = RefCell<Option<WordDocidsBalancedCaches<'extractor>>>;
+
+    fn init_data(&self, extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
+        Ok(RefCell::new(Some(WordDocidsBalancedCaches::new_in(
+            self.buckets,
+            self.grenad_parameters.max_memory_by_thread(),
+            extractor_alloc,
+        ))))
+    }
+
+    fn process<'doc>(
+        &self,
+        changes: impl Iterator<Item = Result<DocumentChange<'doc>>>,
+        context: &DocumentChangeContext<Self::Data>,
+    ) -> Result<()> {
+        for change in changes {
+            let change = change?;
+            WordDocidsExtractors::extract_document_change(context, self.tokenizer, change)?;
+        }
+        Ok(())
+    }
+}
+
+pub struct WordDocidsExtractors;
+
+impl WordDocidsExtractors {
+    pub fn run_extraction<
+        'pl,
+        'fid,
+        'indexer,
+        'index,
+        'extractor,
+        DC: DocumentChanges<'pl>,
+        MSP,
+        SP,
+    >(
+        grenad_parameters: GrenadParameters,
+        document_changes: &DC,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP, SP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: Step,
+    ) -> Result<WordDocidsCaches<'extractor>>
+    where
+        MSP: Fn() -> bool + Sync,
+        SP: Fn(Progress) + Sync,
+    {
+        let index = indexing_context.index;
+        let rtxn = index.read_txn()?;
+
+        let stop_words = index.stop_words(&rtxn)?;
+        let allowed_separators = index.allowed_separators(&rtxn)?;
+        let allowed_separators: Option<Vec<_>> =
+            allowed_separators.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let dictionary = index.dictionary(&rtxn)?;
+        let dictionary: Option<Vec<_>> =
+            dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let builder = tokenizer_builder(
+            stop_words.as_ref(),
+            allowed_separators.as_deref(),
+            dictionary.as_deref(),
+        );
+        let tokenizer = builder.into_tokenizer();
+
+        let attributes_to_extract = Self::attributes_to_extract(&rtxn, index)?;
+        let attributes_to_skip = Self::attributes_to_skip(&rtxn, index)?;
+        let localized_attributes_rules =
+            index.localized_attributes_rules(&rtxn)?.unwrap_or_default();
+
+        let document_tokenizer = DocumentTokenizer {
+            tokenizer: &tokenizer,
+            attribute_to_extract: attributes_to_extract.as_deref(),
+            attribute_to_skip: attributes_to_skip.as_slice(),
+            localized_attributes_rules: &localized_attributes_rules,
+            max_positions_per_attributes: MAX_POSITION_PER_ATTRIBUTE,
+        };
+
+        let datastore = ThreadLocal::new();
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::extract", "docids_extraction");
+            let _entered = span.enter();
+
+            let extractor = WordDocidsExtractorData {
+                tokenizer: &document_tokenizer,
+                grenad_parameters,
+                buckets: rayon::current_num_threads(),
+            };
+
+            extract(
+                document_changes,
+                &extractor,
+                indexing_context,
+                extractor_allocs,
+                &datastore,
+                step,
+            )?;
+        }
+
+        let mut merger = WordDocidsCaches::new();
+        for cache in datastore.into_iter().flat_map(RefCell::into_inner) {
+            merger.push(cache)?;
+        }
+
+        Ok(merger)
+    }
+
+    fn extract_document_change(
+        context: &DocumentChangeContext<RefCell<Option<WordDocidsBalancedCaches>>>,
+        document_tokenizer: &DocumentTokenizer,
+        document_change: DocumentChange,
+    ) -> Result<()> {
+        let index = &context.index;
+        let rtxn = &context.rtxn;
+        let mut cached_sorter_ref = context.data.borrow_mut_or_yield();
+        let cached_sorter = cached_sorter_ref.as_mut().unwrap();
+        let mut new_fields_ids_map = context.new_fields_ids_map.borrow_mut_or_yield();
+        let new_fields_ids_map = new_fields_ids_map.deref_mut();
+        let doc_alloc = &context.doc_alloc;
+
+        let exact_attributes = index.exact_attributes(rtxn)?;
+        let is_exact_attribute =
+            |fname: &str| exact_attributes.iter().any(|attr| contained_in(fname, attr));
+        match document_change {
+            DocumentChange::Deletion(inner) => {
+                let mut token_fn = |fname: &str, fid, pos, word: &str| {
+                    cached_sorter.insert_del_u32(
+                        fid,
+                        pos,
+                        word,
+                        is_exact_attribute(fname),
+                        inner.docid(),
+                        doc_alloc,
+                    )
+                };
+                document_tokenizer.tokenize_document(
+                    inner.current(rtxn, index, context.db_fields_ids_map)?,
+                    new_fields_ids_map,
+                    &mut token_fn,
+                )?;
+            }
+            DocumentChange::Update(inner) => {
+                let mut token_fn = |fname: &str, fid, pos, word: &str| {
+                    cached_sorter.insert_del_u32(
+                        fid,
+                        pos,
+                        word,
+                        is_exact_attribute(fname),
+                        inner.docid(),
+                        doc_alloc,
+                    )
+                };
+                document_tokenizer.tokenize_document(
+                    inner.current(rtxn, index, context.db_fields_ids_map)?,
+                    new_fields_ids_map,
+                    &mut token_fn,
+                )?;
+
+                let mut token_fn = |fname: &str, fid, pos, word: &str| {
+                    cached_sorter.insert_add_u32(
+                        fid,
+                        pos,
+                        word,
+                        is_exact_attribute(fname),
+                        inner.docid(),
+                        doc_alloc,
+                    )
+                };
+                document_tokenizer.tokenize_document(
+                    inner.merged(rtxn, index, context.db_fields_ids_map)?,
+                    new_fields_ids_map,
+                    &mut token_fn,
+                )?;
+            }
+            DocumentChange::Insertion(inner) => {
+                let mut token_fn = |fname: &str, fid, pos, word: &str| {
+                    cached_sorter.insert_add_u32(
+                        fid,
+                        pos,
+                        word,
+                        is_exact_attribute(fname),
+                        inner.docid(),
+                        doc_alloc,
+                    )
+                };
+                document_tokenizer.tokenize_document(
+                    inner.inserted(),
+                    new_fields_ids_map,
+                    &mut token_fn,
+                )?;
+            }
+        }
+
+        let buffer_size = size_of::<FieldId>();
+        let mut buffer = BumpVec::with_capacity_in(buffer_size, &context.doc_alloc);
+        cached_sorter.flush_fid_word_count(&mut buffer)
+    }
+
+    fn attributes_to_extract<'a>(
+        rtxn: &'a RoTxn,
+        index: &'a Index,
+    ) -> Result<Option<Vec<&'a str>>> {
+        index.user_defined_searchable_fields(rtxn).map_err(Into::into)
+    }
+
+    fn attributes_to_skip<'a>(_rtxn: &'a RoTxn, _index: &'a Index) -> Result<Vec<&'a str>> {
+        Ok(vec!["_geo"])
+    }
+}

crates/milli/src/update/new/extract/searchable/extract_word_pair_proximity_docids.rs

@@ -0,0 +1,190 @@
+use std::cell::RefCell;
+use std::collections::VecDeque;
+use std::rc::Rc;
+
+use heed::RoTxn;
+
+use super::tokenize_document::DocumentTokenizer;
+use super::SearchableExtractor;
+use crate::proximity::{index_proximity, MAX_DISTANCE};
+use crate::update::new::document::Document;
+use crate::update::new::extract::cache::BalancedCaches;
+use crate::update::new::indexer::document_changes::DocumentChangeContext;
+use crate::update::new::ref_cell_ext::RefCellExt as _;
+use crate::update::new::DocumentChange;
+use crate::{FieldId, GlobalFieldsIdsMap, Index, Result};
+
+pub struct WordPairProximityDocidsExtractor;
+
+impl SearchableExtractor for WordPairProximityDocidsExtractor {
+    fn attributes_to_extract<'a>(
+        rtxn: &'a RoTxn,
+        index: &'a Index,
+    ) -> Result<Option<Vec<&'a str>>> {
+        index.user_defined_searchable_fields(rtxn).map_err(Into::into)
+    }
+
+    fn attributes_to_skip<'a>(_rtxn: &'a RoTxn, _index: &'a Index) -> Result<Vec<&'a str>> {
+        Ok(vec!["_geo"])
+    }
+
+    // This method is reimplemented to count the number of words in the document in each field
+    // and to store the docids of the documents that have a number of words in a given field
+    // equal to or under than MAX_COUNTED_WORDS.
+    fn extract_document_change(
+        context: &DocumentChangeContext<RefCell<BalancedCaches>>,
+        document_tokenizer: &DocumentTokenizer,
+        document_change: DocumentChange,
+    ) -> Result<()> {
+        let doc_alloc = &context.doc_alloc;
+
+        let index = context.index;
+        let rtxn = &context.rtxn;
+
+        let mut key_buffer = bumpalo::collections::Vec::new_in(doc_alloc);
+        let mut del_word_pair_proximity = bumpalo::collections::Vec::new_in(doc_alloc);
+        let mut add_word_pair_proximity = bumpalo::collections::Vec::new_in(doc_alloc);
+
+        let mut new_fields_ids_map = context.new_fields_ids_map.borrow_mut_or_yield();
+        let new_fields_ids_map = &mut *new_fields_ids_map;
+
+        let mut cached_sorter = context.data.borrow_mut_or_yield();
+        let cached_sorter = &mut *cached_sorter;
+
+        // is a vecdequeue, and will be smol, so can stay on the heap for now
+        let mut word_positions: VecDeque<(Rc<str>, u16)> =
+            VecDeque::with_capacity(MAX_DISTANCE as usize);
+
+        let docid = document_change.docid();
+        match document_change {
+            DocumentChange::Deletion(inner) => {
+                let document = inner.current(rtxn, index, context.db_fields_ids_map)?;
+                process_document_tokens(
+                    document,
+                    document_tokenizer,
+                    new_fields_ids_map,
+                    &mut word_positions,
+                    &mut |(w1, w2), prox| {
+                        del_word_pair_proximity.push(((w1, w2), prox));
+                    },
+                )?;
+            }
+            DocumentChange::Update(inner) => {
+                let document = inner.current(rtxn, index, context.db_fields_ids_map)?;
+                process_document_tokens(
+                    document,
+                    document_tokenizer,
+                    new_fields_ids_map,
+                    &mut word_positions,
+                    &mut |(w1, w2), prox| {
+                        del_word_pair_proximity.push(((w1, w2), prox));
+                    },
+                )?;
+                let document = inner.merged(rtxn, index, context.db_fields_ids_map)?;
+                process_document_tokens(
+                    document,
+                    document_tokenizer,
+                    new_fields_ids_map,
+                    &mut word_positions,
+                    &mut |(w1, w2), prox| {
+                        add_word_pair_proximity.push(((w1, w2), prox));
+                    },
+                )?;
+            }
+            DocumentChange::Insertion(inner) => {
+                let document = inner.inserted();
+                process_document_tokens(
+                    document,
+                    document_tokenizer,
+                    new_fields_ids_map,
+                    &mut word_positions,
+                    &mut |(w1, w2), prox| {
+                        add_word_pair_proximity.push(((w1, w2), prox));
+                    },
+                )?;
+            }
+        }
+
+        del_word_pair_proximity.sort_unstable();
+        del_word_pair_proximity.dedup_by(|(k1, _), (k2, _)| k1 == k2);
+        for ((w1, w2), prox) in del_word_pair_proximity.iter() {
+            let key = build_key(*prox, w1, w2, &mut key_buffer);
+            cached_sorter.insert_del_u32(key, docid)?;
+        }
+
+        add_word_pair_proximity.sort_unstable();
+        add_word_pair_proximity.dedup_by(|(k1, _), (k2, _)| k1 == k2);
+        for ((w1, w2), prox) in add_word_pair_proximity.iter() {
+            let key = build_key(*prox, w1, w2, &mut key_buffer);
+            cached_sorter.insert_add_u32(key, docid)?;
+        }
+        Ok(())
+    }
+}
+
+fn build_key<'a>(
+    prox: u8,
+    w1: &str,
+    w2: &str,
+    key_buffer: &'a mut bumpalo::collections::Vec<u8>,
+) -> &'a [u8] {
+    key_buffer.clear();
+    key_buffer.push(prox);
+    key_buffer.extend_from_slice(w1.as_bytes());
+    key_buffer.push(0);
+    key_buffer.extend_from_slice(w2.as_bytes());
+    key_buffer.as_slice()
+}
+
+fn word_positions_into_word_pair_proximity(
+    word_positions: &mut VecDeque<(Rc<str>, u16)>,
+    word_pair_proximity: &mut impl FnMut((Rc<str>, Rc<str>), u8),
+) {
+    let (head_word, head_position) = word_positions.pop_front().unwrap();
+    for (word, position) in word_positions.iter() {
+        let prox = index_proximity(head_position as u32, *position as u32) as u8;
+        if prox > 0 && prox < MAX_DISTANCE as u8 {
+            word_pair_proximity((head_word.clone(), word.clone()), prox);
+        }
+    }
+}
+
+fn drain_word_positions(
+    word_positions: &mut VecDeque<(Rc<str>, u16)>,
+    word_pair_proximity: &mut impl FnMut((Rc<str>, Rc<str>), u8),
+) {
+    while !word_positions.is_empty() {
+        word_positions_into_word_pair_proximity(word_positions, word_pair_proximity);
+    }
+}
+
+fn process_document_tokens<'doc>(
+    document: impl Document<'doc>,
+    document_tokenizer: &DocumentTokenizer,
+    fields_ids_map: &mut GlobalFieldsIdsMap,
+    word_positions: &mut VecDeque<(Rc<str>, u16)>,
+    word_pair_proximity: &mut impl FnMut((Rc<str>, Rc<str>), u8),
+) -> Result<()> {
+    let mut field_id = None;
+    let mut token_fn = |_fname: &str, fid: FieldId, pos: u16, word: &str| {
+        if field_id != Some(fid) {
+            field_id = Some(fid);
+            drain_word_positions(word_positions, word_pair_proximity);
+        }
+        // drain the proximity window until the head word is considered close to the word we are inserting.
+        while word_positions
+            .front()
+            .map_or(false, |(_w, p)| index_proximity(*p as u32, pos as u32) >= MAX_DISTANCE)
+        {
+            word_positions_into_word_pair_proximity(word_positions, word_pair_proximity);
+        }
+
+        // insert the new word.
+        word_positions.push_back((Rc::from(word), pos));
+        Ok(())
+    };
+    document_tokenizer.tokenize_document(document, fields_ids_map, &mut token_fn)?;
+
+    drain_word_positions(word_positions, word_pair_proximity);
+    Ok(())
+}

crates/milli/src/update/new/extract/searchable/mod.rs

@@ -0,0 +1,156 @@
+mod extract_word_docids;
+mod extract_word_pair_proximity_docids;
+mod tokenize_document;
+
+use std::cell::RefCell;
+use std::marker::PhantomData;
+
+use bumpalo::Bump;
+pub use extract_word_docids::{WordDocidsCaches, WordDocidsExtractors};
+pub use extract_word_pair_proximity_docids::WordPairProximityDocidsExtractor;
+use heed::RoTxn;
+use tokenize_document::{tokenizer_builder, DocumentTokenizer};
+
+use super::cache::BalancedCaches;
+use super::DocidsExtractor;
+use crate::update::new::indexer::document_changes::{
+    extract, DocumentChangeContext, DocumentChanges, Extractor, IndexingContext, Progress,
+};
+use crate::update::new::steps::Step;
+use crate::update::new::thread_local::{FullySend, ThreadLocal};
+use crate::update::new::DocumentChange;
+use crate::update::GrenadParameters;
+use crate::{Index, Result, MAX_POSITION_PER_ATTRIBUTE};
+
+pub struct SearchableExtractorData<'a, EX: SearchableExtractor> {
+    tokenizer: &'a DocumentTokenizer<'a>,
+    grenad_parameters: GrenadParameters,
+    buckets: usize,
+    _ex: PhantomData<EX>,
+}
+
+impl<'a, 'extractor, EX: SearchableExtractor + Sync> Extractor<'extractor>
+    for SearchableExtractorData<'a, EX>
+{
+    type Data = RefCell<BalancedCaches<'extractor>>;
+
+    fn init_data(&self, extractor_alloc: &'extractor Bump) -> Result<Self::Data> {
+        Ok(RefCell::new(BalancedCaches::new_in(
+            self.buckets,
+            self.grenad_parameters.max_memory_by_thread(),
+            extractor_alloc,
+        )))
+    }
+
+    fn process<'doc>(
+        &self,
+        changes: impl Iterator<Item = Result<DocumentChange<'doc>>>,
+        context: &DocumentChangeContext<Self::Data>,
+    ) -> Result<()> {
+        for change in changes {
+            let change = change?;
+            EX::extract_document_change(context, self.tokenizer, change)?;
+        }
+        Ok(())
+    }
+}
+
+pub trait SearchableExtractor: Sized + Sync {
+    fn run_extraction<'pl, 'fid, 'indexer, 'index, 'extractor, DC: DocumentChanges<'pl>, MSP, SP>(
+        grenad_parameters: GrenadParameters,
+        document_changes: &DC,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP, SP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: Step,
+    ) -> Result<Vec<BalancedCaches<'extractor>>>
+    where
+        MSP: Fn() -> bool + Sync,
+        SP: Fn(Progress) + Sync,
+    {
+        let rtxn = indexing_context.index.read_txn()?;
+        let stop_words = indexing_context.index.stop_words(&rtxn)?;
+        let allowed_separators = indexing_context.index.allowed_separators(&rtxn)?;
+        let allowed_separators: Option<Vec<_>> =
+            allowed_separators.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let dictionary = indexing_context.index.dictionary(&rtxn)?;
+        let dictionary: Option<Vec<_>> =
+            dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let mut builder = tokenizer_builder(
+            stop_words.as_ref(),
+            allowed_separators.as_deref(),
+            dictionary.as_deref(),
+        );
+        let tokenizer = builder.build();
+
+        let attributes_to_extract = Self::attributes_to_extract(&rtxn, indexing_context.index)?;
+        let attributes_to_skip = Self::attributes_to_skip(&rtxn, indexing_context.index)?;
+        let localized_attributes_rules =
+            indexing_context.index.localized_attributes_rules(&rtxn)?.unwrap_or_default();
+
+        let document_tokenizer = DocumentTokenizer {
+            tokenizer: &tokenizer,
+            attribute_to_extract: attributes_to_extract.as_deref(),
+            attribute_to_skip: attributes_to_skip.as_slice(),
+            localized_attributes_rules: &localized_attributes_rules,
+            max_positions_per_attributes: MAX_POSITION_PER_ATTRIBUTE,
+        };
+
+        let extractor_data: SearchableExtractorData<Self> = SearchableExtractorData {
+            tokenizer: &document_tokenizer,
+            grenad_parameters,
+            buckets: rayon::current_num_threads(),
+            _ex: PhantomData,
+        };
+
+        let datastore = ThreadLocal::new();
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::extract", "docids_extraction");
+            let _entered = span.enter();
+            extract(
+                document_changes,
+                &extractor_data,
+                indexing_context,
+                extractor_allocs,
+                &datastore,
+                step,
+            )?;
+        }
+
+        Ok(datastore.into_iter().map(RefCell::into_inner).collect())
+    }
+
+    fn extract_document_change(
+        context: &DocumentChangeContext<RefCell<BalancedCaches>>,
+        document_tokenizer: &DocumentTokenizer,
+        document_change: DocumentChange,
+    ) -> Result<()>;
+
+    fn attributes_to_extract<'a>(rtxn: &'a RoTxn, index: &'a Index)
+        -> Result<Option<Vec<&'a str>>>;
+
+    fn attributes_to_skip<'a>(rtxn: &'a RoTxn, index: &'a Index) -> Result<Vec<&'a str>>;
+}
+
+impl<T: SearchableExtractor> DocidsExtractor for T {
+    fn run_extraction<'pl, 'fid, 'indexer, 'index, 'extractor, DC: DocumentChanges<'pl>, MSP, SP>(
+        grenad_parameters: GrenadParameters,
+        document_changes: &DC,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP, SP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: Step,
+    ) -> Result<Vec<BalancedCaches<'extractor>>>
+    where
+        MSP: Fn() -> bool + Sync,
+        SP: Fn(Progress) + Sync,
+    {
+        Self::run_extraction(
+            grenad_parameters,
+            document_changes,
+            indexing_context,
+            extractor_allocs,
+            step,
+        )
+    }
+}

crates/milli/src/update/new/extract/searchable/tokenize_document.rs

@@ -0,0 +1,282 @@
+use std::collections::HashMap;
+
+use charabia::{SeparatorKind, Token, TokenKind, Tokenizer, TokenizerBuilder};
+use serde_json::Value;
+
+use crate::update::new::document::Document;
+use crate::update::new::extract::perm_json_p::{
+    seek_leaf_values_in_array, seek_leaf_values_in_object, select_field, Depth, Selection,
+};
+use crate::{
+    FieldId, GlobalFieldsIdsMap, InternalError, LocalizedAttributesRule, Result, UserError,
+    MAX_WORD_LENGTH,
+};
+
+// todo: should be crate::proximity::MAX_DISTANCE but it has been forgotten
+const MAX_DISTANCE: u32 = 8;
+
+pub struct DocumentTokenizer<'a> {
+    pub tokenizer: &'a Tokenizer<'a>,
+    pub attribute_to_extract: Option<&'a [&'a str]>,
+    pub attribute_to_skip: &'a [&'a str],
+    pub localized_attributes_rules: &'a [LocalizedAttributesRule],
+    pub max_positions_per_attributes: u32,
+}
+
+impl<'a> DocumentTokenizer<'a> {
+    pub fn tokenize_document<'doc>(
+        &self,
+        document: impl Document<'doc>,
+        field_id_map: &mut GlobalFieldsIdsMap,
+        token_fn: &mut impl FnMut(&str, FieldId, u16, &str) -> Result<()>,
+    ) -> Result<()> {
+        let mut field_position = HashMap::new();
+
+        for entry in document.iter_top_level_fields() {
+            let (field_name, value) = entry?;
+
+            let mut tokenize_field = |field_name: &str, _depth, value: &Value| {
+                let Some(field_id) = field_id_map.id_or_insert(field_name) else {
+                    return Err(UserError::AttributeLimitReached.into());
+                };
+
+                if select_field(field_name, self.attribute_to_extract, self.attribute_to_skip)
+                    != Selection::Select
+                {
+                    return Ok(());
+                }
+
+                let position = field_position
+                    .entry(field_id)
+                    .and_modify(|counter| *counter += MAX_DISTANCE)
+                    .or_insert(0);
+                if *position >= self.max_positions_per_attributes {
+                    return Ok(());
+                }
+
+                let text;
+                let tokens = match value {
+                    Value::Number(n) => {
+                        text = n.to_string();
+                        self.tokenizer.tokenize(text.as_str())
+                    }
+                    Value::Bool(b) => {
+                        text = b.to_string();
+                        self.tokenizer.tokenize(text.as_str())
+                    }
+                    Value::String(text) => {
+                        let locales = self
+                            .localized_attributes_rules
+                            .iter()
+                            .find(|rule| rule.match_str(field_name))
+                            .map(|rule| rule.locales());
+                        self.tokenizer.tokenize_with_allow_list(text.as_str(), locales)
+                    }
+                    _ => return Ok(()),
+                };
+
+                // create an iterator of token with their positions.
+                let tokens = process_tokens(*position, tokens)
+                    .take_while(|(p, _)| *p < self.max_positions_per_attributes);
+
+                for (index, token) in tokens {
+                    // keep a word only if it is not empty and fit in a LMDB key.
+                    let token = token.lemma().trim();
+                    if !token.is_empty() && token.len() <= MAX_WORD_LENGTH {
+                        *position = index;
+                        if let Ok(position) = (*position).try_into() {
+                            token_fn(field_name, field_id, position, token)?;
+                        }
+                    }
+                }
+
+                Ok(())
+            };
+
+            // parse json.
+            match serde_json::to_value(value).map_err(InternalError::SerdeJson)? {
+                Value::Object(object) => seek_leaf_values_in_object(
+                    &object,
+                    None,
+                    &[],
+                    field_name,
+                    Depth::OnBaseKey,
+                    &mut tokenize_field,
+                )?,
+                Value::Array(array) => seek_leaf_values_in_array(
+                    &array,
+                    None,
+                    &[],
+                    field_name,
+                    Depth::OnBaseKey,
+                    &mut tokenize_field,
+                )?,
+                value => tokenize_field(field_name, Depth::OnBaseKey, &value)?,
+            }
+        }
+
+        Ok(())
+    }
+}
+
+/// take an iterator on tokens and compute their relative position depending on separator kinds
+/// if it's an `Hard` separator we add an additional relative proximity of MAX_DISTANCE between words,
+/// else we keep the standard proximity of 1 between words.
+fn process_tokens<'a>(
+    start_offset: u32,
+    tokens: impl Iterator<Item = Token<'a>>,
+) -> impl Iterator<Item = (u32, Token<'a>)> {
+    tokens
+        .skip_while(|token| token.is_separator())
+        .scan((start_offset, None), |(offset, prev_kind), mut token| {
+            match token.kind {
+                TokenKind::Word | TokenKind::StopWord if !token.lemma().is_empty() => {
+                    *offset += match *prev_kind {
+                        Some(TokenKind::Separator(SeparatorKind::Hard)) => MAX_DISTANCE,
+                        Some(_) => 1,
+                        None => 0,
+                    };
+                    *prev_kind = Some(token.kind)
+                }
+                TokenKind::Separator(SeparatorKind::Hard) => {
+                    *prev_kind = Some(token.kind);
+                }
+                TokenKind::Separator(SeparatorKind::Soft)
+                    if *prev_kind != Some(TokenKind::Separator(SeparatorKind::Hard)) =>
+                {
+                    *prev_kind = Some(token.kind);
+                }
+                _ => token.kind = TokenKind::Unknown,
+            }
+            Some((*offset, token))
+        })
+        .filter(|(_, t)| t.is_word())
+}
+
+/// Factorize tokenizer building.
+pub fn tokenizer_builder<'a>(
+    stop_words: Option<&'a fst::Set<&'a [u8]>>,
+    allowed_separators: Option<&'a [&str]>,
+    dictionary: Option<&'a [&str]>,
+) -> TokenizerBuilder<'a, &'a [u8]> {
+    let mut tokenizer_builder = TokenizerBuilder::new();
+    if let Some(stop_words) = stop_words {
+        tokenizer_builder.stop_words(stop_words);
+    }
+    if let Some(dictionary) = dictionary {
+        tokenizer_builder.words_dict(dictionary);
+    }
+    if let Some(separators) = allowed_separators {
+        tokenizer_builder.separators(separators);
+    }
+
+    tokenizer_builder
+}
+
+#[cfg(test)]
+mod test {
+    use bumpalo::Bump;
+    use charabia::TokenizerBuilder;
+    use meili_snap::snapshot;
+    use raw_collections::RawMap;
+    use serde_json::json;
+    use serde_json::value::RawValue;
+
+    use super::*;
+    use crate::fields_ids_map::metadata::{FieldIdMapWithMetadata, MetadataBuilder};
+    use crate::update::new::document::{DocumentFromVersions, Versions};
+    use crate::FieldsIdsMap;
+
+    #[test]
+    fn test_tokenize_document() {
+        let mut fields_ids_map = FieldsIdsMap::new();
+
+        let document = json!({
+            "doggo": {                "name": "doggo",
+            "age": 10,},
+            "catto": {
+                "catto": {
+                    "name": "pesti",
+                    "age": 23,
+                }
+            },
+            "doggo.name": ["doggo", "catto"],
+            "not-me": "UNSEARCHABLE",
+            "me-nether": {"nope": "unsearchable"}
+        });
+
+        let _field_1_id = fields_ids_map.insert("doggo").unwrap();
+        let _field_2_id = fields_ids_map.insert("catto").unwrap();
+        let _field_3_id = fields_ids_map.insert("doggo.name").unwrap();
+        let _field_4_id = fields_ids_map.insert("not-me").unwrap();
+        let _field_5_id = fields_ids_map.insert("me-nether").unwrap();
+
+        let mut tb = TokenizerBuilder::default();
+        let document_tokenizer = DocumentTokenizer {
+            tokenizer: &tb.build(),
+            attribute_to_extract: None,
+            attribute_to_skip: &["not-me", "me-nether.nope"],
+            localized_attributes_rules: &[],
+            max_positions_per_attributes: 1000,
+        };
+
+        let fields_ids_map = FieldIdMapWithMetadata::new(
+            fields_ids_map,
+            MetadataBuilder::new(Default::default(), Default::default(), Default::default(), None),
+        );
+
+        let fields_ids_map_lock = std::sync::RwLock::new(fields_ids_map);
+        let mut global_fields_ids_map = GlobalFieldsIdsMap::new(&fields_ids_map_lock);
+
+        let mut words = std::collections::BTreeMap::new();
+
+        let document = document.to_string();
+
+        let bump = Bump::new();
+        let document: &RawValue = serde_json::from_str(&document).unwrap();
+        let document = RawMap::from_raw_value(document, &bump).unwrap();
+
+        let document = Versions::single(document);
+        let document = DocumentFromVersions::new(&document);
+
+        document_tokenizer
+            .tokenize_document(
+                document,
+                &mut global_fields_ids_map,
+                &mut |_fname, fid, pos, word| {
+                    words.insert([fid, pos], word.to_string());
+                    Ok(())
+                },
+            )
+            .unwrap();
+
+        snapshot!(format!("{:#?}", words), @r###"
+        {
+            [
+                2,
+                0,
+            ]: "doggo",
+            [
+                2,
+                8,
+            ]: "doggo",
+            [
+                2,
+                16,
+            ]: "catto",
+            [
+                5,
+                0,
+            ]: "10",
+            [
+                7,
+                0,
+            ]: "pesti",
+            [
+                8,
+                0,
+            ]: "23",
+        }
+        "###);
+    }
+}

crates/milli/src/update/new/extract/vectors/mod.rs

@@ -0,0 +1,489 @@
+use std::cell::RefCell;
+
+use bumpalo::collections::Vec as BVec;
+use bumpalo::Bump;
+use hashbrown::{DefaultHashBuilder, HashMap};
+
+use super::cache::DelAddRoaringBitmap;
+use crate::error::FaultSource;
+use crate::prompt::Prompt;
+use crate::update::new::channel::EmbeddingSender;
+use crate::update::new::indexer::document_changes::{DocumentChangeContext, Extractor};
+use crate::update::new::thread_local::MostlySend;
+use crate::update::new::vector_document::VectorDocument;
+use crate::update::new::DocumentChange;
+use crate::vector::error::{
+    EmbedErrorKind, PossibleEmbeddingMistakes, UnusedVectorsDistributionBump,
+};
+use crate::vector::{Embedder, Embedding, EmbeddingConfigs};
+use crate::{DocumentId, FieldDistribution, InternalError, Result, ThreadPoolNoAbort, UserError};
+
+pub struct EmbeddingExtractor<'a> {
+    embedders: &'a EmbeddingConfigs,
+    sender: &'a EmbeddingSender<'a>,
+    possible_embedding_mistakes: PossibleEmbeddingMistakes,
+    threads: &'a ThreadPoolNoAbort,
+}
+
+impl<'a> EmbeddingExtractor<'a> {
+    pub fn new(
+        embedders: &'a EmbeddingConfigs,
+        sender: &'a EmbeddingSender<'a>,
+        field_distribution: &'a FieldDistribution,
+        threads: &'a ThreadPoolNoAbort,
+    ) -> Self {
+        let possible_embedding_mistakes = PossibleEmbeddingMistakes::new(field_distribution);
+        Self { embedders, sender, threads, possible_embedding_mistakes }
+    }
+}
+
+pub struct EmbeddingExtractorData<'extractor>(
+    pub HashMap<String, DelAddRoaringBitmap, DefaultHashBuilder, &'extractor Bump>,
+);
+
+unsafe impl MostlySend for EmbeddingExtractorData<'_> {}
+
+impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
+    type Data = RefCell<EmbeddingExtractorData<'extractor>>;
+
+    fn init_data<'doc>(&'doc self, extractor_alloc: &'extractor Bump) -> crate::Result<Self::Data> {
+        Ok(RefCell::new(EmbeddingExtractorData(HashMap::new_in(extractor_alloc))))
+    }
+
+    fn process<'doc>(
+        &'doc self,
+        changes: impl Iterator<Item = crate::Result<DocumentChange<'doc>>>,
+        context: &'doc DocumentChangeContext<Self::Data>,
+    ) -> crate::Result<()> {
+        let embedders = self.embedders.inner_as_ref();
+        let mut unused_vectors_distribution =
+            UnusedVectorsDistributionBump::new_in(&context.doc_alloc);
+
+        let mut all_chunks = BVec::with_capacity_in(embedders.len(), &context.doc_alloc);
+        for (embedder_name, (embedder, prompt, _is_quantized)) in embedders {
+            let embedder_id =
+                context.index.embedder_category_id.get(&context.rtxn, embedder_name)?.ok_or_else(
+                    || InternalError::DatabaseMissingEntry {
+                        db_name: "embedder_category_id",
+                        key: None,
+                    },
+                )?;
+            all_chunks.push(Chunks::new(
+                embedder,
+                embedder_id,
+                embedder_name,
+                prompt,
+                context.data,
+                &self.possible_embedding_mistakes,
+                self.threads,
+                self.sender,
+                &context.doc_alloc,
+            ))
+        }
+
+        for change in changes {
+            let change = change?;
+            match change {
+                DocumentChange::Deletion(deletion) => {
+                    // vector deletion is handled by document sender,
+                    // we still need to accomodate deletion from user_provided
+                    for chunks in &mut all_chunks {
+                        // regenerate: true means we delete from user_provided
+                        chunks.set_regenerate(deletion.docid(), true);
+                    }
+                }
+                DocumentChange::Update(update) => {
+                    let old_vectors = update.current_vectors(
+                        &context.rtxn,
+                        context.index,
+                        context.db_fields_ids_map,
+                        &context.doc_alloc,
+                    )?;
+                    let new_vectors = update.updated_vectors(&context.doc_alloc, self.embedders)?;
+
+                    if let Some(new_vectors) = &new_vectors {
+                        unused_vectors_distribution.append(new_vectors)?;
+                    }
+
+                    for chunks in &mut all_chunks {
+                        let embedder_name = chunks.embedder_name();
+                        let prompt = chunks.prompt();
+
+                        let old_vectors = old_vectors.vectors_for_key(embedder_name)?.unwrap();
+                        if let Some(new_vectors) = new_vectors.as_ref().and_then(|new_vectors| {
+                            new_vectors.vectors_for_key(embedder_name).transpose()
+                        }) {
+                            let new_vectors = new_vectors?;
+                            if old_vectors.regenerate != new_vectors.regenerate {
+                                chunks.set_regenerate(update.docid(), new_vectors.regenerate);
+                            }
+                            // do we have set embeddings?
+                            if let Some(embeddings) = new_vectors.embeddings {
+                                chunks.set_vectors(
+                                    update.docid(),
+                                    embeddings
+                                        .into_vec(&context.doc_alloc, embedder_name)
+                                        .map_err(|error| UserError::InvalidVectorsEmbedderConf {
+                                            document_id: update.external_document_id().to_string(),
+                                            error: error.to_string(),
+                                        })?,
+                                );
+                            } else if new_vectors.regenerate {
+                                let new_rendered = prompt.render_document(
+                                    update.current(
+                                        &context.rtxn,
+                                        context.index,
+                                        context.db_fields_ids_map,
+                                    )?,
+                                    context.new_fields_ids_map,
+                                    &context.doc_alloc,
+                                )?;
+                                let old_rendered = prompt.render_document(
+                                    update.merged(
+                                        &context.rtxn,
+                                        context.index,
+                                        context.db_fields_ids_map,
+                                    )?,
+                                    context.new_fields_ids_map,
+                                    &context.doc_alloc,
+                                )?;
+                                if new_rendered != old_rendered {
+                                    chunks.set_autogenerated(
+                                        update.docid(),
+                                        update.external_document_id(),
+                                        new_rendered,
+                                        &unused_vectors_distribution,
+                                    )?;
+                                }
+                            }
+                        } else if old_vectors.regenerate {
+                            let old_rendered = prompt.render_document(
+                                update.current(
+                                    &context.rtxn,
+                                    context.index,
+                                    context.db_fields_ids_map,
+                                )?,
+                                context.new_fields_ids_map,
+                                &context.doc_alloc,
+                            )?;
+                            let new_rendered = prompt.render_document(
+                                update.merged(
+                                    &context.rtxn,
+                                    context.index,
+                                    context.db_fields_ids_map,
+                                )?,
+                                context.new_fields_ids_map,
+                                &context.doc_alloc,
+                            )?;
+                            if new_rendered != old_rendered {
+                                chunks.set_autogenerated(
+                                    update.docid(),
+                                    update.external_document_id(),
+                                    new_rendered,
+                                    &unused_vectors_distribution,
+                                )?;
+                            }
+                        }
+                    }
+                }
+                DocumentChange::Insertion(insertion) => {
+                    let new_vectors =
+                        insertion.inserted_vectors(&context.doc_alloc, self.embedders)?;
+                    if let Some(new_vectors) = &new_vectors {
+                        unused_vectors_distribution.append(new_vectors)?;
+                    }
+
+                    for chunks in &mut all_chunks {
+                        let embedder_name = chunks.embedder_name();
+                        let prompt = chunks.prompt();
+                        // if no inserted vectors, then regenerate: true + no embeddings => autogenerate
+                        if let Some(new_vectors) = new_vectors.as_ref().and_then(|new_vectors| {
+                            new_vectors.vectors_for_key(embedder_name).transpose()
+                        }) {
+                            let new_vectors = new_vectors?;
+                            chunks.set_regenerate(insertion.docid(), new_vectors.regenerate);
+                            if let Some(embeddings) = new_vectors.embeddings {
+                                chunks.set_vectors(
+                                    insertion.docid(),
+                                    embeddings
+                                        .into_vec(&context.doc_alloc, embedder_name)
+                                        .map_err(|error| UserError::InvalidVectorsEmbedderConf {
+                                            document_id: insertion
+                                                .external_document_id()
+                                                .to_string(),
+                                            error: error.to_string(),
+                                        })?,
+                                );
+                            } else if new_vectors.regenerate {
+                                let rendered = prompt.render_document(
+                                    insertion.inserted(),
+                                    context.new_fields_ids_map,
+                                    &context.doc_alloc,
+                                )?;
+                                chunks.set_autogenerated(
+                                    insertion.docid(),
+                                    insertion.external_document_id(),
+                                    rendered,
+                                    &unused_vectors_distribution,
+                                )?;
+                            }
+                        } else {
+                            let rendered = prompt.render_document(
+                                insertion.inserted(),
+                                context.new_fields_ids_map,
+                                &context.doc_alloc,
+                            )?;
+                            chunks.set_autogenerated(
+                                insertion.docid(),
+                                insertion.external_document_id(),
+                                rendered,
+                                &unused_vectors_distribution,
+                            )?;
+                        }
+                    }
+                }
+            }
+        }
+
+        for chunk in all_chunks {
+            chunk.drain(&unused_vectors_distribution)?;
+        }
+        Ok(())
+    }
+}
+
+// **Warning**: the destructor of this struct is not normally run, make sure that all its fields:
+// 1. don't have side effects tied to they destructors
+// 2. if allocated, are allocated inside of the bumpalo
+//
+// Currently this is the case as:
+// 1. BVec are inside of the bumaplo
+// 2. All other fields are either trivial (u8) or references.
+struct Chunks<'a, 'extractor> {
+    texts: BVec<'a, &'a str>,
+    ids: BVec<'a, DocumentId>,
+
+    embedder: &'a Embedder,
+    embedder_id: u8,
+    embedder_name: &'a str,
+    prompt: &'a Prompt,
+    possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
+    user_provided: &'a RefCell<EmbeddingExtractorData<'extractor>>,
+    threads: &'a ThreadPoolNoAbort,
+    sender: &'a EmbeddingSender<'a>,
+    has_manual_generation: Option<&'a str>,
+}
+
+impl<'a, 'extractor> Chunks<'a, 'extractor> {
+    #[allow(clippy::too_many_arguments)]
+    pub fn new(
+        embedder: &'a Embedder,
+        embedder_id: u8,
+        embedder_name: &'a str,
+        prompt: &'a Prompt,
+        user_provided: &'a RefCell<EmbeddingExtractorData<'extractor>>,
+        possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
+        threads: &'a ThreadPoolNoAbort,
+        sender: &'a EmbeddingSender<'a>,
+        doc_alloc: &'a Bump,
+    ) -> Self {
+        let capacity = embedder.prompt_count_in_chunk_hint() * embedder.chunk_count_hint();
+        let texts = BVec::with_capacity_in(capacity, doc_alloc);
+        let ids = BVec::with_capacity_in(capacity, doc_alloc);
+        Self {
+            texts,
+            ids,
+            embedder,
+            prompt,
+            possible_embedding_mistakes,
+            threads,
+            sender,
+            embedder_id,
+            embedder_name,
+            user_provided,
+            has_manual_generation: None,
+        }
+    }
+
+    pub fn set_autogenerated(
+        &mut self,
+        docid: DocumentId,
+        external_docid: &'a str,
+        rendered: &'a str,
+        unused_vectors_distribution: &UnusedVectorsDistributionBump,
+    ) -> Result<()> {
+        let is_manual = matches!(&self.embedder, &Embedder::UserProvided(_));
+        if is_manual {
+            self.has_manual_generation.get_or_insert(external_docid);
+        }
+
+        if self.texts.len() < self.texts.capacity() {
+            self.texts.push(rendered);
+            self.ids.push(docid);
+            return Ok(());
+        }
+
+        Self::embed_chunks(
+            &mut self.texts,
+            &mut self.ids,
+            self.embedder,
+            self.embedder_id,
+            self.embedder_name,
+            self.possible_embedding_mistakes,
+            unused_vectors_distribution,
+            self.threads,
+            self.sender,
+            self.has_manual_generation.take(),
+        )
+    }
+
+    pub fn drain(
+        mut self,
+        unused_vectors_distribution: &UnusedVectorsDistributionBump,
+    ) -> Result<()> {
+        let res = Self::embed_chunks(
+            &mut self.texts,
+            &mut self.ids,
+            self.embedder,
+            self.embedder_id,
+            self.embedder_name,
+            self.possible_embedding_mistakes,
+            unused_vectors_distribution,
+            self.threads,
+            self.sender,
+            self.has_manual_generation,
+        );
+        // optimization: don't run bvec dtors as they only contain bumpalo allocated stuff
+        std::mem::forget(self);
+        res
+    }
+
+    #[allow(clippy::too_many_arguments)]
+    pub fn embed_chunks(
+        texts: &mut BVec<'a, &'a str>,
+        ids: &mut BVec<'a, DocumentId>,
+        embedder: &Embedder,
+        embedder_id: u8,
+        embedder_name: &str,
+        possible_embedding_mistakes: &PossibleEmbeddingMistakes,
+        unused_vectors_distribution: &UnusedVectorsDistributionBump,
+        threads: &ThreadPoolNoAbort,
+        sender: &EmbeddingSender<'a>,
+        has_manual_generation: Option<&'a str>,
+    ) -> Result<()> {
+        if let Some(external_docid) = has_manual_generation {
+            let mut msg = format!(
+                r"While embedding documents for embedder `{embedder_name}`: no vectors provided for document `{}`{}",
+                external_docid,
+                if ids.len() > 1 {
+                    format!(" and at least {} other document(s)", ids.len() - 1)
+                } else {
+                    "".to_string()
+                }
+            );
+
+            msg += &format!("\n- Note: `{embedder_name}` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.{embedder_name}`.");
+
+            let mut hint_count = 0;
+
+            for (vector_misspelling, count) in possible_embedding_mistakes.vector_mistakes().take(2)
+            {
+                msg += &format!("\n- Hint: try replacing `{vector_misspelling}` by `_vectors` in {count} document(s).");
+                hint_count += 1;
+            }
+
+            for (embedder_misspelling, count) in possible_embedding_mistakes
+                .embedder_mistakes_bump(embedder_name, unused_vectors_distribution)
+                .take(2)
+            {
+                msg += &format!("\n- Hint: try replacing `_vectors.{embedder_misspelling}` by `_vectors.{embedder_name}` in {count} document(s).");
+                hint_count += 1;
+            }
+
+            if hint_count == 0 {
+                msg += &format!(
+                    "\n- Hint: opt-out for a document with `_vectors.{embedder_name}: null`"
+                );
+            }
+
+            return Err(crate::Error::UserError(crate::UserError::DocumentEmbeddingError(msg)));
+        }
+
+        let res = match embedder.embed_chunks_ref(texts.as_slice(), threads) {
+            Ok(embeddings) => {
+                for (docid, embedding) in ids.into_iter().zip(embeddings) {
+                    sender.set_vector(*docid, embedder_id, embedding).unwrap();
+                }
+                Ok(())
+            }
+            Err(error) => {
+                if let FaultSource::Bug = error.fault {
+                    Err(crate::Error::InternalError(crate::InternalError::VectorEmbeddingError(
+                        error.into(),
+                    )))
+                } else {
+                    let mut msg = format!(
+                        r"While embedding documents for embedder `{embedder_name}`: {error}"
+                    );
+
+                    if let EmbedErrorKind::ManualEmbed(_) = &error.kind {
+                        msg += &format!("\n- Note: `{embedder_name}` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.{embedder_name}`.");
+                    }
+
+                    let mut hint_count = 0;
+
+                    for (vector_misspelling, count) in
+                        possible_embedding_mistakes.vector_mistakes().take(2)
+                    {
+                        msg += &format!("\n- Hint: try replacing `{vector_misspelling}` by `_vectors` in {count} document(s).");
+                        hint_count += 1;
+                    }
+
+                    for (embedder_misspelling, count) in possible_embedding_mistakes
+                        .embedder_mistakes_bump(embedder_name, unused_vectors_distribution)
+                        .take(2)
+                    {
+                        msg += &format!("\n- Hint: try replacing `_vectors.{embedder_misspelling}` by `_vectors.{embedder_name}` in {count} document(s).");
+                        hint_count += 1;
+                    }
+
+                    if hint_count == 0 {
+                        if let EmbedErrorKind::ManualEmbed(_) = &error.kind {
+                            msg += &format!(
+                                "\n- Hint: opt-out for a document with `_vectors.{embedder_name}: null`"
+                            );
+                        }
+                    }
+
+                    Err(crate::Error::UserError(crate::UserError::DocumentEmbeddingError(msg)))
+                }
+            }
+        };
+        texts.clear();
+        ids.clear();
+        res
+    }
+
+    pub fn prompt(&self) -> &'a Prompt {
+        self.prompt
+    }
+
+    pub fn embedder_name(&self) -> &'a str {
+        self.embedder_name
+    }
+
+    fn set_regenerate(&self, docid: DocumentId, regenerate: bool) {
+        let mut user_provided = self.user_provided.borrow_mut();
+        let user_provided = user_provided.0.entry_ref(self.embedder_name).or_default();
+        if regenerate {
+            // regenerate == !user_provided
+            user_provided.insert_del_u32(docid);
+        } else {
+            user_provided.insert_add_u32(docid);
+        }
+    }
+
+    fn set_vectors(&self, docid: DocumentId, embeddings: Vec<Embedding>) {
+        self.sender.set_vectors(docid, self.embedder_id, embeddings).unwrap();
+    }
+}