improve the way we handle the fsts

milli/Cargo.toml

@@ -56,7 +56,6 @@ itertools = "0.10.5"
 log = "0.4.17"
 logging_timer = "1.1.0"
 csv = "1.2.1"
-fastrand = "1.9.0"
 
 [dev-dependencies]
 mimalloc = { version = "0.1.29", default-features = false }
@@ -65,7 +64,10 @@ insta = "1.29.0"
 maplit = "1.0.2"
 md5 = "0.7.0"
 rand = {version = "0.8.5", features = ["small_rng"] }
+
+# fuzzing
 arbitrary = { version = "1.3.0", features = ["derive"] }
+fastrand = "1.9.0"
 
 [target.'cfg(fuzzing)'.dev-dependencies]
 fuzzcheck = "0.12.1"

milli/src/external_documents_ids.rs

@@ -106,8 +106,11 @@ impl<'a> ExternalDocumentsIds<'a> {
         map
     }
 
-    fn merge_soft_into_hard(&mut self) -> fst::Result<()> {
-        if self.soft.len() >= self.hard.len() / 2 {
+    /// Return an fst of the combined hard and soft deleted ID.
+    pub fn to_fst<'b>(&'b self) -> fst::Result<Cow<'b, fst::Map<Cow<'a, [u8]>>>> {
+        if self.soft.is_empty() {
+            return Ok(Cow::Borrowed(&self.hard));
+        }
         let union_op = self.hard.op().add(&self.soft).r#union();
 
         let mut iter = union_op.into_stream();
@@ -121,7 +124,12 @@ impl<'a> ExternalDocumentsIds<'a> {
 
         drop(iter);
 
-            self.hard = new_hard_builder.into_map().map_data(Cow::Owned)?;
+        Ok(Cow::Owned(new_hard_builder.into_map().map_data(Cow::Owned)?))
+    }
+
+    fn merge_soft_into_hard(&mut self) -> fst::Result<()> {
+        if self.soft.len() >= self.hard.len() / 2 {
+            self.hard = self.to_fst()?.into_owned();
             self.soft = fst::Map::default().map_data(Cow::Owned)?;
         }
 

milli/src/update/index_documents/mod.rs

@@ -198,7 +198,6 @@ where
             let number_of_documents = self.index.number_of_documents(self.wtxn)?;
             return Ok(DocumentAdditionResult { indexed_documents: 0, number_of_documents });
         }
-
         let output = self
             .transform
             .take()
@@ -221,7 +220,6 @@ where
         }
 
         let indexed_documents = output.documents_count as u64;
-
         let number_of_documents = self.execute_raw(output)?;
 
         Ok(DocumentAdditionResult { indexed_documents, number_of_documents })

milli/src/update/index_documents/transform.rs

@@ -57,8 +57,8 @@ pub struct Transform<'a, 'i> {
     original_sorter: grenad::Sorter<MergeFn>,
     flattened_sorter: grenad::Sorter<MergeFn>,
 
-    pub replaced_documents_ids: RoaringBitmap,
-    pub new_documents_ids: RoaringBitmap,
+    replaced_documents_ids: RoaringBitmap,
+    new_documents_ids: RoaringBitmap,
     // To increase the cache locality and decrease the heap usage we use compact smartstring.
     new_external_documents_ids_builder: FxHashMap<SmartString<smartstring::Compact>, u64>,
     documents_count: usize,
@@ -653,9 +653,7 @@ impl<'a, 'i> Transform<'a, 'i> {
             primary_key,
             fields_ids_map: self.fields_ids_map,
             field_distribution,
-            new_external_documents_ids: new_external_documents_ids
-                .map_data(|c| Cow::Owned(c))
-                .unwrap(),
+            new_external_documents_ids: new_external_documents_ids.map_data(Cow::Owned).unwrap(),
             new_documents_ids: self.new_documents_ids,
             replaced_documents_ids: self.replaced_documents_ids,
             documents_count: self.documents_count,
@@ -689,8 +687,8 @@ impl<'a, 'i> Transform<'a, 'i> {
         let new_external_documents_ids = {
             let mut external_documents_ids = self.index.external_documents_ids(wtxn)?;
             external_documents_ids.delete_soft_deleted_documents_ids_from_fsts()?;
-            // it is safe to get the hard document IDs
-            external_documents_ids.into_static().hard
+            // This call should be free and can't fail since the previous method merged both fsts.
+            external_documents_ids.into_static().to_fst()?.into_owned()
         };
 
         let documents_ids = self.index.documents_ids(wtxn)?;