From 2ae3f4097130646880c0d58733dce11a76deceee Mon Sep 17 00:00:00 2001
From: Kerollmops <clement@meilisearch.com>
Date: Wed, 1 Jul 2020 17:49:46 +0200
Subject: [PATCH] Make the indexer ignore certain words

This is a preparation for making the indexing fully parallel by making the
indexer only be aware of certain words for each threads to avoid postings lists
conflicts for each words
---
 Cargo.lock         | 10 ++++++----
 Cargo.toml         |  1 +
 src/bin/indexer.rs | 16 +++++++++++++---
 3 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index e3c89585f..519a2b791 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -817,6 +817,7 @@ dependencies = [
  "linked-hash-map",
  "memmap",
  "once_cell",
+ "rayon",
  "roaring",
  "serde",
  "slice-group-by",
@@ -1342,10 +1343,11 @@ dependencies = [
 
 [[package]]
 name = "rayon"
-version = "1.3.0"
+version = "1.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "db6ce3297f9c85e16621bb8cca38a06779ffc31bb8184e1be4bed2be4678a098"
+checksum = "62f02856753d04e03e26929f820d0a0a337ebe71f849801eea335d464b349080"
 dependencies = [
+ "autocfg 1.0.0",
  "crossbeam-deque",
  "either",
  "rayon-core",
@@ -1353,9 +1355,9 @@ dependencies = [
 
 [[package]]
 name = "rayon-core"
-version = "1.7.0"
+version = "1.7.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "08a89b46efaf957e52b18062fb2f4660f8b8a4dde1807ca002690868ef2c85a9"
+checksum = "e92e15d89083484e11353891f1af602cc661426deb9564c298b270c726973280"
 dependencies = [
  "crossbeam-deque",
  "crossbeam-queue",
diff --git a/Cargo.toml b/Cargo.toml
index 714b1b158..db45da11e 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -19,6 +19,7 @@ levenshtein_automata = { version = "0.2.0", features = ["fst_automaton"] }
 linked-hash-map = "0.5.3"
 memmap = "0.7.0"
 once_cell = "1.4.0"
+rayon = "1.3.1"
 roaring = { git = "https://github.com/Kerollmops/roaring-rs.git", branch = "deserialize-from-slice" }
 slice-group-by = "0.2.6"
 smallstr = "0.2.0"
diff --git a/src/bin/indexer.rs b/src/bin/indexer.rs
index 5a2a2ccfb..af45d27d5 100644
--- a/src/bin/indexer.rs
+++ b/src/bin/indexer.rs
@@ -77,7 +77,14 @@ where
     a
 }
 
-fn index_csv<R: io::Read>(wtxn: &mut heed::RwTxn, mut rdr: csv::Reader<R>, index: &Index) -> anyhow::Result<()> {
+fn index_csv<R: io::Read>(
+    wtxn: &mut heed::RwTxn,
+    mut rdr: csv::Reader<R>,
+    index: &Index,
+    num_threads: usize,
+    thread_index: usize,
+) -> anyhow::Result<()>
+{
     eprintln!("Indexing into LMDB...");
 
     let mut words_cache = ArcCache::<_, (RoaringBitmap, FastMap4<_, RoaringBitmap>)>::new(100_000);
@@ -100,6 +107,9 @@ fn index_csv<R: io::Read>(wtxn: &mut heed::RwTxn, mut rdr: csv::Reader<R>, index
                     let word = word.to_lowercase(); // TODO cow_to_lowercase
                     let position = (attr * 1000 + pos) as u32;
 
+                    // If this indexing process is not concerned by this word, then ignore it.
+                    if fxhash::hash32(&word) as usize % num_threads != thread_index { continue; }
+
                     match words_cache.get_mut(&word) {
                         (Some(entry), evicted) => {
                             let (ids, positions) = entry;
@@ -214,11 +224,11 @@ fn main() -> anyhow::Result<()> {
     match opt.csv_file {
         Some(path) => {
             let rdr = csv::Reader::from_path(path)?;
-            index_csv(&mut wtxn, rdr, &index)?;
+            index_csv(&mut wtxn, rdr, &index, 1, 0)?;
         },
         None => {
             let rdr = csv::Reader::from_reader(io::stdin());
-            index_csv(&mut wtxn, rdr, &index)?;
+            index_csv(&mut wtxn, rdr, &index, 1, 0)?;
         }
     };