Merge #628

628: Make sure that long words are ignored r=ManyTheFish a=Kerollmops

This PR is related to https://github.com/meilisearch/meilisearch/issues/2743 and is fixing it.

Co-authored-by: Kerollmops <clement@meilisearch.com>

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

@@ -7,7 +7,9 @@ use charabia::{SeparatorKind, Token, TokenKind, TokenizerBuilder};
 use roaring::RoaringBitmap;
 use serde_json::Value;
 
-use super::helpers::{concat_u32s_array, create_sorter, sorter_into_reader, GrenadParameters};
+use super::helpers::{
+    concat_u32s_array, create_sorter, sorter_into_reader, GrenadParameters, MAX_WORD_LENGTH,
+};
 use crate::error::{InternalError, SerializationError};
 use crate::{absolute_from_relative_position, FieldId, Result, MAX_POSITION_PER_ATTRIBUTE};
 
@@ -68,7 +70,7 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
 
                     for (index, token) in tokens {
                         let token = token.lemma().trim();
-                        if !token.is_empty() {
+                        if !token.is_empty() && token.len() <= MAX_WORD_LENGTH {
                             key_buffer.truncate(mem::size_of::<u32>());
                             key_buffer.extend_from_slice(token.as_bytes());
 

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

@@ -18,8 +18,11 @@ pub use merge_functions::{
     roaring_bitmap_from_u32s_array, serialize_roaring_bitmap, MergeFn,
 };
 
+/// The maximum length a word can be
+pub const MAX_WORD_LENGTH: usize = 250;
+
 pub fn valid_lmdb_key(key: impl AsRef<[u8]>) -> bool {
-    key.as_ref().len() <= 511 && !key.as_ref().is_empty()
+    key.as_ref().len() <= MAX_WORD_LENGTH * 2 && !key.as_ref().is_empty()
 }
 
 /// Divides one slice into two at an index, returns `None` if mid is out of bounds.

milli/src/update/index_documents/mod.rs

@@ -1741,4 +1741,22 @@ mod tests {
         index.add_documents(doc3).unwrap_err();
         index.add_documents(doc4).unwrap_err();
     }
+
+    #[test]
+    fn long_words_must_be_skipped() {
+        let index = TempIndex::new();
+
+        // this is obviousy too long
+        let long_word = "lol".repeat(1000);
+        let doc1 = documents! {[{
+            "id": "1",
+            "title": long_word.clone(),
+        }]};
+
+        index.add_documents(doc1).unwrap();
+
+        let rtxn = index.read_txn().unwrap();
+        let words_fst = index.words_fst(&rtxn).unwrap();
+        assert!(!words_fst.contains(&long_word));
+    }
 }