Merge pull request #272 from meilisearch/fix-long-words

Ignore words that are too long

meilidb-core/src/database.rs

@@ -387,6 +387,56 @@ mod tests {
         assert_matches!(result, UpdateStatus::Processed(status) if status.result.is_err());
     }
 
+    #[test]
+    fn ignored_words_to_long() {
+        let dir = tempfile::tempdir().unwrap();
+
+        let database = Database::open_or_create(dir.path()).unwrap();
+        let env = &database.env;
+
+        let (sender, receiver) = mpsc::sync_channel(100);
+        let update_fn = move |update: ProcessedUpdateResult| sender.send(update.update_id).unwrap();
+        let index = database.create_index("test").unwrap();
+
+        let done = database.set_update_callback("test", Box::new(update_fn));
+        assert!(done, "could not set the index update function");
+
+        let schema = {
+            let data = r#"
+                identifier = "id"
+
+                [attributes."name"]
+                displayed = true
+                indexed = true
+            "#;
+            toml::from_str(data).unwrap()
+        };
+
+        let mut writer = env.write_txn().unwrap();
+        let _update_id = index.schema_update(&mut writer, schema).unwrap();
+        writer.commit().unwrap();
+
+        let mut additions = index.documents_addition();
+
+        let doc1 = serde_json::json!({
+            "id": 123,
+            "name": "s̷̡̢̡̧̺̜̞͕͉͉͕̜͔̟̼̥̝͍̟̖͔͔̪͉̲̹̝̣̖͎̞̤̥͓͎̭̩͕̙̩̿̀̋̅̈́̌́̏̍̄̽͂̆̾̀̿̕̚̚͜͠͠ͅͅļ̵̨̨̨̰̦̻̳̖̳͚̬̫͚̦͖͈̲̫̣̩̥̻̙̦̱̼̠̖̻̼̘̖͉̪̜̠̙͖̙̩͔̖̯̩̲̿̽͋̔̿̍̓͂̍̿͊͆̃͗̔̎͐͌̾̆͗́̆̒̔̾̅̚̚͜͜ͅͅī̵̛̦̅̔̓͂͌̾́͂͛̎̋͐͆̽̂̋̋́̾̀̉̓̏̽́̑̀͒̇͋͛̈́̃̉̏͊̌̄̽̿̏̇͘̕̚̕p̶̧̛̛̖̯̗͕̝̗̭̱͙̖̗̟̟̐͆̊̂͐̋̓̂̈́̓͊̆͌̾̾͐͋͗͌̆̿̅͆̈́̈́̉͋̍͊͗̌̓̅̈̎̇̃̎̈́̉̐̋͑̃͘̕͘d̴̢̨̛͕̘̯͖̭̮̝̝̐̊̈̅̐̀͒̀́̈́̀͌̽͛͆͑̀̽̿͛̃̋̇̎̀́̂́͘͠͝ǫ̵̨̛̮̩̘͚̬̯̖̱͍̼͑͑̓̐́̑̿̈́̔͌̂̄͐͝ģ̶̧̜͇̣̭̺̪̺̖̻͖̮̭̣̙̻͒͊͗̓̓͒̀̀ͅ",
+        });
+
+        additions.update_document(doc1);
+
+        let mut writer = env.write_txn().unwrap();
+        let update_id = additions.finalize(&mut writer).unwrap();
+        writer.commit().unwrap();
+
+        // block until the transaction is processed
+        let _ = receiver.into_iter().find(|id| *id == update_id);
+
+        let reader = env.read_txn().unwrap();
+        let result = index.update_status(&reader, update_id).unwrap();
+        assert_matches!(result, UpdateStatus::Processed(status) if status.result.is_ok());
+    }
+
     #[test]
     fn add_schema_attributes_at_end() {
         let dir = tempfile::tempdir().unwrap();

meilidb-core/src/raw_indexer.rs

@@ -7,6 +7,8 @@ use meilidb_schema::SchemaAttr;
 use meilidb_tokenizer::{is_cjk, SeqTokenizer, Token, Tokenizer};
 use sdset::SetBuf;
 
+const WORD_LENGTH_LIMIT: usize = 80;
+
 type Word = Vec<u8>; // TODO make it be a SmallVec
 
 pub struct RawIndexer {
@@ -128,21 +130,26 @@ fn index_token(
         match token_to_docindex(id, attr, token) {
             Some(docindex) => {
                 let word = Vec::from(token.word);
-                words_doc_indexes
-                    .entry(word.clone())
-                    .or_insert_with(Vec::new)
-                    .push(docindex);
-                docs_words.entry(id).or_insert_with(Vec::new).push(word);
 
-                if !lower.contains(is_cjk) {
-                    let unidecoded = deunicode_with_tofu(&lower, "");
-                    if unidecoded != lower && !unidecoded.is_empty() {
-                        let word = Vec::from(unidecoded);
-                        words_doc_indexes
-                            .entry(word.clone())
-                            .or_insert_with(Vec::new)
-                            .push(docindex);
-                        docs_words.entry(id).or_insert_with(Vec::new).push(word);
+                if word.len() <= WORD_LENGTH_LIMIT {
+                    words_doc_indexes
+                        .entry(word.clone())
+                        .or_insert_with(Vec::new)
+                        .push(docindex);
+                    docs_words.entry(id).or_insert_with(Vec::new).push(word);
+
+                    if !lower.contains(is_cjk) {
+                        let unidecoded = deunicode_with_tofu(&lower, "");
+                        if unidecoded != lower && !unidecoded.is_empty() {
+                            let word = Vec::from(unidecoded);
+                            if word.len() <= WORD_LENGTH_LIMIT {
+                                words_doc_indexes
+                                    .entry(word.clone())
+                                    .or_insert_with(Vec::new)
+                                    .push(docindex);
+                                docs_words.entry(id).or_insert_with(Vec::new).push(word);
+                            }
+                        }
                     }
                 }
             }