Open Source alternative to Algolia + Pinecone and an Easier-to-Use alternative to ElasticSearch ⚡ 🔍 ✨ Fast, typo tolerant, in-memory fuzzy Search Engine for building delightful search experiences
C++ algoliadatastoreelasticsearchenterprise-searchfacetingfull-text-searchfuzzy-searchgeosearchin-memoryinstantsearchmerchandisingpineconesearchsearch-enginesemantic-searchsimilarity-searchsite-searchsynonymstypo-tolerancevector-search
Fix quirk in single field prefix searching.
Kishore Nallan9d566088194a
diff --git a/src/index.cpp b/src/index.cppindex 5deee436..c10cdf76 100644--- a/src/index.cpp+++ b/src/index.cpp@@ -1164,16 +1164,9 @@ void Index::search_all_candidates(const size_t num_search_fields,std::set<uint64>& query_hashes,std::vector<uint32_t>& id_buff) const {- auto product = []( long long a, tok_candidates & b ) { return a*b.candidates.size(); };- long long int N = std::accumulate(token_candidates_vec.begin(), token_candidates_vec.end(), 1LL, product);-- // escape hatch to prevent too much looping but subject to being overriden explicitly via `max_candidates`- long long combination_limit = std::max<size_t>(Index::COMBINATION_MIN_LIMIT, max_candidates);-/*if(!token_candidates_vec.empty()) {LOG(INFO) << "Prefix candidates size: " << token_candidates_vec.back().candidates.size();LOG(INFO) << "max_candidates: " << max_candidates;- LOG(INFO) << "combination_limit: " << combination_limit;LOG(INFO) << "token_candidates_vec.size(): " << token_candidates_vec.size();}*/@@ -1251,9 +1244,15 @@ void Index::search_all_candidates(const size_t num_search_fields,token_candidates_vec.back().candidates.clear();token_candidates_vec.back().candidates.assign(trimmed_candidates.begin(), trimmed_candidates.end());-}+ auto product = []( long long a, tok_candidates & b ) { return a*b.candidates.size(); };+ long long int N = std::accumulate(token_candidates_vec.begin(), token_candidates_vec.end(), 1LL, product);++ // escape hatch to prevent too much looping but subject to being overriden explicitly via `max_candidates`+ long long combination_limit = (num_search_fields == 1 && prefixes[0]) ? max_candidates :+ std::max<size_t>(Index::COMBINATION_MIN_LIMIT, max_candidates);+for(long long n = 0; n < N && n < combination_limit; ++n) {RETURN_CIRCUIT_BREAKER@@ -2809,8 +2808,7 @@ void Index::fuzzy_search_fields(const std::vector<search_field_t>& the_fields,continue;}- size_t max_words = (num_search_fields == 1 && prefix_search) ? max_candidates : 100000;- // need less candidates for filtered searches since we already only pick tokens with results+ size_t max_words = 100000;art_fuzzy_search(search_index.at(the_field.name), (const unsigned char *) token.c_str(), token_len,costs[token_index], costs[token_index], max_words, token_order, prefix_search,filter_ids, filter_ids_length, leaves, unique_tokens);diff --git a/test/collection_specific_more_test.cpp b/test/collection_specific_more_test.cppindex 0a36fae0..fd68f719 100644--- a/test/collection_specific_more_test.cpp+++ b/test/collection_specific_more_test.cpp@@ -86,6 +86,35 @@ TEST_F(CollectionSpecificMoreTest, PrefixExpansionWhenExactMatchExists) {collectionManager.drop_collection("coll1");}+TEST_F(CollectionSpecificMoreTest, PrefixExpansionOnSingleField) {+ // when no default sorting field is provided, tokens must be ordered on frequency+ Collection *coll1;+ std::vector<field> fields = {field("title", field_types::STRING, false),+ field("points", field_types::INT32, false)};++ coll1 = collectionManager.get_collection("coll1").get();+ if(coll1 == nullptr) {+ coll1 = collectionManager.create_collection("coll1", 1, fields, "points").get();+ }++ std::vector<std::string> tokens = {+ "Mark Jack", "John Jack", "John James", "John Joseph", "John Jim", "John Jordan",+ "Mark Nicholas", "Mark Abbey", "Mark Boucher", "Mark Bicks", "Mark Potter"+ };++ for(size_t i = 0; i < tokens.size(); i++) {+ std::string title = tokens[i];+ nlohmann::json doc;+ doc["title"] = title;+ doc["points"] = i;+ coll1->add(doc.dump());+ }++ // max candidates as default 4+ auto results = coll1->search("mark j", {"title"}, "", {}, {}, {0}, 100, 1, MAX_SCORE, {true}).get();+ ASSERT_EQ("0", results["hits"][0]["document"]["id"].get<std::string>());+}+TEST_F(CollectionSpecificMoreTest, ArrayElementMatchShouldBeMoreImportantThanTotalMatch) {std::vector<field> fields = {field("title", field_types::STRING, false),field("author", field_types::STRING, false),