Skip to content

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

Don't consider global context for text match in arrays.

Kishore Nallancf5f71a0ba19
diff --git a/src/index.cpp b/src/index.cpp
index c10cdf76..cabc3d13 100644
--- a/src/index.cpp
+++ b/src/index.cpp
@@ -4170,8 +4170,8 @@ int64_t Index::score_results2(const std::vector<sort_by> & sort_fields, const ui
const Match &match = Match(seq_id, token_positions, false, prioritize_exact_match);
uint64_t this_match_score = match.get_match_score(total_cost, posting_lists.size());
- auto unique_words = ((this_match_score >> 32) & 0xFF);
auto this_words_present = ((this_match_score >> 24) & 0xFF);
+ auto unique_words = field_is_array ? this_words_present : ((this_match_score >> 32) & 0xFF);
auto typo_score = ((this_match_score >> 16) & 0xFF);
auto proximity = ((this_match_score >> 8) & 0xFF);
auto verbatim = (this_match_score & 0xFF);
diff --git a/test/collection_specific_more_test.cpp b/test/collection_specific_more_test.cpp
index fd68f719..c27d8ccf 100644
--- a/test/collection_specific_more_test.cpp
+++ b/test/collection_specific_more_test.cpp
@@ -157,6 +157,40 @@ TEST_F(CollectionSpecificMoreTest, ArrayElementMatchShouldBeMoreImportantThanTot
ASSERT_EQ("1", results["hits"][2]["document"]["id"].get<std::string>());
}
+TEST_F(CollectionSpecificMoreTest, ArrayMatchAcrossElementsMustNotMatter) {
+ std::vector<field> fields = {field("title", field_types::STRING, false),
+ field("author", field_types::STRING, false),
+ field("tags", field_types::STRING_ARRAY, false),};
+
+ Collection* coll1 = collectionManager.create_collection("coll1", 1, fields).get();
+
+ nlohmann::json doc1;
+ doc1["id"] = "0";
+ doc1["title"] = "Por do sol immateur";
+ doc1["author"] = "Vermelho";
+ doc1["tags"] = {"por do sol", "immateur", "gemsor", "praia", "sol", "vermelho", "suyay"};
+
+ nlohmann::json doc2;
+ doc2["id"] = "1";
+ doc2["title"] = "Sunset Rising";
+ doc2["author"] = "Vermelho";
+ doc2["tags"] = {"sunset", "por do sol", "praia", "somao", "vermelho"};
+
+ ASSERT_TRUE(coll1->add(doc1.dump()).ok());
+ ASSERT_TRUE(coll1->add(doc2.dump()).ok());
+
+ auto results = coll1->search("praia por sol vermelho", {"title", "author", "tags"},
+ "", {}, {}, {2}, 10,
+ 1, FREQUENCY, {true},
+ 1, spp::sparse_hash_set<std::string>(),
+ spp::sparse_hash_set<std::string>(), 10, "", 30, 4, "title", 5, {}, {}, {}, 0,
+ "<mark>", "</mark>", {}, 1000, true).get();
+
+ ASSERT_EQ(2, results["hits"].size());
+ ASSERT_EQ("0", results["hits"][0]["document"]["id"].get<std::string>());
+ ASSERT_EQ("1", results["hits"][1]["document"]["id"].get<std::string>());
+}
+
TEST_F(CollectionSpecificMoreTest, MatchedSegmentMoreImportantThanTotalMatches) {
std::vector<field> fields = {field("title", field_types::STRING, false),
field("author", field_types::STRING, false)};