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

Fix cascade deletion bug. (#1742)

Harpreet Sangar8e808d4e29f6
diff --git a/include/collection.h b/include/collection.h
index cadfe47e..956d2d9c 100644
--- a/include/collection.h
+++ b/include/collection.h
@@ -601,6 +601,8 @@ public:
Option<nlohmann::json> get(const std::string & id) const;
+ void remove_ref_docs(const uint32_t& count, uint32_t const* const docs, bool remove_from_store = true);
+
Option<std::string> remove(const std::string & id, bool remove_from_store = true);
Option<bool> remove_if_found(uint32_t seq_id, bool remove_from_store = true);
diff --git a/src/collection.cpp b/src/collection.cpp
index 00290da5..c00b6522 100644
--- a/src/collection.cpp
+++ b/src/collection.cpp
@@ -4435,12 +4435,33 @@ void Collection::remove_document(const nlohmann::json & document, const uint32_t
for (const auto &item: referenced_in) {
auto ref_coll = collectionManager.get_collection(item.first);
if (ref_coll != nullptr) {
+ // Delete all the docs where reference helper field has value `seq_id`.
filter_result_t filter_result;
- ref_coll->get_filter_ids(item.second + ":=" + id, filter_result);
- for (uint32_t i = 0; i < filter_result.count; i++) {
- ref_coll->remove(std::to_string(filter_result.docs[i]));
+ ref_coll->get_filter_ids(item.second + ":" + std::to_string(seq_id), filter_result);
+
+ ref_coll->remove_ref_docs(filter_result.count, filter_result.docs, remove_from_store);
+ }
+ }
+}
+
+void Collection::remove_ref_docs(const uint32_t& count, uint32_t const* const docs, bool remove_from_store) {
+ for (uint32_t i = 0; i < count; i++) {
+ auto const& seq_id = docs[i];
+
+ nlohmann::json document;
+ auto get_doc_op = get_document_from_store(get_seq_id_key(seq_id), document);
+
+ if( !get_doc_op.ok()) {
+ if (get_doc_op.code() == 404) {
+ LOG(ERROR) << "`" << name << "` collection: Sequence ID `" << seq_id << "` exists, but document is missing.";
+ continue;
}
+
+ LOG(ERROR) << "`" << name << "` collection: " << get_doc_op.error();
+ continue;
}
+
+ remove_document(document, seq_id, remove_from_store);
}
}
diff --git a/test/collection_join_test.cpp b/test/collection_join_test.cpp
index d77ce55f..f1c6dd04 100644
--- a/test/collection_join_test.cpp
+++ b/test/collection_join_test.cpp
@@ -4280,6 +4280,176 @@ TEST_F(CollectionJoinTest, CascadeDeletion) {
ASSERT_EQ(1, res_obj["found"].get<size_t>());
ASSERT_EQ("product_b", res_obj["hits"][0]["document"].at("product_id"));
ASSERT_EQ("user_a", res_obj["hits"][0]["document"].at("user_id"));
+
+ schema_json =
+ R"({
+ "name": "document",
+ "fields": [
+ {"name": "name", "type": "string"}
+ ]
+ })"_json;
+ documents = {
+ R"({
+ "id": "1",
+ "name": "doc_1"
+ })"_json,
+ R"({
+ "id": "2",
+ "name": "doc_2"
+ })"_json,
+ R"({
+ "id": "3",
+ "name": "doc_3"
+ })"_json,
+ };
+ collection_create_op = collectionManager.create_collection(schema_json);
+ ASSERT_TRUE(collection_create_op.ok());
+ for (auto const &json: documents) {
+ auto add_op = collection_create_op.get()->add(json.dump());
+ ASSERT_TRUE(add_op.ok());
+ }
+
+ schema_json =
+ R"({
+ "name": "lead",
+ "fields": [
+ {"name": "name", "type": "string"}
+ ]
+ })"_json;
+ documents = {
+ R"({
+ "id": "1",
+ "name": "lead_1"
+ })"_json,
+ R"({
+ "id": "2",
+ "name": "lead_2"
+ })"_json,
+ R"({
+ "id": "3",
+ "name": "lead_3"
+ })"_json
+ };
+ collection_create_op = collectionManager.create_collection(schema_json);
+ ASSERT_TRUE(collection_create_op.ok());
+ for (auto const &json: documents) {
+ auto add_op = collection_create_op.get()->add(json.dump());
+ ASSERT_TRUE(add_op.ok());
+ }
+
+ schema_json =
+ R"({
+ "name": "lead_document",
+ "fields": [
+ {"name": "leadId", "type": "string", "reference":"lead.id"},
+ {"name": "documentId", "type": "string", "reference":"document.id"}
+ ]
+ })"_json;
+ documents = {
+ R"({
+ "id": "1",
+ "leadId": "1",
+ "documentId": "1"
+ })"_json,
+ R"({
+ "id": "2",
+ "leadId": "2",
+ "documentId": "2"
+ })"_json,
+ R"({
+ "id": "3",
+ "leadId": "3",
+ "documentId": "2"
+ })"_json
+ };
+ collection_create_op = collectionManager.create_collection(schema_json);
+ ASSERT_TRUE(collection_create_op.ok());
+ for (auto const &json: documents) {
+ auto add_op = collection_create_op.get()->add(json.dump());
+ ASSERT_TRUE(add_op.ok());
+ }
+
+ req_params = {
+ {"collection", "lead_document"},
+ {"q", "*"},
+ };
+ search_op = collectionManager.do_search(req_params, embedded_params, json_res, now_ts);
+ ASSERT_TRUE(search_op.ok());
+
+ res_obj = nlohmann::json::parse(json_res);
+ ASSERT_EQ(3, res_obj["found"].get<size_t>());
+
+ ASSERT_EQ("3", res_obj["hits"][0]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][0]["document"].at("documentId"));
+
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("documentId"));
+
+ ASSERT_EQ("1", res_obj["hits"][2]["document"].at("leadId"));
+ ASSERT_EQ("1", res_obj["hits"][2]["document"].at("documentId"));
+
+ collectionManager.get_collection_unsafe("document")->remove("1");
+
+ req_params = {
+ {"collection", "lead_document"},
+ {"q", "*"},
+ };
+ search_op = collectionManager.do_search(req_params, embedded_params, json_res, now_ts);
+ ASSERT_TRUE(search_op.ok());
+
+ res_obj = nlohmann::json::parse(json_res);
+ ASSERT_EQ(2, res_obj["found"].get<size_t>());
+
+ ASSERT_EQ("3", res_obj["hits"][0]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][0]["document"].at("documentId"));
+
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("documentId"));
+
+ auto doc = R"({
+ "id": "1",
+ "leadId": "1",
+ "documentId": "3"
+ })"_json;
+ auto add_doc_op = collectionManager.get_collection_unsafe("lead_document")->add(doc.dump());
+ ASSERT_TRUE(add_doc_op.ok());
+
+ req_params = {
+ {"collection", "lead_document"},
+ {"q", "*"},
+ };
+ search_op = collectionManager.do_search(req_params, embedded_params, json_res, now_ts);
+ ASSERT_TRUE(search_op.ok());
+
+ res_obj = nlohmann::json::parse(json_res);
+ ASSERT_EQ(3, res_obj["found"].get<size_t>());
+
+ ASSERT_EQ("1", res_obj["hits"][0]["document"].at("leadId"));
+ ASSERT_EQ("3", res_obj["hits"][0]["document"].at("documentId"));
+
+ ASSERT_EQ("3", res_obj["hits"][1]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("documentId"));
+
+ ASSERT_EQ("2", res_obj["hits"][2]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][2]["document"].at("documentId"));
+
+ collectionManager.get_collection_unsafe("lead")->remove("1");
+
+ req_params = {
+ {"collection", "lead_document"},
+ {"q", "*"},
+ };
+ search_op = collectionManager.do_search(req_params, embedded_params, json_res, now_ts);
+ ASSERT_TRUE(search_op.ok());
+
+ res_obj = nlohmann::json::parse(json_res);
+ ASSERT_EQ(2, res_obj["found"].get<size_t>());
+
+ ASSERT_EQ("3", res_obj["hits"][0]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][0]["document"].at("documentId"));
+
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("leadId"));
+ ASSERT_EQ("2", res_obj["hits"][1]["document"].at("documentId"));
}
TEST_F(CollectionJoinTest, SortByReference) {