# ruff: noqa: E712,E731,F401,F403,F405,F541,F841,I001,UP031,UP032,W291,W292,W293 # fmt: off import random import threading import time import numpy as np import pandas as pd import pytest from base.client_v2_base import TestMilvusClientV2Base from common import common_func as cf from common import common_type as ct from common.common_type import CaseLabel, CheckTasks from pymilvus import DataType, Function, FunctionType from utils.util_log import test_log as log from utils.util_pymilvus import * # noqa: F403 prefix = "milvus_client_api_query" epsilon = ct.epsilon default_nb = ct.default_nb default_nb_medium = ct.default_nb_medium default_nq = ct.default_nq default_dim = ct.default_dim default_limit = ct.default_limit default_search_exp = "id >= 0" default_term_expr = f"{ct.default_int64_field_name} in [0, 1]" exp_res = "exp_res" default_search_string_exp = 'varchar >= "0"' default_search_mix_exp = 'int64 >= 0 && varchar >= "0"' default_invalid_string_exp = "varchar >= 0" default_json_search_exp = 'json_field["number"] >= 0' perfix_expr = 'varchar like "0%"' suffix_expr = 'varchar like "%0"' inner_match_expr = 'varchar like "%0%"' default_search_field = ct.default_float_vec_field_name default_search_params = ct.default_search_params default_primary_key_field_name = "id" default_vector_field_name = "vector" default_float_field_name = ct.default_float_field_name default_bool_field_name = ct.default_bool_field_name default_string_field_name = ct.default_string_field_name default_int32_array_field_name = ct.default_int32_array_field_name default_string_array_field_name = ct.default_string_array_field_name # ===================================================================== # Shared collection for TestMilvusClientQueryInvalidShared # Used by invalid-parameter query tests that do not mutate data/schema/state. # ===================================================================== INVALID_SHARED_COLLECTION = "test_query_invalid_shared_" + cf.gen_unique_str("_") INVALID_SHARED_NB = 2000 QUERY_COUNT_DEFAULT_SHARED_COLLECTION = "test_query_count_default_shared_" + cf.gen_unique_str("_") QUERY_COUNT_EXPR_SHARED_COLLECTION = "test_query_count_expr_shared_" + cf.gen_unique_str("_") QUERY_COUNT_EXPR_SHARED_ROWS = [] QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION = "test_query_string_non_primary_shared_" + cf.gen_unique_str("_") QUERY_STRING_NON_PRIMARY_SHARED_ROWS = [] QUERY_STRING_PRIMARY_SHARED_COLLECTION = "test_query_string_primary_shared_" + cf.gen_unique_str("_") QUERY_STRING_PRIMARY_SHARED_ROWS = [] def build_query_invalid_schema(client): """Schema for TestMilvusClientQueryInvalidShared. int64 PK (auto_id=False) + float_vec + float + varchar + json fields, enable_dynamic_field=True. enable_dynamic_field=True matches pymilvus quick-setup's default — unknown tokens in filter expressions (e.g. "s" in "12-s", "int64" when not a schema field) are treated as dynamic field references, so parser errors surface at the intended later stages ("predicate is not a boolean expression", "not can only apply on boolean", etc.) rather than bailing out at "field X not exist".""" schema = client.create_schema(auto_id=False, enable_dynamic_field=True) schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(default_float_field_name, DataType.FLOAT) schema.add_field(default_string_field_name, DataType.VARCHAR, max_length=256) schema.add_field(ct.default_json_field_name, DataType.JSON) return schema def build_query_count_default_schema(client): """Build the default schema used by read-only count(*) tests.""" schema = client.create_schema(auto_id=False, enable_dynamic_field=False) schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) return schema def build_query_count_expression_schema(client): """Build a superset schema for read-only count(*) expression tests.""" schema = client.create_schema(auto_id=False, enable_dynamic_field=False) schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=65535) schema.add_field(ct.default_bool_field_name, DataType.BOOL) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) return schema def build_query_string_non_primary_schema(client): """Build a schema with int64 primary key and varchar as a scalar field.""" schema = client.create_schema(auto_id=False, enable_dynamic_field=False) schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(default_string_field_name, DataType.VARCHAR, max_length=64) return schema def build_query_string_primary_schema(client): """Build a schema with varchar primary key plus numeric fields for string query tests.""" schema = client.create_schema(auto_id=False, enable_dynamic_field=False) schema.add_field(default_string_field_name, DataType.VARCHAR, max_length=100, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_int64_field_name, DataType.INT64) return schema def gen_query_string_primary_rows(nb=default_nb): """Generate deterministic rows where float <= int64 for compare-field tests.""" rows = [] vectors = cf.gen_vectors(nb, default_dim) for i in range(nb): rows.append( { default_string_field_name: str(i), default_vector_field_name: vectors[i], ct.default_float_field_name: float(i), ct.default_int64_field_name: i + 10, } ) return rows class TestMilvusClientQueryInvalid(TestMilvusClientV2Base): """Test case of search interface""" @pytest.fixture(scope="function", params=[False, True]) def auto_id(self, request): yield request.param @pytest.fixture(scope="function", params=["COSINE", "L2"]) def metric_type(self, request): yield request.param """ ****************************************************************** # The following are invalid base cases ****************************************************************** """ @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_non_constant_array_term(self): """ target: test query with non-constant array term expr method: query with non-constant array expr expected: raise exception Stays independent (not in TestMilvusClientQueryInvalidShared): the test expects a parse error for 'int64 in [[1]]', which requires enable_dynamic_field=False so the parser rejects the non-constant list element. With dynamic=True the server accepts the expression (int64 becomes a dynamic field) and just returns empty results. """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection( client, collection_name, default_dim, consistency_level="Strong", enable_dynamic_field=False ) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. test non-constant array term expressions constants = [[1], (), {}] for constant in constants: error = {ct.err_code: 1100, ct.err_msg: f"cannot parse expression: int64 in [{constant}]"} term_expr = f"{ct.default_int64_field_name} in [{constant}]" self.query(client, collection_name, filter=term_expr, check_task=CheckTasks.err_res, check_items=error) # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_not_existed_field(self, enable_dynamic_field): """ target: test query with not existed field method: query by term expr with fake field expected: raise exception Please note that when the query request contains non-existent output fields: - If enable_dynamic_field=True, the query will succeed but only return empty results. - If enable_dynamic_field=False, an error will be thrown: field not exist. """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection( client, collection_name, default_dim, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field ) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. query with non-existent field term_expr = "invalid_field in [1, 2]" if enable_dynamic_field: self.query( client, collection_name, filter=term_expr, check_task=CheckTasks.check_query_results, check_items={exp_res: [], "pk_name": ct.default_int64_field_name}, ) else: error = { ct.err_code: 65535, ct.err_msg: "field invalid_field not exist", } self.query(client, collection_name, filter=term_expr, check_task=CheckTasks.err_res, check_items=error) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_output_not_existed_field(self, enable_dynamic_field): """ target: test query output not existed field method: query with not existed output field expected: raise exception """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection( client, collection_name, default_dim, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field ) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. test query with non-existent fields output_fields = [["int"], [default_primary_key_field_name, "int"]] for output_field in output_fields: if enable_dynamic_field: exp_res = [{default_primary_key_field_name: i} for i in range(default_nb)] self.query( client, collection_name, filter=default_search_exp, output_fields=output_field, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "pk_name": ct.default_int64_field_name}, ) else: error = {ct.err_code: 65535, ct.err_msg: "field int not exist"} self.query( client, collection_name, filter=default_search_exp, output_fields=output_field, check_task=CheckTasks.err_res, check_items=error, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_json_contains_invalid_type(self, enable_dynamic_field): """ target: test query with expression using json_contains_any method: query with expression using json_contains_any expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data nb = 10 rows = cf.gen_row_data_by_schema(nb=nb, schema=schema) for i in range(nb): rows[i][ct.default_json_field_name] = {"number": i, "list": [m for m in range(i, i + 10)]} self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with invalid type expr_prefix_s = ["json_contains_any", "JSON_CONTAINS_ANY", "json_contains_all", "JSON_CONTAINS_ALL"] not_list_s = ["str", {1, 2, 3}, (1, 2, 3), 10] for expr_prefix in expr_prefix_s: for not_list in not_list_s: expression = f"{expr_prefix}({ct.default_json_field_name}['list'], {not_list})" error = { ct.err_code: 1100, ct.err_msg: f"failed to create query plan: cannot parse expression: {expression}", } self.query(client, collection_name, filter=expression, check_task=CheckTasks.err_res, check_items=error) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_partition_without_loading_collection(self): """ target: test query on partition without loading collection method: query on partition and no loading collection expected: raise exception """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) # 3. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_name=partition_name) # 4. verify partition has data self.flush(client, collection_name) partition_info = self.get_partition_stats(client, collection_name, partition_name)[0] assert partition_info["row_count"] == default_nb, ( f"Expected {default_nb} entities in partition, got {partition_info['row_count']}" ) # 5. query on partition without loading collection error = {ct.err_code: 65535, ct.err_msg: "collection not loaded"} self.query( client, collection_name, filter=default_term_expr, partition_names=[partition_name], check_task=CheckTasks.err_res, check_items=error, ) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_partition_without_loading_partiton(self): """ target: Verify that querying an unloaded partition raises an exception. method: 1. Create a collection and two partitions. 2. Insert data into both partitions. 3. Load only one partition. 4. Attempt to query the other partition that is not loaded. expected: The query should fail with an error indicating the partition is not loaded. """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name1 = cf.gen_unique_str("partition1") partition_name2 = cf.gen_unique_str("partition2") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name1) self.create_partition(client, collection_name, partition_name2) self.release_collection(client, collection_name) # 2. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_names=[partition_name1, partition_name2]) self.flush(client, collection_name) self.load_partitions(client, collection_name, partition_name1) # 3. query on partition without loading error = {ct.err_code: 65535, ct.err_msg: "partition not loaded"} self.query( client, collection_name, filter=default_search_exp, partition_names=[partition_name2], check_task=CheckTasks.err_res, check_items=error, ) # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.xdist_group("TestMilvusClientQueryInvalidShared") class TestMilvusClientQueryInvalidShared(TestMilvusClientV2Base): """Invalid-parameter query tests that share one static-schema collection. Every test here only verifies error paths for query arguments — none of them mutates the collection's data, schema, index, or load state, so a single module-scoped collection is safe to share across them.""" @pytest.fixture(scope="module", autouse=True) def prepare_invalid_shared_collection(self, request): client = self._client() collection_name = INVALID_SHARED_COLLECTION if client.has_collection(collection_name): client.drop_collection(collection_name) schema = build_query_invalid_schema(client) index_params = client.prepare_index_params() index_params.add_index( field_name=default_vector_field_name, index_type="HNSW", metric_type="L2", M=16, efConstruction=200 ) client.create_collection( collection_name=collection_name, schema=schema, index_params=index_params, consistency_level="Strong" ) rng = np.random.default_rng(seed=19530) rows = [] for i in range(INVALID_SHARED_NB): rows.append( { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: float(i), default_string_field_name: str(i), ct.default_json_field_name: {"number": i, "list": list(range(i, i + 10))}, } ) client.insert(collection_name=collection_name, data=rows) client.flush(collection_name=collection_name) def teardown(): try: if self.has_collection(client, INVALID_SHARED_COLLECTION)[0]: self.drop_collection(client, INVALID_SHARED_COLLECTION) except Exception: pass request.addfinalizer(teardown) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_not_all_required_params(self): """ target: test query without required limit parameter method: call query with no filter and no limit expected: raise error (empty filter requires explicit limit) """ client = self._client() error = {ct.err_code: 65535, ct.err_msg: "empty expression should be used with limit"} self.query(client, INVALID_SHARED_COLLECTION, check_task=CheckTasks.err_res, check_items=error) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_no_collection(self): """ target: test the scenario which query a non-existent collection method: query a collection name that was never created expected: raise collection-not-found error """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() error = {"err_code": 1, "err_msg": "collection not found"} self.query(client, collection_name, filter=default_search_exp, check_task=CheckTasks.err_res, check_items=error) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("filter", [None, ""]) def test_milvus_client_query_expr_none(self, filter): """ target: test query with None/empty filter and no limit method: 1. query with filter=None/"" without limit 2. query with filter=None/"" with offset but no limit expected: raise error in both cases (empty filter requires explicit limit) """ client = self._client() error = {ct.err_code: 0, ct.err_msg: "empty expression should be used with limit: invalid parameter"} self.query(client, INVALID_SHARED_COLLECTION, filter=filter, check_task=CheckTasks.err_res, check_items=error) self.query( client, INVALID_SHARED_COLLECTION, filter=filter, offset=1, check_task=CheckTasks.err_res, check_items=error ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("expr", ["12-s", "中文", "a"]) def test_milvus_client_query_expr_invalid_string(self, expr): """ target: test query with invalid filter expressions method: query with syntactically invalid filter strings expected: raise parse error """ client = self._client() if expr == "中文": error = {ct.err_code: 1100, ct.err_msg: "cannot parse expression"} elif expr == "a" or expr == "12-s": error = {ct.err_code: 1100, ct.err_msg: f"predicate is not a boolean expression: {expr}"} self.query(client, INVALID_SHARED_COLLECTION, filter=expr, check_task=CheckTasks.err_res, check_items=error) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_wrong_term_keyword(self): """ target: test query with wrong term expr keywords method: query with mistyped / invalid term keywords (e.g. "inn", "in not") expected: raise parse error """ client = self._client() expr_1 = f"{ct.default_int64_field_name} inn [1, 2]" error_1 = { ct.err_code: 65535, ct.err_msg: "invalid expression: int64 inn [1, 2]", } self.query(client, INVALID_SHARED_COLLECTION, filter=expr_1, check_task=CheckTasks.err_res, check_items=error_1) expr_2 = f"{ct.default_int64_field_name} in not [1, 2]" error_2 = { ct.err_code: 65535, ct.err_msg: "not can only apply on boolean", } self.query(client, INVALID_SHARED_COLLECTION, filter=expr_2, check_task=CheckTasks.err_res, check_items=error_2) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_non_array_term(self): """ target: test query with non-array RHS in term expr method: query with term expressions whose RHS is not a list expected: raise parse error """ client = self._client() exprs = [f"{ct.default_int64_field_name} in 1", f'{ct.default_int64_field_name} in "in"'] for expr in exprs: error = { ct.err_code: 1100, ct.err_msg: "the right-hand side of 'in' must be a list", } self.query(client, INVALID_SHARED_COLLECTION, filter=expr, check_task=CheckTasks.err_res, check_items=error) expr = f"{ct.default_int64_field_name} in (mn)" error = { ct.err_code: 1100, ct.err_msg: "value '(mn)' in list cannot be a non-const expression", } self.query(client, INVALID_SHARED_COLLECTION, filter=expr, check_task=CheckTasks.err_res, check_items=error) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_inconsistent_mix_term_array(self): """ target: test query with mismatched types in term expr on int64 PK method: 1. query with all-float values against int64 PK 2. query with mixed int and float values against int64 PK expected: raise casting error in both cases """ client = self._client() values = [1.0, 2.0] term_expr = f"{default_primary_key_field_name} in {values}" error = { ct.err_code: 1100, ct.err_msg: "value 'float_val:1' in list cannot be casted to Int64", } self.query( client, INVALID_SHARED_COLLECTION, filter=term_expr, check_task=CheckTasks.err_res, check_items=error ) values = [1, 2.0] term_expr = f"{default_primary_key_field_name} in {values}" error = { ct.err_code: 1100, ct.err_msg: "value 'float_val:2' in list cannot be casted to Int64", } self.query( client, INVALID_SHARED_COLLECTION, filter=term_expr, check_task=CheckTasks.err_res, check_items=error ) @pytest.mark.tags(CaseLabel.L0) def test_milvus_client_query_expr_with_limit_offset_out_of_range(self): """ target: test query with limit/offset outside the [1, 16384] window method: 1. limit > 16384 2. offset + limit > 16384 3. offset < 0 expected: raise invalid-window error in all cases """ client = self._client() error = { ct.err_code: 1, ct.err_msg: "invalid max query result window, (offset+limit) should be in range [1, 16384]", } self.query( client, INVALID_SHARED_COLLECTION, filter="", limit=16385, check_task=CheckTasks.err_res, check_items=error ) self.query( client, INVALID_SHARED_COLLECTION, filter="", limit=1, offset=16384, check_task=CheckTasks.err_res, check_items=error, ) self.query( client, INVALID_SHARED_COLLECTION, filter="", limit=16384, offset=1, check_task=CheckTasks.err_res, check_items=error, ) error = { ct.err_code: 1, ct.err_msg: "invalid max query result window, offset [-1] is invalid, should be gte than 0", } self.query( client, INVALID_SHARED_COLLECTION, filter="", limit=2, offset=-1, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("output_fields", [["*%"], ["**"], ["*", "@"]]) def test_milvus_client_query_invalid_wildcard(self, output_fields): """ target: test query with malformed wildcards in output_fields method: query with output_fields containing invalid wildcard patterns expected: raise parse error """ client = self._client() error = {ct.err_code: 65535, ct.err_msg: "parse output field name failed"} self.query( client, INVALID_SHARED_COLLECTION, filter=default_term_expr, output_fields=output_fields, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_not_existed_partition(self): """ target: test query against a non-existent partition name method: query with partition_names referencing a partition that was never created expected: raise partition-not-found error """ client = self._client() partition_name = cf.gen_unique_str() error = {ct.err_code: 65535, ct.err_msg: f"partition name {partition_name} not found"} self.query( client, INVALID_SHARED_COLLECTION, filter=default_term_expr, partition_names=[partition_name], check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("ignore_growing", [2.3, "str"]) def test_milvus_client_query_invalid_ignore_growing_param(self, ignore_growing): """ target: test query with invalid ignore_growing param type method: query with ignore_growing set to a non-bool value expected: raise parse error for ignore_growing field """ client = self._client() error = {ct.err_code: 999, ct.err_msg: "parse ignore growing field failed"} self.query( client, INVALID_SHARED_COLLECTION, filter=default_search_exp, ignore_growing=ignore_growing, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("limit", ["12 s", " ", [0, 1], {2}]) def test_milvus_client_query_pagination_with_invalid_limit_type(self, limit): """ target: test query pagination with invalid limit type method: query with limit set to non-integer types expected: raise invalid-limit error """ client = self._client() # milvus strips leading/trailing whitespace before echoing the value # in the error message, so " " becomes "" in `limit [] is invalid` displayed = limit.strip() if isinstance(limit, str) else limit error = {ct.err_code: 1, ct.err_msg: f"limit [{displayed}] is invalid"} self.query( client, INVALID_SHARED_COLLECTION, filter=default_search_exp, offset=10, limit=limit, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("limit", [-1, 67890]) def test_milvus_client_query_pagination_with_invalid_limit_value(self, limit): """ target: test query pagination with out-of-range limit value method: query with negative limit or offset+limit exceeding 16384 expected: raise invalid-window error """ client = self._client() error = { ct.err_code: 65535, ct.err_msg: "invalid max query result window, (offset+limit) should be in range [1, 16384], but got 67900", } if limit == -1: error = { ct.err_code: 65535, ct.err_msg: f"invalid max query result window, limit [{limit}] is invalid, should be greater than 0", } self.query( client, INVALID_SHARED_COLLECTION, filter=default_search_exp, offset=10, limit=limit, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("offset", ["12 s", " ", [0, 1], {2}]) def test_milvus_client_query_pagination_with_invalid_offset_type(self, offset): """ target: test query pagination with invalid offset type method: query with offset set to non-integer types expected: raise invalid-offset error """ client = self._client() # milvus strips leading/trailing whitespace before echoing the value # in the error message, so " " becomes "" in `offset [] is invalid` displayed = offset.strip() if isinstance(offset, str) else offset error = {ct.err_code: 1, ct.err_msg: f"offset [{displayed}] is invalid"} self.query( client, INVALID_SHARED_COLLECTION, filter=default_search_exp, offset=offset, limit=10, check_task=CheckTasks.err_res, check_items=error, ) class TestMilvusClientQueryValid(TestMilvusClientV2Base): """Test case of search interface""" @pytest.fixture(scope="function", params=[False, True]) def auto_id(self, request): yield request.param @pytest.fixture(scope="function", params=["COSINE", "L2"]) def metric_type(self, request): yield request.param @pytest.fixture(scope="function", params=[True, False]) def enable_dynamic_field(self, request): yield request.param """ ****************************************************************** # The following are valid base cases ****************************************************************** """ @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_default(self): """ target: test query (high level api) normal case method: create connection, collection, insert and search expected: search/query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) # 3. query using ids self.query( client, collection_name, ids=[i for i in range(default_nb)], check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 4. query using filter self.query( client, collection_name, filter=default_search_exp, check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_with_multiple_namespaces(self): """ target: test query with namespace enabled method: create namespace-enabled collection with multiple shards, insert rows into multiple namespaces, then query by namespace expected: query only returns rows from the requested namespace """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() shards_num = 4 self.create_collection( client, collection_name, default_dim, consistency_level="Strong", enable_namespace=True, shards_num=shards_num, ) desc = self.describe_collection(client, collection_name)[0] assert desc["enable_namespace"] is True assert desc["num_shards"] == shards_num rng = np.random.default_rng(seed=19530) namespaces = [f"namespace_{i}" for i in range(8)] nb_per_namespace = 10 ids_by_namespace = { namespace: [idx * 1000 + i for i in range(nb_per_namespace)] for idx, namespace in enumerate(namespaces) } for namespace, ids in ids_by_namespace.items(): rows = [ { default_primary_key_field_name: pk, default_vector_field_name: list(rng.random((1, default_dim))[0]), } for pk in ids ] self.insert(client, collection_name, rows, namespace=namespace) for namespace, expected_ids in ids_by_namespace.items(): res = self.query( client, collection_name, filter=default_search_exp, output_fields=[default_primary_key_field_name], namespace=namespace, )[0] assert {r[default_primary_key_field_name] for r in res} == set(expected_ids) first_id_per_namespace = [ids[0] for ids in ids_by_namespace.values()] for namespace, expected_ids in ids_by_namespace.items(): filtered = self.query( client, collection_name, filter=f"{default_primary_key_field_name} in {first_id_per_namespace}", output_fields=[default_primary_key_field_name], namespace=namespace, )[0] assert {r[default_primary_key_field_name] for r in filtered} == {expected_ids[0]} self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_output_fields(self): """ target: test query (high level api) normal case method: create connection, collection, insert and search expected: search/query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) # 3. query using ids self.query( client, collection_name, ids=[i for i in range(default_nb)], check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 4. query using filter res = self.query( client, collection_name, filter=default_search_exp, output_fields=[ default_primary_key_field_name, default_float_field_name, default_string_field_name, default_vector_field_name, ], check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] assert set(res[0].keys()) == { default_primary_key_field_name, default_vector_field_name, default_float_field_name, default_string_field_name, } self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_output_fields_dynamic_name(self): """ target: test query (high level api) normal case method: create connection, collection, insert, add field name(same as dynamic name) and query expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() dim = 8 # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, max_length=64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=dim) schema.add_field(default_float_field_name, DataType.FLOAT, nullable=True) index_params = self.prepare_index_params(client)[0] index_params.add_index(default_vector_field_name, metric_type="COSINE") self.create_collection(client, collection_name, dimension=dim, schema=schema, index_params=index_params) # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) self.add_collection_field( client, collection_name, field_name=default_string_field_name, data_type=DataType.VARCHAR, nullable=True, default_value="default", max_length=64, ) for row in rows: row[default_string_field_name] = "default" # 3. query using ids self.query( client, collection_name, ids=[i for i in range(default_nb)], check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 4. query using filter res = self.query( client, collection_name, filter=default_search_exp, output_fields=[f'$meta["{default_string_field_name}"]'], check_task=CheckTasks.check_query_results, check_items={ exp_res: [{"id": item["id"]} for item in rows], "with_vec": True, "pk_name": default_primary_key_field_name, }, )[0] assert set(res[0].keys()) == {default_primary_key_field_name} self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_output_fields_nullable_vector_fields(self): """ target: test query with nullable vector fields method: create collection with nullable vector fields, insert data with nullable vector fields, query with nullable vector fields expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim, nullable=True) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 3. query with nullable vector fields res = self.query( client, collection_name, filter=default_search_exp, output_fields=[default_primary_key_field_name, default_vector_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] assert set(res[0].keys()) == {default_primary_key_field_name, default_vector_field_name} self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) @pytest.mark.parametrize( "vector_field, vector_type", [ (ct.default_float_vec_field_name, DataType.FLOAT_VECTOR), (ct.default_float16_vec_field_name, DataType.FLOAT16_VECTOR), (ct.default_bfloat16_vec_field_name, DataType.BFLOAT16_VECTOR), ], ) def test_milvus_client_query_output_fields_all(self, enable_dynamic_field, vector_field, vector_type): """ target: test query (high level api) normal case method: create connection, collection, insert and search expected: search/query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_int32_field_name, DataType.INT32) schema.add_field(ct.default_int16_field_name, DataType.INT16) schema.add_field(ct.default_int8_field_name, DataType.INT8) schema.add_field(ct.default_bool_field_name, DataType.BOOL) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_double_field_name, DataType.DOUBLE) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=10) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(vector_field, vector_type, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert rows = cf.gen_row_data_by_schema(nb=10, schema=schema, skip_field_names=[ct.default_json_field_name]) for i in range(10): rows[i][ct.default_json_field_name] = {"key": i} self.insert(client, collection_name, rows) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=vector_field, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) all_fields = [ default_primary_key_field_name, ct.default_int32_field_name, ct.default_int16_field_name, ct.default_int8_field_name, ct.default_bool_field_name, ct.default_float_field_name, ct.default_double_field_name, ct.default_string_field_name, ct.default_json_field_name, vector_field, ] # 3. query using ids self.query( client, collection_name, ids=[i for i in range(default_nb)], output_fields=["*"], check_task=CheckTasks.check_query_results, check_items={ exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name, "vector_field": vector_field, "vector_type": vector_type, }, ) # 4. query using filter res = self.query( client, collection_name, filter=default_search_exp, output_fields=["*"], check_task=CheckTasks.check_query_results, check_items={ exp_res: rows, "with_vec": True, "pk_name": default_primary_key_field_name, "vector_field": vector_field, "vector_type": vector_type, }, )[0] assert set(res[0].keys()) == set(all_fields) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_output_field_none_or_empty(self, enable_dynamic_field): """ target: test query with none and empty output field method: query with output field=None, field=[] expected: return primary field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection( client, collection_name, default_dim, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field ) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with None and empty output fields fields = [None, [], ""] for field in fields: res = self.query(client, collection_name, filter=default_search_exp, output_fields=field)[0] assert res[0].keys() == {default_primary_key_field_name, default_vector_field_name} # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L0) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_output_one_field(self, enable_dynamic_field): """ target: test query with output one field method: query with output one field expected: return one field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection( client, collection_name, schema=schema, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field, ) rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with one output field res = self.query(client, collection_name, filter=default_search_exp, output_fields=[default_float_field_name])[ 0 ] # verify primary field and specified field are returned assert set(res[0].keys()) == {default_primary_key_field_name, default_float_field_name} # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_limit(self): """ target: test query (high level api) normal case method: create connection, collection, insert and search expected: search/query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) # 3. query using ids limit = 5 self.query( client, collection_name, ids=[i for i in range(default_nb)], limit=limit, check_task=CheckTasks.check_query_results, check_items={exp_res: rows[:limit], "with_vec": True, "pk_name": default_primary_key_field_name[:limit]}, ) # 4. query using filter self.query( client, collection_name, filter=default_search_exp, limit=limit, check_task=CheckTasks.check_query_results, check_items={exp_res: rows[:limit], "with_vec": True, "pk_name": default_primary_key_field_name[:limit]}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("sample_rate", [0.7, 0.5, 0.01]) def test_milvus_client_query_random_sample(self, sample_rate): """ target: test query random sample method: create connection, collection, insert and query expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert vectors = cf.gen_vectors(default_nb, default_dim) rows = [ { default_primary_key_field_name: i, default_vector_field_name: vectors[i], default_float_field_name: i * 1.0, default_string_field_name: cf.generate_random_sentence("English"), } for i in range(default_nb) ] self.insert(client, collection_name, rows) expr = f"{default_string_field_name} like '%red%'" # 3. query without sample rate all_res = self.query(client, collection_name, filter=expr)[0] exp_num = max(1, int(len(all_res) * sample_rate)) # 4. query using sample rate expr = expr + f" && random_sample({sample_rate})" sample_res = self.query(client, collection_name, filter=expr)[0] log.info(exp_num) assert len(sample_res) == exp_num @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("sample_rate", [1, 0, -9]) def test_milvus_client_query_invalid_sample_rate(self, sample_rate): """ target: test query random sample method: create connection, collection, insert and query expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert vectors = cf.gen_vectors(1, default_dim) rows = [ { default_primary_key_field_name: i, default_vector_field_name: vectors[i], default_float_field_name: i * 1.0, default_string_field_name: cf.generate_random_sentence("English"), } for i in range(1) ] self.insert(client, collection_name, rows) expr = f"{default_string_field_name} like '%red%' && random_sample({sample_rate})" # 3. query error = { ct.err_code: 999, ct.err_msg: "the sample factor should be between 0 and 1 and not too close to 0 or 1", } self.query(client, collection_name, filter=expr, check_task=CheckTasks.err_res, check_items=error) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_json_modulo_operator(self): """ target: test query with modulo operator on JSON field values method: create collection with JSON field, insert data with numeric values, test modulo filters expected: modulo operations should work correctly and return all matching results """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with JSON field schema, _ = self.create_schema(client, enable_dynamic_field=False) schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field("json_field", DataType.JSON, nullable=True) index_params, _ = self.prepare_index_params(client) index_params.add_index(default_vector_field_name, metric_type="COSINE") self.create_collection( client, collection_name, schema=schema, index_params=index_params, consistency_level="Strong" ) # 2. insert 3000 rows with various numeric values nb = 3000 rng = np.random.default_rng(seed=19530) # Store original data for verification rows = [] all_numbers = [] all_nested_numbers = [] for i in range(nb): # Generate diverse numeric values if i % 5 == 0: # Some large numbers numeric_value = random.randint(100000, 999999) elif i % 3 == 0: # Some medium numbers numeric_value = random.randint(1000, 9999) else: # Regular sequential numbers numeric_value = i nested_value = i * 7 + (i % 13) # Different pattern for nested all_numbers.append(numeric_value) all_nested_numbers.append(nested_value) rows.append( { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), "json_field": {"number": numeric_value, "index": i, "data": {"nested_number": nested_value}}, } ) self.insert(client, collection_name, rows) # 3. Test modulo operations with different modulo values and remainders test_cases = [ (10, list(range(10))), # mod 10 - all remainders 0-9 (2, [0, 1]), # mod 2 - even/odd (3, [0, 1, 2]), # mod 3 (7, list(range(7))), # mod 7 - all remainders 0-6 (13, [0, 5, 12]), # mod 13 - selected remainders ] for modulo, remainders in test_cases: log.info(f"\nTesting modulo {modulo}") for remainder in remainders: # Calculate expected results from original data expected_ids = [i for i, num in enumerate(all_numbers) if num % modulo == remainder] expected_count = len(expected_ids) # Query and get results filter_expr = f'json_field["number"] % {modulo} == {remainder}' res, _ = self.query( client, collection_name, filter=filter_expr, output_fields=["json_field", default_primary_key_field_name], ) # Extract actual IDs from results actual_ids = sorted([item[default_primary_key_field_name] for item in res]) expected_ids_sorted = sorted(expected_ids) log.info(f"Modulo {modulo} remainder {remainder}: expected {expected_count}, got {len(res)} results") # Verify we got exactly the expected results assert len(res) == expected_count, ( f"Expected {expected_count} results for % {modulo} == {remainder}, got {len(res)}" ) assert actual_ids == expected_ids_sorted, ( f"Result IDs don't match expected IDs for % {modulo} == {remainder}" ) # Also verify each result is correct for item in res: actual_remainder = item["json_field"]["number"] % modulo assert actual_remainder == remainder, ( f"Number {item['json_field']['number']} % {modulo} = {actual_remainder}, expected {remainder}" ) # Test nested field expected_nested_ids = [i for i, num in enumerate(all_nested_numbers) if num % modulo == remainder] nested_filter = f'json_field["data"]["nested_number"] % {modulo} == {remainder}' nested_res, _ = self.query( client, collection_name, filter=nested_filter, output_fields=["json_field", default_primary_key_field_name], ) actual_nested_ids = sorted([item[default_primary_key_field_name] for item in nested_res]) expected_nested_ids_sorted = sorted(expected_nested_ids) assert len(nested_res) == len(expected_nested_ids), ( f"Expected {len(expected_nested_ids)} nested results for % {modulo} == {remainder}, got {len(nested_res)}" ) assert actual_nested_ids == expected_nested_ids_sorted, ( f"Nested result IDs don't match expected IDs for % {modulo} == {remainder}" ) # Test combining modulo with other conditions combined_expr = 'json_field["number"] % 2 == 0 && json_field["index"] < 100' expected_combined = [i for i in range(min(100, nb)) if all_numbers[i] % 2 == 0] res_combined, _ = self.query( client, collection_name, filter=combined_expr, output_fields=["json_field", default_primary_key_field_name] ) actual_combined_ids = sorted([item[default_primary_key_field_name] for item in res_combined]) assert len(res_combined) == len(expected_combined), ( f"Expected {len(expected_combined)} results for combined filter, got {len(res_combined)}" ) assert actual_combined_ids == sorted(expected_combined), "Results for combined filter don't match expected" log.info(f"All modulo tests passed! Combined filter returned {len(res_combined)} results") self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_empty_collection(self): """ target: test query empty collection method: query on an empty collection expected: empty result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, "vector") index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name="vector", index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) res = self.query(client, collection_name, filter=default_search_exp)[0] assert len(res) == 0 @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_empty_term_array(self): """ target: test query with empty array term expr method: query with empty term expr expected: empty result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. query with empty array term_expr = f"{ct.default_int64_field_name} in []" res = self.query(client, collection_name, filter=term_expr)[0] assert len(res) == 0 # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("dup_times", [1, 2, 3]) @pytest.mark.parametrize("dim", [8, 128]) def test_milvus_client_query_with_dup_primary_key(self, dim, dup_times): """ target: test query with duplicate primary key method: 1.insert same data twice 2.search expected: query results are de-duplicated """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, dim, consistency_level="Strong") # 2. insert duplicate data multiple times schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) for _ in range(dup_times): self.insert(client, collection_name, rows) # 3. query res = self.query(client, collection_name, filter=default_search_exp)[0] # assert that query results are de-duplicated res_ids = [m[default_primary_key_field_name] for m in res] assert sorted(list(set(res_ids))) == sorted(res_ids) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_auto_id_not_existed_primary_values(self): """ target: test query on auto_id true collection method: 1.create auto_id true collection 2.query with not existed primary keys expected: query result is empty """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with auto_id=True self.create_collection(client, collection_name, default_dim, auto_id=True, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, "vector") index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name="vector", index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. query with non-existent primary keys term_expr = f"{default_primary_key_field_name} in [0, 1, 2]" res = self.query(client, collection_name, filter=term_expr)[0] assert len(res) == 0 self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_non_primary_fields(self): """ target: test query on non-primary non-vector fields method: query on non-primary non-vector fields expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with all field types schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_int32_field_name, DataType.INT32) schema.add_field(ct.default_int16_field_name, DataType.INT16) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_double_field_name, DataType.DOUBLE) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query by non-primary non-vector scalar fields non_primary_fields = [ ct.default_int32_field_name, ct.default_int16_field_name, ct.default_float_field_name, ct.default_double_field_name, ct.default_string_field_name, ] # exp res: first two rows and all fields except vector field exp_res = rows[:2] for field in non_primary_fields: if field in schema.fields: filter_values = [rows[0][field], rows[1][field]] if field != ct.default_string_field_name: term_expr = f"{field} in {filter_values}" else: term_expr = f"{field} in {filter_values}" term_expr = term_expr.replace("'", '"') self.query( client, collection_name, filter=term_expr, output_fields=["*"], check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "with_vec": True, "pk_name": ct.default_int64_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_by_bool_field(self): """ target: test query by bool field and output bool field method: 1.create and insert with [int64, float, bool, float_vec] fields 2.query by bool field, and output all int64, bool fields expected: verify query result and output fields """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with bool field schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_bool_field_name, DataType.BOOL) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with bool values rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) # Update bool field values to alternate between True and False for i, row in enumerate(rows): row[ct.default_bool_field_name] = True if i % 2 == 0 else False self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. output bool field res = self.query(client, collection_name, filter=default_term_expr, output_fields=[ct.default_bool_field_name])[ 0 ] assert set(res[0].keys()) == {ct.default_int64_field_name, ct.default_bool_field_name} # 5. not support filter bool field with expr 'bool in [0/ 1]' not_support_expr = f"{ct.default_bool_field_name} in [0]" error = { ct.err_code: 65535, ct.err_msg: "value 'int64_val:0' in list cannot be casted to Bool", } self.query( client, collection_name, filter=not_support_expr, output_fields=[ct.default_bool_field_name], check_task=CheckTasks.err_res, check_items=error, ) # 6. filter bool field by bool term expr for bool_value in [True, False]: exprs = [f"{ct.default_bool_field_name} in [{bool_value}]", f"{ct.default_bool_field_name} == {bool_value}"] for expr in exprs: res = self.query(client, collection_name, filter=expr, output_fields=[ct.default_bool_field_name])[0] assert len(res) == default_nb // 2 for _r in res: assert _r[ct.default_bool_field_name] == bool_value self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_by_int64(self): """ target: test query through int64 field and output int64 field method: use int64 as query expr parameter expected: verify query output number """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection and index self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, "vector") index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb * 10, schema=schema_info) self.insert(client, collection_name, rows) assert len(rows) == default_nb * 10 # 3. filter on int64 fields expr_list = [ f"{default_primary_key_field_name} > 8192 && {default_primary_key_field_name} < 8194", f"{default_primary_key_field_name} > 16384 && {default_primary_key_field_name} < 16386", ] for expr in expr_list: res = self.query(client, collection_name, filter=expr, output_fields=[ct.default_int64_field_name])[0] assert len(res) == 1 # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_by_int8(self): """ target: test query through int8 field and output all scalar fields method: 1. create collection with fields [int64, float, int8, float_vec] 2. insert data 3. query by int8 field expected: query result matches expected rows """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection and index schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_int8_field_name, DataType.INT8) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) # Cycle through the full INT8 range without relying on numpy overflow casting. for i, row in enumerate(rows): row[ct.default_int8_field_name] = ((i + 128) % 256) - 128 self.insert(client, collection_name, rows) assert len(rows) == default_nb # 3. query expression term_expr = f"{ct.default_int8_field_name} in [0]" # expected query result: # int8 range [-128, 127], so repeated values every 256 expected_rows = [] for i in range(0, default_nb, 256): expected_rows.append(rows[i]) res = self.query( client, collection_name, filter=term_expr, output_fields=["float", "int64", "int8", "varchar"] )[0] assert len(res) == len(expected_rows) returned_ids = {r[ct.default_int64_field_name] for r in res} expected_ids = {r[ct.default_int64_field_name] for r in expected_rows} assert returned_ids == expected_ids # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_expr_not_in_term(self): """ target: test query with `not in` expr method: query with not in expr expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # query with not in expression fields = [ct.default_int64_field_name, ct.default_float_field_name] for field in fields: values = [int(row[field]) if field == ct.default_int64_field_name else float(row[field]) for row in rows] pos = 100 term_expr = f"{field} not in {values[pos:]}" df = pd.DataFrame(rows) res = df.iloc[:pos, :3].to_dict("records") self.query( client, collection_name, filter=term_expr, output_fields=[ct.default_float_field_name, ct.default_int64_field_name, ct.default_string_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_expr_not_in_empty_and_all(self): """ target: test query with `not in` expr method: query with `not in` expr for (non)empty collection expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with not in expression int64_values = [row[ct.default_int64_field_name] for row in rows] pos_s = [0, default_nb] for pos in pos_s: term_expr = f"{ct.default_int64_field_name} not in {int64_values[pos:]}" df = pd.DataFrame(rows) res = df.iloc[:pos, :1].to_dict("records") self.query( client, collection_name, filter=term_expr, output_fields=[ct.default_int64_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_expr_random_values(self): """ target: test query with random filter values method: query with random filter values, like [0, 2, 4, 3] expected: correct query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=100, schema=schema_info) self.insert(client, collection_name, rows) assert len(rows) == 100 # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with random values random_values = [0, 2, 4, 3] term_expr = f"{ct.default_int64_field_name} in {random_values}" df = pd.DataFrame(rows) res = df.iloc[random_values, :1].to_dict("records") self.query( client, collection_name, filter=term_expr, output_fields=[ct.default_int64_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_not_in_random(self): """ target: test query with fixed filter values method: query with fixed filter values expected: correct query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=50, schema=schema_info) self.insert(client, collection_name, rows) assert len(rows) == 50 # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with not in random values random_values = [i for i in range(10, 50)] log.debug(f"random values: {random_values}") random.shuffle(random_values) term_expr = f"{ct.default_int64_field_name} not in {random_values}" df = pd.DataFrame(rows) res = df.iloc[:10, :1].to_dict("records") self.query( client, collection_name, filter=term_expr, output_fields=[ct.default_int64_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) # @pytest.mark.parametrize("expr_prefix", ["json_contains", "JSON_CONTAINS", # "array_contains", "ARRAY_CONTAINS"]) def test_milvus_client_query_expr_json_contains(self, enable_dynamic_field): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) limit = 99 for i in range(default_nb): rows[i][ct.default_json_field_name] = {"number": i, "list": [m for m in range(i, i + limit)]} self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query expr_prefix_s = ["json_contains", "JSON_CONTAINS", "array_contains", "ARRAY_CONTAINS"] for expr_prefix in expr_prefix_s: expression = f"{expr_prefix}({ct.default_json_field_name}['list'], 1000)" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_expr_list_json_contains(self): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data limit = default_nb // 4 rows = [] for i in range(default_nb): data = { ct.default_int64_field_name: i, ct.default_json_field_name: [str(m) for m in range(i, i + limit)], ct.default_vector_field_name: cf.gen_vectors(1, default_dim)[0], } rows.append(data) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query expr_prefix_s = ["json_contains", "JSON_CONTAINS"] for expr_prefix in expr_prefix_s: expression = f"{expr_prefix}({ct.default_json_field_name}, '1000')" res = self.query(client, collection_name, filter=expression, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == limit # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_json_contains_combined_with_normal(self, enable_dynamic_field): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) limit = default_nb // 3 tar = 1000 for i in range(default_nb): # Set float field values to ensure some records satisfy the query condition rows[i][ct.default_json_field_name] = {"number": i, "list": [m for m in range(i, i + limit)]} if i % 2 == 0: rows[i][ct.default_float_field_name] = tar - limit else: rows[i][ct.default_float_field_name] = tar self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with combined expression expr_prefix_s = ["json_contains", "JSON_CONTAINS"] for expr_prefix in expr_prefix_s: expression = f"{expr_prefix}({ct.default_json_field_name}['list'], {tar}) && {ct.default_float_field_name} > {tar - limit // 2}" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit // 2 # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_all_datatype_json_contains_all(self, enable_dynamic_field): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) limit = 10 for i in range(default_nb): content = { "listInt": [m for m in range(i, i + limit)], "listStr": [str(m) for m in range(i, i + limit)], "listFlt": [m * 1.0 for m in range(i, i + limit)], "listBool": [bool(i % 2)], "listList": [[i, str(i + 1)], [i * 1.0, i + 1]], "listMix": [i, i * 1.111, str(i), bool(i % 2), [i, str(i)]], } rows[i][ct.default_json_field_name] = content self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with different data types _id = random.randint(limit, default_nb - limit) expr_prefix_s = ["json_contains_all", "JSON_CONTAINS_ALL", "array_contains_all", "ARRAY_CONTAINS_ALL"] for expr_prefix in expr_prefix_s: # test for int ids = [i for i in range(_id, _id + limit)] expression = f"{expr_prefix}({ct.default_json_field_name}['listInt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # test for string ids = [str(_id), str(_id + 1), str(_id + 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listStr'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit - len(ids) + 1 # test for float ids = [_id * 1.0] expression = f"{expr_prefix}({ct.default_json_field_name}['listFlt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit # test for bool ids = [True] expression = f"{expr_prefix}({ct.default_json_field_name}['listBool'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == default_nb // 2 # test for list ids = [[_id, str(_id + 1)]] expression = f"{expr_prefix}({ct.default_json_field_name}['listList'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # test for mixed data ids = [[_id, str(_id)], bool(_id % 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listMix'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("expr_prefix", ["json_contains_all", "JSON_CONTAINS_ALL"]) def test_milvus_client_query_expr_list_all_datatype_json_contains_all(self, expr_prefix): """ target: test query with expression using json_contains_all method: query with expression using json_contains_all expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema, skip_field_names=[ct.default_json_field_name]) limit = 10 for i in range(default_nb): rows[i][ct.default_json_field_name] = { "listInt": [m for m in range(i, i + limit)], "listStr": [str(m) for m in range(i, i + limit)], "listFlt": [m * 1.0 for m in range(i, i + limit)], "listBool": [bool(i % 2)], "listList": [[i, str(i + 1)], [i * 1.0, i + 1]], "listMix": [i, i * 1.1, str(i), bool(i % 2), [i, str(i)]], } self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with different data types _id = random.randint(limit, default_nb - limit) # test for int ids = [i for i in range(_id, _id + limit)] expression = f"{expr_prefix}({ct.default_json_field_name}['listInt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # test for string ids = [str(_id), str(_id + 1), str(_id + 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listStr'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit - len(ids) + 1 # test for float ids = [_id * 1.0] expression = f"{expr_prefix}({ct.default_json_field_name}['listFlt'], {ids})" res = self.query(client, collection_name, filter=expression, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == limit # test for bool ids = [True] expression = f"{expr_prefix}({ct.default_json_field_name}['listBool'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == default_nb // 2 # test for list ids = [[_id, str(_id + 1)]] expression = f"{expr_prefix}({ct.default_json_field_name}['listList'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # test for mixed data ids = [[_id, str(_id)], bool(_id % 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listMix'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_all_datatype_json_contains_any(self, enable_dynamic_field): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) limit = 10 for i in range(default_nb): rows[i][ct.default_json_field_name] = { "listInt": [m for m in range(i, i + limit)], "listStr": [str(m) for m in range(i, i + limit)], "listFlt": [m * 1.0 for m in range(i, i + limit)], "listBool": [bool(i % 2)], "listList": [[i, str(i + 1)], [i * 1.0, i + 1]], "listMix": [i, i * 1.1, str(i), bool(i % 2), [i, str(i)]], } self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with different data types _id = random.randint(limit, default_nb - limit) expr_prefix_s = ["json_contains_any", "JSON_CONTAINS_ANY"] for expr_prefix in expr_prefix_s: # test for int ids = [i for i in range(_id, _id + limit)] expression = f"{expr_prefix}({ct.default_json_field_name}['listInt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 2 * limit - 1 # test for string ids = [str(_id), str(_id + 1), str(_id + 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listStr'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit + len(ids) - 1 # test for float ids = [_id * 1.0] expression = f"{expr_prefix}({ct.default_json_field_name}['listFlt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == limit # test for bool ids = [True] expression = f"{expr_prefix}({ct.default_json_field_name}['listBool'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == default_nb // 2 # test for list ids = [[_id, str(_id + 1)]] expression = f"{expr_prefix}({ct.default_json_field_name}['listList'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 # test for mixed data ids = [_id, bool(_id % 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listMix'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == default_nb // 2 # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L0) @pytest.mark.parametrize("auto_id", [True, False]) def test_milvus_client_query_expr_empty(self, auto_id): """ target: test query with empty expression method: query empty expression with a limit expected: return topK results by order """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=auto_id) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with empty expression and limit limits = [10, 100, 1000] for limit in limits: res = self.query(client, collection_name, filter="", limit=limit)[0] # 5. verify results are ordered by primary key if not auto_id: primary_keys = [entity[default_primary_key_field_name] for entity in res] assert primary_keys == sorted(primary_keys) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_empty_pk_string(self): """ target: test query with empty expression method: query empty expression with a limit expected: return topK results by order """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with string primary key schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=10, is_primary=True) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with limit - string field is sorted by lexicographical order exp_ids = ["0", "1", "10", "100", "1000", "1001", "1002", "1003", "1004", "1005"] res = self.query(client, collection_name, filter="", limit=ct.default_limit)[0] # verify results are in lexicographical order primary_keys = [entity[ct.default_string_field_name] for entity in res] assert primary_keys == exp_ids # 5. query with limit + offset res = self.query(client, collection_name, filter="", limit=5, offset=5)[0] primary_keys = [entity[ct.default_string_field_name] for entity in res] assert primary_keys == exp_ids[5:] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("auto_id", [True, False]) def test_milvus_client_query_expr_empty_with_pagination(self, auto_id): """ target: test query with empty expression method: query empty expression with a limit expected: return topK results by order """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=auto_id) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with limit and offset offsets = [100, 1000] limits = [100, 1000] for offset in offsets: for limit in limits: res = self.query(client, collection_name, filter="", limit=limit, offset=offset)[0] # 5. verify results are ordered by primary key if not auto_id: primary_keys = [entity[default_primary_key_field_name] for entity in res] expected_ids = list(range(offset, offset + limit)) assert primary_keys == expected_ids # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_expr_empty_with_random_pk(self): """ target: test query with empty expression method: create a collection using random pk, query empty expression with a limit expected: return topK results by order """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=10) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. generate unordered pk array and insert unordered_ids = [i for i in range(default_nb)] random.shuffle(unordered_ids) rows = [] for i in range(default_nb): row = { ct.default_int64_field_name: unordered_ids[i], ct.default_float_field_name: np.float32(unordered_ids[i]), ct.default_string_field_name: str(unordered_ids[i]), ct.default_vector_field_name: cf.gen_vectors(nb=1, dim=default_dim)[0], } rows.append(row) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with empty expr and check the result limits = [100, 1000] offsets = [100, 1000] for limit in limits: for offset in offsets: exp_ids, res = sorted(unordered_ids)[:limit], [] for ids in exp_ids: res.append({ct.default_int64_field_name: ids, ct.default_string_field_name: str(ids)}) res = self.query( client, collection_name, filter="", limit=limit, output_fields=[ct.default_string_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, )[0] # 5. query with pagination exp_ids, res = sorted(unordered_ids)[: limit + offset][offset:], [] for ids in exp_ids: res.append({ct.default_int64_field_name: ids, ct.default_string_field_name: str(ids)}) res = self.query( client, collection_name, filter="", limit=limit, offset=offset, output_fields=[ct.default_string_field_name], check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, )[0] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("expression", cf.gen_integer_overflow_expressions()) def test_milvus_client_query_expr_out_of_range(self, expression): """ target: test query with integer overflow in boolean expressions method: create a collection with various integer fields, insert data, execute query using expressions with out-of-range integer constants, and compare the number of results with local Python evaluation. expected: query executes successfully and the result count matches local evaluation """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with all data types schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_int32_field_name, DataType.INT32) schema.add_field(ct.default_int16_field_name, DataType.INT16) schema.add_field(ct.default_int8_field_name, DataType.INT8) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert valid bounded integer values start = default_nb // 2 rows = [] for i in range(default_nb): row = { ct.default_int64_field_name: start + i, ct.default_int32_field_name: start + i, ct.default_int16_field_name: (start + i) % 32768, ct.default_int8_field_name: (start + i) % 128, ct.default_float_vec_field_name: cf.gen_vectors(nb=1, dim=default_dim)[0], } rows.append(row) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. filter result with expression in collection expression = expression.replace("&&", "and").replace("||", "or") filter_ids = [] for i in range(default_nb): local_values = { "int8": (start + i) % 128, "int16": (start + i) % 32768, "int32": start + i, } if not expression or eval(expression, {}, local_values): filter_ids.append(start + i) # 5. query and verify result res = self.query(client, collection_name, filter=expression, output_fields=["int8"])[0] assert len(res) == len(filter_ids) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize( "expr_prefix", ["json_contains_any", "JSON_CONTAINS_ANY", "array_contains_any", "ARRAY_CONTAINS_ANY"] ) def test_milvus_client_query_expr_list_all_datatype_json_contains_any(self, expr_prefix): """ target: test query with expression using json_contains_any method: query with expression using json_contains_any expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema, skip_field_names=[ct.default_json_field_name]) limit = random.randint(10, 20) int_data = [[m for m in range(i, i + limit)] for i in range(default_nb)] str_data = [[str(m) for m in range(i, i + limit)] for i in range(default_nb)] flt_data = [[m * 1.0 for m in range(i, i + limit)] for i in range(default_nb)] bool_data = [[bool(i % 2)] for i in range(default_nb)] list_data = [[[i, str(i + 1)], [i * 1.0, i + 1]] for i in range(default_nb)] mix_data = [[i, i * 1.1, str(i), bool(i % 2), [i, str(i)]] for i in range(default_nb)] for i in range(default_nb): rows[i][ct.default_json_field_name] = { "listInt": int_data[i], "listStr": str_data[i], "listFlt": flt_data[i], "listBool": bool_data[i], "listList": list_data[i], "listMix": mix_data[i], } self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with different data types _id = random.randint(limit, default_nb - limit) # test for int ids = [i for i in range(_id, _id + limit)] expression = f"{expr_prefix}({ct.default_json_field_name}['listInt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert [entity[ct.default_int64_field_name] for entity in res] == cf.assert_json_contains(expression, int_data) # test for string ids = [str(_id), str(_id + 1), str(_id + 2)] expression = f"{expr_prefix}({ct.default_json_field_name}['listStr'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert [entity[ct.default_int64_field_name] for entity in res] == cf.assert_json_contains(expression, str_data) # test for float ids = [_id * 1.0] expression = f"{expr_prefix}({ct.default_json_field_name}['listFlt'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert [entity[ct.default_int64_field_name] for entity in res] == cf.assert_json_contains(expression, flt_data) # test for bool ids = [True] expression = f"{expr_prefix}({ct.default_json_field_name}['listBool'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert [entity[ct.default_int64_field_name] for entity in res] == cf.assert_json_contains(expression, bool_data) # test for list ids = [[_id, str(_id + 1)]] expression = f"{expr_prefix}({ct.default_json_field_name}['listList'], {ids})" res = self.query(client, collection_name, filter=expression, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == 1 # test for mixed data ids = [str(_id)] expression = f"{expr_prefix}({ct.default_json_field_name}['listMix'], {ids})" res = self.query(client, collection_name, filter=expression, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == 1 # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("expr_prefix", ["json_contains_any", "json_contains_all"]) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_expr_json_contains_list_in_list(self, expr_prefix, enable_dynamic_field): """ target: test query with expression using json_contains_any method: query with expression using json_contains_any expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema, skip_field_names=[ct.default_json_field_name]) for i in range(default_nb): rows[i][ct.default_json_field_name] = {"list": [[i, i + 1], [i, i + 2], [i, i + 3]]} self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with list in list _id = random.randint(3, default_nb - 3) ids = [[_id, _id + 1]] expression = f"{expr_prefix}({ct.default_json_field_name}['list'], {ids})" res = self.query(client, collection_name, filter=expression)[0] assert len(res) == 1 ids = [[_id + 4, _id], [_id]] expression = f"{expr_prefix}({ct.default_json_field_name}['list'], {ids})" self.query(client, collection_name, filter=expression, check_task=CheckTasks.check_query_empty) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) @pytest.mark.parametrize("expr_prefix", ["json_contains", "JSON_CONTAINS"]) def test_milvus_client_query_expr_json_contains_pagination(self, enable_dynamic_field, expr_prefix): """ target: test query with expression using json_contains method: query with expression using json_contains expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) limit = default_nb // 3 for i in range(default_nb): rows[i][ct.default_json_field_name] = {"number": i, "list": [m for m in range(i, i + limit)]} self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with pagination expression = f"{expr_prefix}({ct.default_json_field_name}['list'], 1000)" offset = random.randint(1, limit) res = self.query(client, collection_name, filter=expression, limit=limit, offset=offset)[0] assert len(res) == limit - offset # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) @pytest.mark.parametrize("array_length", ["ARRAY_LENGTH", "array_length"]) @pytest.mark.parametrize("op", ["==", "!=", ">", "<="]) def test_milvus_client_query_expr_array_length(self, array_length, op, enable_dynamic_field): """ target: test query with expression using array_length method: query with expression using array_length array_length only support == , != expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field( ct.default_float_array_field_name, DataType.ARRAY, element_type=DataType.FLOAT, max_capacity=1024 ) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema( nb=default_nb, schema=schema, skip_field_names=[ct.default_float_array_field_name] ) length = [] for i in range(default_nb): ran_int = random.randint(50, 53) length.append(ran_int) rows[i][ct.default_float_array_field_name] = [np.float32(j) for j in range(ran_int)] self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query expression = f"{array_length}({ct.default_float_array_field_name}) {op} 51" res = self.query(client, collection_name, filter=expression)[0] # 5. check with local filter expression = expression.replace(f"{array_length}({ct.default_float_array_field_name})", "array_length") filter_ids = [] for i in range(default_nb): array_length = length[i] if eval(expression): filter_ids.append(i) assert len(res) == len(filter_ids) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize( "wildcard_output_fields", [ ["*"], ["*", default_float_field_name], ["*", ct.default_int64_field_name], ["*", ct.default_string_field_name], ], ) def test_milvus_client_query_output_field_wildcard_with_mmap(self, wildcard_output_fields): """ Purpose: Verify that queries with wildcard output fields return correct results, both with and without mmap enabled. Steps: 1. Create a collection with multiple fields and enable dynamic fields. 2. Insert data and build indexes. 3. Query using wildcard output fields and validate the returned fields. 4. Enable mmap, reload the collection, and repeat the query to ensure consistency. Expected: The query results with wildcard output fields should be correct and consistent before and after enabling mmap. """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_int64_field_name, DataType.INT64) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_vector_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name=ct.default_int64_field_name, index_type="AUTOINDEX") index_params.add_index(field_name=ct.default_float_field_name, index_type="AUTOINDEX") index_params.add_index(field_name=ct.default_string_field_name, index_type="AUTOINDEX") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with wildcard output fields actual_res = self.query( client, collection_name, filter=default_search_exp, output_fields=wildcard_output_fields )[0] # 5. test mmap self.release_collection(client, collection_name) self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) self.alter_index_properties( client, collection_name, ct.default_vector_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_int64_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_float_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_string_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) mmap_res = self.query(client, collection_name, filter=default_search_exp, output_fields=wildcard_output_fields)[ 0 ] assert actual_res == mmap_res # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_output_binary_vec_field(self): """ target: test query with binary vec output field method: specify binary vec field as output field expected: return primary field and binary vec field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with binary vector schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_binary_vec_field_name, DataType.BINARY_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index( field_name=ct.default_binary_vec_field_name, index_type="BIN_FLAT", metric_type="JACCARD" ) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. test different output field combinations fields = [ [ct.default_binary_vec_field_name], [default_primary_key_field_name, ct.default_binary_vec_field_name], ] for output_fields in fields: res = self.query(client, collection_name, filter=default_search_exp, output_fields=output_fields)[0] assert set(res[0].keys()) == set(fields[-1]) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_output_primary_field(self): """ target: test query with output field only primary field method: specify int64 primary field as output field expected: return int64 field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with only primary field as output res = self.query( client, collection_name, filter=default_search_exp, output_fields=[default_primary_key_field_name] )[0] # 5. verify only primary field is returned assert set(res[0].keys()) == {default_primary_key_field_name} # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_output_multi_float_vec_field(self): """ target: test query and output multi float vec fields method: a.specify multi vec field as output b.specify output_fields with wildcard % expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with two float vector fields schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field("float_vector1", DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name="float_vector1", index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with multi vec output_fields output_fields = [ct.default_int64_field_name, ct.default_float_vec_field_name, "float_vector1"] exp_res = [] for i in range(2): result_item = {} for field_name in output_fields: result_item[field_name] = rows[i][field_name] exp_res.append(result_item) # Query and verify self.query( client, collection_name, filter=default_term_expr, output_fields=output_fields, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "with_vec": True, "pk_name": ct.default_int64_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_output_multi_sparse_vec_field(self): """ target: test query and output multi sparse vec fields method: a.specify multi vec field as output b.specify output_fields with wildcard % expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with two float vector fields schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field("document", DataType.VARCHAR, max_length=10000, enable_analyzer=True) schema.add_field("sparse1", DataType.SPARSE_FLOAT_VECTOR) schema.add_field("sparse2", DataType.SPARSE_FLOAT_VECTOR) schema.add_field("bm25", DataType.SPARSE_FLOAT_VECTOR) # add bm25 function bm25_function = Function( name="bm25", input_field_names=["document"], output_field_names="bm25", function_type=FunctionType.BM25, ) schema.add_function(bm25_function) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name="sparse1", index_type="SPARSE_INVERTED_INDEX", metric_type="IP") index_params.add_index(field_name="sparse2", index_type="SPARSE_INVERTED_INDEX", metric_type="IP") index_params.add_index( field_name="bm25", index_type="SPARSE_INVERTED_INDEX", metric_type="BM25", params={"bm25_k1": 1.2, "bm25_b": 0.75}, ) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with multi vec output_fields output_fields = [ct.default_int64_field_name, "sparse1", "sparse2"] exp_res = [] for i in range(2): result_item = {} for field_name in output_fields: result_item[field_name] = rows[i][field_name] exp_res.append(result_item) self.query( client, collection_name, filter=default_term_expr, output_fields=output_fields, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "with_vec": True, "pk_name": ct.default_int64_field_name}, )[0] self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("vec_fields", [["binary_vector"], ["binary_vector", "binary_vec1"]]) def test_milvus_client_query_output_mix_float_binary_field(self, vec_fields): """ target: test query and output mix float and binary vec fields method: a.specify mix vec field as output b.specify output_fields with wildcard % expected: output binary vector and float vec """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with float and binary vector fields schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) for vec_field_name in vec_fields: schema.add_field(vec_field_name, DataType.BINARY_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") for vec_field_name in vec_fields: index_params.add_index(field_name=vec_field_name, index_type="BIN_FLAT", metric_type="JACCARD") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with mix vec output_fields output_fields = [ct.default_int64_field_name, ct.default_float_vec_field_name] for vec_field_name in vec_fields: output_fields.append(vec_field_name) res = self.query(client, collection_name, filter=default_term_expr, output_fields=output_fields)[0] assert len(res) == 2 for result_item in res: assert set(result_item.keys()) == set(output_fields) # Query and verify with wildcard res_wildcard = self.query(client, collection_name, filter=default_term_expr, output_fields=["*"])[0] assert len(res_wildcard) == 2 for result_item in res_wildcard: assert set(result_item.keys()) == set(output_fields) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("invalid_fields", ["12-s", 1, [1, "2", 3], (1,), {1: 1}]) def test_milvus_client_query_invalid_output_fields(self, invalid_fields): """ target: test query with invalid output fields method: query with invalid field fields expected: raise exception """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. test invalid output_fields type if invalid_fields == [1, "2", 3]: error = {ct.err_code: 1, ct.err_msg: "Unexpected error, message="} else: error = {ct.err_code: 1, ct.err_msg: "Invalid query format. 'output_fields' must be a list"} self.query( client, collection_name, filter=default_term_expr, output_fields=invalid_fields, check_task=CheckTasks.err_res, check_items=error, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L0) def test_milvus_client_query_partition(self): """ target: test query on partition method: create a partition and query expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) # 2. create partition self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 3. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_name=partition_name) # 4. create index and load partition index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name]) # 5. query on partition and verify results self.query( client, collection_name, filter=default_search_exp, partition_names=[partition_name], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_default_partition(self): """ target: test query on default partition method: query on default partition expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data (will go to default partition by default) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query on default partition and verify results self.query( client, collection_name, filter=default_search_exp, partition_names=[ct.default_partition_name], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows, "pk_name": default_primary_key_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_ignore_growing(self): """ target: test search ignoring growing segment method: 1. create a collection, insert data, create index and load 2. insert data again 3. query with param ignore_growing expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong", auto_id=False) # 2. insert initial data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. insert data again new_rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema, start=10000) self.insert(client, collection_name, new_rows) self.flush(client, collection_name) # 5. query with param ignore_growing ignore_growings = [True, False] for ignore_growing in ignore_growings: if ignore_growing: res = self.query(client, collection_name, filter=default_search_exp, ignore_growing=ignore_growing)[0] assert len(res) == default_nb # verify that only original data (id < 10000) is returned for item in res: assert item[default_primary_key_field_name] < 10000 else: res = self.query(client, collection_name, filter=default_search_exp, ignore_growing=ignore_growing)[0] assert len(res) == default_nb * 2 for item in res: assert item[default_primary_key_field_name] < 10000 or item[default_primary_key_field_name] >= 10000 # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("ignore_growing", [True, False]) def test_milvus_client_query_ignore_growing_after_upsert(self, ignore_growing): """ target: test query ignoring growing segment after upsert method: 1. create a collection, insert data, create index and load 2. upsert the inserted data 3. query with param ignore_growing expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong", auto_id=False) # 2. insert initial data and flush rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. upsert data (which creates growing segment data) upsert_data = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.upsert(client, collection_name, upsert_data) self.flush(client, collection_name) # 5. query with param ignore_growing if ignore_growing: exp_len = 0 else: exp_len = default_nb res = self.query(client, collection_name, filter=default_search_exp, ignore_growing=ignore_growing)[0] assert len(res) == exp_len # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_pagination(self): """ target: test query pagination method: create collection and query with pagination params, verify if the result is ordered by primary key expected: query successfully and verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 3. prepare pagination query int_values = [row[default_primary_key_field_name] for row in rows] pos = 10 for offset in [0, 10, 100]: term_expr = f"{default_primary_key_field_name} in {int_values[offset : pos + offset]}" # Expected results: primary key values from offset to pos+offset res = [] for i in range(offset, min(pos + offset, len(rows))): res.append({default_primary_key_field_name: rows[i][default_primary_key_field_name]}) # 4. query with pagination params query_res = self.query( client, collection_name, filter=term_expr, output_fields=[default_primary_key_field_name], limit=10, check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": default_primary_key_field_name}, )[0] # 5. verify primary key order key_res = [item[key] for item in query_res for key in item] assert key_res == int_values[offset : pos + offset] self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_binary_pagination(self): """ target: test query binary pagination method: create collection and query with pagination params, verify if the result is ordered by primary key expected: query successfully and verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with binary vector schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_binary_vec_field_name, DataType.BINARY_VECTOR, dim=default_dim) self.create_collection(client, collection_name, default_dim, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) index_params = self.prepare_index_params(client)[0] index_params.add_index( field_name=ct.default_binary_vec_field_name, index_type="BIN_FLAT", metric_type="JACCARD" ) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 3. prepare pagination query int_values = [row[default_primary_key_field_name] for row in rows] pos = 10 for offset in [0, 10, 100]: term_expr = f"{default_primary_key_field_name} in {int_values[offset : pos + offset]}" # Expected results: primary key values from offset to pos+offset res = [] for i in range(offset, min(pos + offset, len(rows))): res.append({default_primary_key_field_name: rows[i][default_primary_key_field_name]}) # 4. query with pagination params query_res = self.query( client, collection_name, filter=term_expr, output_fields=[default_primary_key_field_name], limit=10, check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": default_primary_key_field_name}, )[0] # 5. verify primary key order key_res = [item[key] for item in query_res for key in item] assert key_res == int_values[offset : pos + offset] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_pagination_with_partition(self): """ target: test query pagination on partition method: create a partition and query with different offset expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name) # 2. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_name=partition_name) # 3. verify entity count self.flush(client, collection_name) stats = self.get_collection_stats(client, collection_name)[0] assert stats["row_count"] == default_nb # 4. create index and load partition self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name]) for offset in [0, 10, 100]: # 5. prepare expected results (first 2 records) res = [] for i in range(offset, min(10 + offset, len(rows))): res.append({default_primary_key_field_name: rows[i][default_primary_key_field_name]}) # 6. query with pagination params on partition self.query( client, collection_name, filter=default_search_exp, output_fields=[default_primary_key_field_name], partition_names=[partition_name], offset=offset, limit=10, check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": default_primary_key_field_name}, ) # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_pagination_with_insert_data(self): """ target: test query pagination on partition method: create a partition and query with pagination expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data df = cf.gen_default_dataframe_data() rows = df.to_dict("records") self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. prepare expected results (first 2 records) int_values = [row[ct.default_int64_field_name] for row in rows] pos = 10 for offset in [0, 10, 100]: term_expr = f"{ct.default_int64_field_name} in {int_values[offset : pos + offset]}" # Expected results: primary key values from offset to pos+offset res = [] for i in range(offset, min(pos + offset, len(rows))): res.append({ct.default_int64_field_name: rows[i][ct.default_int64_field_name]}) # 5. query with pagination params query_res = self.query( client, collection_name, filter=term_expr, output_fields=[ct.default_int64_field_name], limit=10, check_task=CheckTasks.check_query_results, check_items={exp_res: res, "pk_name": ct.default_int64_field_name}, )[0] key_res = [item[ct.default_int64_field_name] for item in query_res] assert key_res == int_values[offset : pos + offset] self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_pagination_without_limit(self): """ target: test query pagination without limit method: create collection and query with pagination params(only offset), compare the result with query without pagination params expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load collection self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with only offset parameter (no limit) query_res_without_pagination = self.query( client, collection_name, filter="id in [0, 1]", check_task=CheckTasks.check_query_results, check_items={exp_res: rows[:2], "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] for offset in [0, 10, 100]: query_res_with_offset = self.query( client, collection_name, filter="id in [0, 1]", offset=offset, check_task=CheckTasks.check_query_results, check_items={exp_res: rows[:2], "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] assert query_res_with_offset == query_res_without_pagination self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_pagination_with_offset_over_num_entities(self): """ target: test query pagination with offset over num_entities method: query with offset over num_entities expected: return an empty list """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load collection self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with offset over num_entities # Use a broader query that could return results, but offset is too large for offset in [3000, 5000]: res = self.query(client, collection_name, filter=default_search_exp, offset=offset, limit=10)[0] # 5. verify empty result assert len(res) == 0 # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_query_with_expression(self, enable_dynamic_field): """ target: test query with different expr method: query with different boolean expr expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection( client, collection_name, schema=schema, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field, ) # 2. insert data nb = ct.default_nb rows = cf.gen_row_data_by_schema(nb=nb, schema=schema) self.insert(client, collection_name, rows) # assert self.num_entities(client, collection_name)[0] == nb self.flush(client, collection_name) # 3. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. filter result with expression in collection for expressions in cf.gen_normal_expressions_and_templates(): log.debug(f"test_milvus_client_query_with_expression: {expressions}") expr = expressions[0].replace("&&", "and").replace("||", "or") filter_ids = [] for i, row in enumerate(rows): int64 = row[ct.default_int64_field_name] # noqa: F841 - referenced by eval(expr) float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) if not expr or eval(expr): filter_ids.append(row[ct.default_int64_field_name]) # query and verify result res = self.query(client, collection_name, filter=expr, limit=nb)[0] query_ids = set(map(lambda x: x[ct.default_int64_field_name], res)) assert query_ids == set(filter_ids) # query again with expression template expr = cf.get_expr_from_template(expressions[1]).replace("&&", "and").replace("||", "or") expr_params = cf.get_expr_params_from_template(expressions[1]) res = self.query(client, collection_name, filter=expr, filter_params=expr_params, limit=nb)[0] query_ids = set(map(lambda x: x[ct.default_int64_field_name], res)) assert query_ids == set(filter_ids) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_pagination_with_expression(self): """ target: test query pagination with different expression method: query with different expression and verify the result expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection( client, collection_name, schema=schema, consistency_level="Strong", enable_dynamic_field=False ) # 2. insert data nb = 1000 rows = cf.gen_row_data_by_schema(nb=nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. filter result with expression in collection for expressions in cf.gen_normal_expressions_and_templates()[1:]: log.debug(f"test_milvus_client_query_pagination_with_expression: {expressions}") expr = expressions[0].replace("&&", "and").replace("||", "or") template_expr = cf.get_expr_from_template(expressions[1]).replace("&&", "and").replace("||", "or") expr_params = cf.get_expr_params_from_template(expressions[1]) filter_ids = [] for i, row in enumerate(rows): int64 = row[ct.default_int64_field_name] # noqa: F841 - referenced by eval(expr) float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) if not expr or eval(expr): filter_ids.append(row[ct.default_int64_field_name]) for offset in [0, 10, 100]: # query and verify result query_params = {"offset": offset, "limit": 10} res = self.query(client, collection_name, filter=expr, **query_params)[0] query_ids = [item[ct.default_int64_field_name] for item in res] expected_ids = filter_ids[offset : offset + 10] assert query_ids == expected_ids # query again with expression template res = self.query( client, collection_name, filter=template_expr, filter_params=expr_params, **query_params )[0] query_ids = [item[ct.default_int64_field_name] for item in res] assert query_ids == expected_ids # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_mmap_query_expr_empty_pk_string(self): """ target: turn on mmap to test queries using empty expression method: enable mmap to query for empty expressions with restrictions. expected: return the first K results in order """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with string primary key schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=65535, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection( client, collection_name, schema=schema, consistency_level="Strong", enable_dynamic_field=False ) # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name=ct.default_string_field_name, index_type="AUTOINDEX") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # prepare expected results exp_ids = ["0", "1", "10", "100", "1000", "1001", "1002", "1003", "1004", "1005"] expected_res = [] for ids in exp_ids: expected_res.append({ct.default_string_field_name: ids}) res = self.query(client, collection_name, filter="", limit=ct.default_limit)[0] query_res1 = [{ct.default_string_field_name: item[ct.default_string_field_name]} for item in res] res = self.query(client, collection_name, filter="", limit=5, offset=5)[0] query_res2 = [{ct.default_string_field_name: item[ct.default_string_field_name]} for item in res] # 4. Set mmap enabled before loading self.release_collection(client, collection_name) self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) describe_res = self.describe_collection(client, collection_name)[0] properties = describe_res.get("properties") assert properties["mmap.enabled"] == "True" self.alter_index_properties( client, collection_name, ct.default_float_vec_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_string_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) # 5. query with empty expression and limit res = self.query(client, collection_name, filter="", limit=ct.default_limit)[0] mmap_query_res = [{ct.default_string_field_name: item[ct.default_string_field_name]} for item in res] assert mmap_query_res == expected_res == query_res1 # 6. query with empty expression, limit + offset expected_res_offset = expected_res[5:] res = self.query(client, collection_name, filter="", limit=5, offset=5)[0] mmap_query_res = [{ct.default_string_field_name: item[ct.default_string_field_name]} for item in res] assert mmap_query_res == expected_res_offset == query_res2 # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) def test_milvus_client_enable_mmap_query_with_expression(self, enable_dynamic_field): """ target: turn on mmap use different expr queries method: turn on mmap and query with different expr expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection( client, collection_name, schema=schema, consistency_level="Strong", enable_dynamic_field=enable_dynamic_field, ) # 2. insert data nb = 1000 rows = cf.gen_row_data_by_schema(nb=nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. enable mmap and create index self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.alter_index_properties( client, collection_name, ct.default_float_vec_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) # 4. filter result with expression in collection for expressions in cf.gen_normal_expressions_and_templates()[1:]: log.debug(f"expr: {expressions}") expr = expressions[0].replace("&&", "and").replace("||", "or") filter_ids = [] for i, row in enumerate(rows): if enable_dynamic_field: int64 = row[ct.default_int64_field_name] # noqa: F841 - referenced by eval(expr) float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) else: int64 = row[ct.default_int64_field_name] # noqa: F841 - referenced by eval(expr) float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) if not expr or eval(expr): filter_ids.append(row[ct.default_int64_field_name]) # query and verify result res = self.query(client, collection_name, filter=expr, limit=nb)[0] query_ids = set(map(lambda x: x[ct.default_int64_field_name], res)) assert query_ids == set(filter_ids) # query again with expression template expr = cf.get_expr_from_template(expressions[1]).replace("&&", "and").replace("||", "or") expr_params = cf.get_expr_params_from_template(expressions[1]) res = self.query(client, collection_name, filter=expr, filter_params=expr_params, limit=nb)[0] query_ids = set(map(lambda x: x[ct.default_int64_field_name], res)) assert query_ids == set(filter_ids) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_mmap_query_string_field_not_primary_is_empty(self): """ target: enable mmap, use string expr to test query, string field is not the main field method: create collection , string field is primary enable mmap collection load and insert empty data with string field collection query uses string expr in string field expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=65535) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with empty string fields nb = 3000 rows = cf.gen_row_data_by_schema(nb=nb, schema=schema) # Set all string fields to empty for row in rows: row[ct.default_string_field_name] = "" self.insert(client, collection_name, rows) self.flush(client, collection_name) assert self.get_collection_stats(client, collection_name)[0]["row_count"] == nb # 3. enable mmap and create index index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name=ct.default_string_field_name, index_type="AUTOINDEX") index_params.add_index(field_name=ct.default_float_field_name, index_type="AUTOINDEX") self.create_index(client, collection_name, index_params) self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) self.alter_index_properties( client, collection_name, ct.default_float_vec_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_string_field_name, properties={"mmap.enabled": True} ) self.alter_index_properties( client, collection_name, ct.default_float_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) # 4. query with empty string expression output_fields = [ct.default_int64_field_name, ct.default_float_field_name, ct.default_string_field_name] expr = 'varchar == ""' res = self.query(client, collection_name, filter=expr, output_fields=output_fields)[0] assert len(res) == nb self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("expression", cf.gen_normal_string_expressions([ct.default_string_field_name])) def test_milvus_client_mmap_query_string_is_primary(self, expression): """ target: test query with output field only primary field method: specify string primary field as output field expected: return string primary field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with string as primary key schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=65535, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) actual_res = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] # 3. enable mmap and create index self.release_collection(client, collection_name) self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) self.alter_index_properties( client, collection_name, ct.default_float_vec_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) # 4. query with string expression and only string field as output mmap_res = self.query(client, collection_name, filter=expression, output_fields=[ct.default_string_field_name])[ 0 ] assert set(mmap_res[0].keys()) == {ct.default_string_field_name} assert actual_res == mmap_res self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_mmap_query_string_expr_with_prefixes_and_suffix(self): """ target: test query with prefix string expression when mmap enabled method: specify string is primary field, use prefix string expr with mmap expected: verify query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with string primary key schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query before enabling mmap expected_results = {} normal_results = {} for varchar_expression in [perfix_expr, suffix_expr, inner_match_expr]: exp_res = [] if varchar_expression == perfix_expr: template_expression = f"{ct.default_string_field_name} like {{pattern}}" template_params = {"pattern": "0%"} elif varchar_expression == suffix_expr: template_expression = f"{ct.default_string_field_name} like {{pattern}}" template_params = {"pattern": "%0"} else: template_expression = f"{ct.default_string_field_name} like {{pattern}}" template_params = {"pattern": "%0%"} if varchar_expression == perfix_expr: # varchar like "0%" for row in rows: varchar_value = row[ct.default_string_field_name] if varchar_value.startswith("0"): exp_res.append( { ct.default_string_field_name: varchar_value, ct.default_float_vec_field_name: row[ct.default_float_vec_field_name], } ) break # Only take the first match like original test elif varchar_expression == suffix_expr: # varchar like "%0" for row in rows: varchar_value = row[ct.default_string_field_name] if varchar_value.endswith("0"): exp_res.append( { ct.default_string_field_name: varchar_value, ct.default_float_vec_field_name: row[ct.default_float_vec_field_name], } ) elif varchar_expression == inner_match_expr: # varchar like "%0%" for row in rows: varchar_value = row[ct.default_string_field_name] if "0" in varchar_value: exp_res.append( { ct.default_string_field_name: varchar_value, ct.default_float_vec_field_name: row[ct.default_float_vec_field_name], } ) expected_results[varchar_expression] = exp_res normal_results[varchar_expression] = self.query( client, collection_name, filter=varchar_expression, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "pk_name": ct.default_string_field_name}, )[0] template_res = self.query( client, collection_name, filter=template_expression, filter_params=template_params, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "pk_name": ct.default_string_field_name}, )[0] assert normal_results[varchar_expression] == template_res # 5. enable mmap and reload self.release_collection(client, collection_name) self.alter_collection_properties(client, collection_name, properties={"mmap.enabled": True}) self.alter_index_properties( client, collection_name, ct.default_float_vec_field_name, properties={"mmap.enabled": True} ) self.load_collection(client, collection_name) # 6. query after enabling mmap for varchar_expression, exp_res in expected_results.items(): mmap_res = self.query( client, collection_name, filter=varchar_expression, check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "pk_name": ct.default_string_field_name}, )[0] assert normal_results[varchar_expression] == mmap_res self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("with_growing", [True, False]) def test_milvus_client_query_to_get_latest_entity_with_dup_ids(self, with_growing): """ target: test query to get latest entity with duplicate primary keys method: 1.create collection and insert dup primary key = 0 2.query with expr=dup_id expected: return the latest entity; verify the result is same as dedup entities """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=16) schema.add_field(ct.default_float_field_name, DataType.FLOAT) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with duplicate primary keys nb = 50 rounds = 10 for i in range(rounds): # Generate data for this round rows = [] for j in range(nb): rows.append( { ct.default_int64_field_name: i, # Same primary key for all entities in this round ct.default_float_vec_field_name: [random.random() for _ in range(16)], ct.default_float_field_name: float(j), } ) self.insert(client, collection_name, rows) # Re-insert the last piece of data to refresh the timestamp last_piece = [ { ct.default_int64_field_name: i, ct.default_float_vec_field_name: [random.random() for _ in range(16)], ct.default_float_field_name: float(nb - 1), # This should be the latest value } ] self.insert(client, collection_name, last_piece) # 3. flush if not testing growing segments if not with_growing: self.flush(client, collection_name) # 4. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 5. verify the result returns the latest entity if there are duplicate primary keys expr = f"{ct.default_int64_field_name} == 0" res = self.query( client, collection_name, filter=expr, output_fields=[ct.default_int64_field_name, ct.default_float_field_name], )[0] assert len(res) == 1, f"Expected 1 result for duplicate primary key 0, got {len(res)}" assert res[0][ct.default_float_field_name] == float(nb - 1), ( f"Expected latest float value {float(nb - 1)}, got {res[0][ct.default_float_field_name]}" ) # 6. verify the result is same as dedup entities (should return one entity per round) expr = f"{ct.default_int64_field_name} >= 0" res = self.query( client, collection_name, filter=expr, output_fields=[ct.default_int64_field_name, ct.default_float_field_name], )[0] assert len(res) == rounds # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("vector_data_type", ct.all_dense_vector_types) def test_milvus_client_query_output_all_vector_type(self, vector_data_type): """ target: test query output different vector type method: create index and specify vec field as output field expected: return primary field and vec field """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with specified vector type schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, vector_data_type, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with vector field as output expected_fields = [ct.default_int64_field_name, ct.default_float_vec_field_name] res = self.query( client, collection_name, filter=default_term_expr, output_fields=[ct.default_float_vec_field_name] )[0] # 5. verify that query returns both primary key and vector field assert set(res[0].keys()) == set(expected_fields) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_growing_segment_data(self): """ target: test query data in the growing segment method: 1. create collection 2. load collection 3. insert without flush 4. query expected: Data can be queried """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.load_collection(client, collection_name) # 2. insert data without flush (data will be in growing segment) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=100, schema=schema_info) self.insert(client, collection_name, rows) # 3. prepare expected result - find the entity with primary key = 1 exp_res = [] for row in rows: if row[default_primary_key_field_name] == 1: exp_res.append(row) break # 4. query for entity with primary key = 1 self.query( client, collection_name, filter=f"{default_primary_key_field_name} in [1]", check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "pk_name": default_primary_key_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("auto_id", [True, False]) def test_milvus_client_query_using_all_types_of_default_value(self, auto_id): """ target: test create collection with default_value method: create a schema with all fields using default value and query expected: query results are as expected """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() schema = self.create_schema(client, enable_dynamic_field=False, auto_id=auto_id)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) # Add various scalar fields with default values schema.add_field(ct.default_int8_field_name, DataType.INT8, default_value=np.int8(8)) schema.add_field(ct.default_int16_field_name, DataType.INT16, default_value=np.int16(16)) schema.add_field(ct.default_int32_field_name, DataType.INT32, default_value=np.int32(32)) schema.add_field(ct.default_int64_field_name, DataType.INT64, default_value=np.int64(64)) schema.add_field(ct.default_float_field_name, DataType.FLOAT, default_value=np.float32(3.14)) schema.add_field(ct.default_double_field_name, DataType.DOUBLE, default_value=np.double(3.1415)) schema.add_field(ct.default_bool_field_name, DataType.BOOL, default_value=False) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100, default_value="abc") self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # Insert data - only provide required fields (pk and vector), other fields will use default values schema_info = self.describe_collection(client, collection_name)[0] skip_fields = [ ct.default_int8_field_name, ct.default_int16_field_name, ct.default_int32_field_name, ct.default_int64_field_name, ct.default_float_field_name, ct.default_double_field_name, ct.default_bool_field_name, ct.default_string_field_name, ] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info, skip_field_names=skip_fields) self.insert(client, collection_name, rows) self.flush(client, collection_name) # Create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # Query with specific primary keys and verify default values res = self.query(client, collection_name, filter=default_search_exp, output_fields=["*"])[0] # Verify all default values are correctly applied first_result = res[0] assert first_result[ct.default_int8_field_name] == 8, ( f"Expected int8 default value 8, got {first_result[ct.default_int8_field_name]}" ) assert first_result[ct.default_int16_field_name] == 16, ( f"Expected int16 default value 16, got {first_result[ct.default_int16_field_name]}" ) assert first_result[ct.default_int32_field_name] == 32, ( f"Expected int32 default value 32, got {first_result[ct.default_int32_field_name]}" ) assert first_result[ct.default_int64_field_name] == 64, ( f"Expected int64 default value 64, got {first_result[ct.default_int64_field_name]}" ) assert first_result[ct.default_float_field_name] == np.float32(3.14), ( f"Expected float default value 3.14, got {first_result[ct.default_float_field_name]}" ) assert first_result[ct.default_double_field_name] == 3.1415, ( f"Expected double default value 3.1415, got {first_result[ct.default_double_field_name]}" ) assert first_result[ct.default_bool_field_name] is False, ( f"Expected bool default value False, got {first_result[ct.default_bool_field_name]}" ) assert first_result[ct.default_string_field_name] == "abc", ( f"Expected string default value 'abc', got {first_result[ct.default_string_field_name]}" ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L0) def test_milvus_client_query_multi_logical_exprs(self): """ target: test the scenario which query with many logical expressions method: 1. create collection 3. query the expr that like: int64 == 0 || int64 == 1 ........ expected: run successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with multi logical expressions # Create expression like: int64 == 0 || int64 == 1 || int64 == 2 ... || int64 == 59 multi_exprs = " || ".join(f"{default_primary_key_field_name} == {i}" for i in range(60)) self.query( client, collection_name, filter=multi_exprs, check_task=CheckTasks.check_query_results, check_items={"exp_res": rows[:60], "pk_name": default_primary_key_field_name}, ) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_empty_partition_names(self): """ target: test query with empty partition_names method: query with partition_names=[] expected: query from all partitions """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert [0, half) into partition_w, [half, nb) into _default half = default_nb // 2 schema_info = self.describe_collection(client, collection_name)[0] # Insert first half into custom partition rows_partition = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=0) self.insert(client, collection_name, rows_partition, partition_name=partition_name) # Insert second half into default partition rows_default = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=half) self.insert(client, collection_name, rows_default) # 3. create index and load both partitions index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name, ct.default_partition_name]) # 4. query from empty partition_names (should query all partitions) term_expr = f"{default_primary_key_field_name} in [0, {half}, {default_nb - 1}]" # Prepare expected results by combining data from both partitions all_rows = rows_partition + rows_default exp_res = [all_rows[0], all_rows[half], all_rows[default_nb - 1]] self.query( client, collection_name, filter=term_expr, partition_names=[], check_task=CheckTasks.check_query_results, check_items={"exp_res": exp_res, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_on_specific_partition(self): """ Target: Verify that querying a specific partition only returns data from that partition. Method: 1. Create a collection and two partitions (partition1 and partition2). 2. Insert the first half of the data into partition1 and the second half into partition2. 3. Create an index and load both partitions. 4. Query only partition1 and check that the returned results only contain data from partition1. 5. Clean up the collection. Expected: Only data from partition1 is returned and matches the inserted data. """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name1 = cf.gen_unique_str("partition1") partition_name2 = cf.gen_unique_str("partition2") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name1) self.create_partition(client, collection_name, partition_name2) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data to partition half = default_nb // 2 schema_info = self.describe_collection(client, collection_name)[0] rows_partition1 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=0) rows_partition2 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=half) self.insert(client, collection_name, rows_partition1, partition_name=partition_name1) self.insert(client, collection_name, rows_partition2, partition_name=partition_name2) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name1, partition_name2]) # 4. query on partition self.query( client, collection_name, filter=default_search_exp, partition_names=[partition_name1], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows_partition1, "with_vec": True, "pk_name": default_primary_key_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_multi_partitions_multi_results(self): """ target: test query on multi partitions and get multi results method: 1.create two partitions and insert entities into them 2.query on two partitions and get multi results expected: query results from two partitions """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name1 = cf.gen_unique_str("partition1") partition_name2 = cf.gen_unique_str("partition2") # 1. create collection and two partitions self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name1) self.create_partition(client, collection_name, partition_name2) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data into two partitions half = default_nb // 2 schema_info = self.describe_collection(client, collection_name)[0] rows_partition1 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=0) self.insert(client, collection_name, rows_partition1, partition_name=partition_name1) rows_partition2 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=half) self.insert(client, collection_name, rows_partition2, partition_name=partition_name2) self.flush(client, collection_name) # 3. create index and load both partitions index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name1, partition_name2]) # 4. query on two partitions to get multi results term_expr = f"{default_primary_key_field_name} in [{half - 1}, {half}]" rows = rows_partition1 + rows_partition2 self.query( client, collection_name, filter=term_expr, partition_names=[partition_name1, partition_name2], check_task=CheckTasks.check_query_results, check_items={ "exp_res": rows[half - 1 : half + 1], "with_vec": True, "pk_name": default_primary_key_field_name, }, )[0] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_multi_partitions_single_results(self): """ target: test query on multi partitions and get multi results method: 1.create two partitions and insert entities into them 2.query on two partitions and query single results expected: query from two partitions and get single result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name1 = cf.gen_unique_str("partition1") partition_name2 = cf.gen_unique_str("partition2") # 1. create collection and two partitions self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name1) self.create_partition(client, collection_name, partition_name2) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data into two partitions half = default_nb // 2 schema_info = self.describe_collection(client, collection_name)[0] rows_partition1 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=0) self.insert(client, collection_name, rows_partition1, partition_name=partition_name1) rows_partition2 = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=half) self.insert(client, collection_name, rows_partition2, partition_name=partition_name2) self.flush(client, collection_name) # 3. create index and load both partitions index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name1, partition_name2]) # 4. query on two partitions to get multi results term_expr = f"{default_primary_key_field_name} in [{half}]" rows = rows_partition1 + rows_partition2 self.query( client, collection_name, filter=term_expr, partition_names=[partition_name1, partition_name2], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows[half : half + 1], "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_empty_partition(self): """ target: test query on empty partition method: query on an empty partition expected: empty query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name) # 2. verify partition is empty partition_info = self.get_partition_stats(client, collection_name, partition_name)[0] assert partition_info["row_count"] == 0 # 3. create index and load partition self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name]) # 4. query on empty partition res = self.query(client, collection_name, filter=default_search_exp, partition_names=[partition_name])[0] assert len(res) == 0 # 5. clean up self.drop_collection(client, collection_name) class TestQueryOperation(TestMilvusClientV2Base): """ ****************************************************************** The following cases are used to test query interface operations ****************************************************************** """ @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_without_connection(self): """ target: test query without connection method: close connect and query expected: raise exception """ client_temp = self._client(alias="client_temp") collection_name = cf.gen_collection_name_by_testcase_name() # Remove connection self.close(client_temp) error = {ct.err_code: 1, ct.err_msg: "should create connection first"} self.query( client_temp, collection_name, filter=default_search_exp, check_task=CheckTasks.err_res, check_items=error ) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("output_fields", [None, ["count(*)"]]) def test_milvus_client_query_without_loading(self, output_fields): """ target: test query without loading method: query without loading expected: raise exception """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.release_collection(client, collection_name) # 3. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) collection_info = self.get_collection_stats(client, collection_name)[0] assert collection_info["row_count"] == default_nb # 4. query without loading error = {ct.err_code: 65535, ct.err_msg: "collection not loaded"} self.query( client, collection_name, filter=default_search_exp, output_fields=output_fields, check_task=CheckTasks.err_res, check_items=error, ) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_partition_repeatedly(self): """ target: test query repeatedly on partition method: query on partition twice expected: verify query result is identical """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong", auto_id=False) self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data to partition schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_name=partition_name) self.flush(client, collection_name) # 3. check number of entities partition_info = self.get_partition_stats(client, collection_name, partition_name)[0] assert partition_info["row_count"] == default_nb # 4. create index and load partition index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name]) # 5. query twice on the same partition res_one = self.query( client, collection_name, filter=default_search_exp, partition_names=[partition_name], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows, "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] res_two = self.query( client, collection_name, filter=default_search_exp, partition_names=[partition_name], check_task=CheckTasks.check_query_results, check_items={"exp_res": rows, "with_vec": True, "pk_name": default_primary_key_field_name}, )[0] assert res_one == res_two, "Query results should be identical when querying the same partition repeatedly" self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.skip(reason="the assert is not good") def test_milvus_client_query_with_bloom_filter(self): """ target: test query with bloom filter in PK method: compare time with filter by bloom filter and without bloom filter expected: query with bloom filter in PK should faster than without bloom filter """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=True)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=5) self.create_collection(client, collection_name=collection_name, schema=schema) index_params = client.prepare_index_params() index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] insert_offset = 0 insert_nb = 1000 for i in range(10): rows = cf.gen_row_data_by_schema(nb=insert_nb, schema=schema_info, start=insert_offset) self.insert(client, collection_name, rows) self.flush(client, collection_name) insert_offset += insert_nb # 3. load self.load_collection(client, collection_name) # 4. query with bloom filter and without bloom filter start_time = time.perf_counter() self.query( client, collection_name=collection_name, filter=f"{default_primary_key_field_name} != -1", output_fields=["count(*)"], ) end_time = time.perf_counter() run_time1 = end_time - start_time # with bloom filter start_time = time.perf_counter() self.query( client, collection_name=collection_name, filter=f"{default_primary_key_field_name} == -1", output_fields=["count(*)"], ) end_time = time.perf_counter() run_time2 = end_time - start_time print(f"rt1: {run_time1}s rt2: {run_time2}s") log.info(f"rt1: {run_time1}s rt2: {run_time2}s") # 5. verify without bloom filter should slower than with bloom filter assert run_time1 > run_time2 # 6. clean up self.drop_collection(client, collection_name) class TestMilvusClientGetInvalid(TestMilvusClientV2Base): """Test case of search interface""" """ ****************************************************************** # The following are invalid base cases ****************************************************************** """ @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize( "name", ["12-s", "12 s", "(mn)", "中文", "%$#", "".join("a" for i in range(ct.max_name_length + 1))] ) def test_milvus_client_get_invalid_collection_name(self, name): """ target: test get interface invalid cases method: invalid collection name expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert default_nb = 1000 rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [i for i in range(default_nb)] # 3. get first primary key error = {ct.err_code: 1100, ct.err_msg: "Invalid collection name"} self.get(client, name, ids=pks[0:1], check_task=CheckTasks.err_res, check_items=error) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_get_not_exist_collection_name(self): """ target: test get interface invalid cases method: invalid collection name expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert default_nb = 1000 rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [i for i in range(default_nb)] # 3. get first primary key name = "invalid" error = {ct.err_code: 100, ct.err_msg: f"can't find collection[database=default][collection={name}]"} self.get(client, name, ids=pks[0:1], check_task=CheckTasks.err_res, check_items=error) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("invalid_ids", ["中文", "%$#"]) def test_milvus_client_get_invalid_ids(self, invalid_ids): """ target: test get interface invalid cases method: invalid collection name expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert default_nb = 1000 rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) # 3. get first primary key error = {ct.err_code: 1100, ct.err_msg: "cannot parse expression"} self.get(client, collection_name, ids=invalid_ids, check_task=CheckTasks.err_res, check_items=error) self.drop_collection(client, collection_name) class TestMilvusClientGetValid(TestMilvusClientV2Base): """Test case of search interface""" @pytest.fixture(scope="function", params=[False, True]) def auto_id(self, request): yield request.param @pytest.fixture(scope="function", params=["COSINE", "L2"]) def metric_type(self, request): yield request.param """ ****************************************************************** # The following are valid base cases ****************************************************************** """ @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_get_normal(self): """ target: test get interface method: create connection, collection, insert delete, and search expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert default_nb = 1000 rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [i for i in range(default_nb)] # 3. get first primary key first_pk_data = self.get(client, collection_name, ids=pks[0:1])[0] assert len(first_pk_data) == len(pks[0:1]) first_pk_data_1 = self.get(client, collection_name, ids=0)[0] assert first_pk_data == first_pk_data_1 self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_get_output_fields(self): """ target: test get interface with output fields method: create connection, collection, insert delete, and search expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") # 2. insert default_nb = 1000 rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: i, default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [i for i in range(default_nb)] # 3. get first primary key output_fields_array = [ default_primary_key_field_name, default_vector_field_name, default_float_field_name, default_string_field_name, ] first_pk_data = self.get(client, collection_name, ids=pks[0:1], output_fields=output_fields_array)[0] assert len(first_pk_data) == len(pks[0:1]) assert len(first_pk_data[0]) == len(output_fields_array) first_pk_data_1 = self.get(client, collection_name, ids=0, output_fields=output_fields_array)[0] assert first_pk_data == first_pk_data_1 assert len(first_pk_data_1[0]) == len(output_fields_array) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.skip(reason="pymilvus issue 2056") def test_milvus_client_get_normal_string(self): """ target: test get interface for string field method: create connection, collection, insert delete, and search expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, id_type="string", max_length=ct.default_length) # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: str(i), default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [str(i) for i in range(default_nb)] # 3. get first primary key first_pk_data = self.get(client, collection_name, ids=pks[0:1])[0] assert len(first_pk_data) == len(pks[0:1]) first_pk_data_1 = self.get(client, collection_name, ids="0")[0] assert first_pk_data == first_pk_data_1 self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.skip(reason="pymilvus issue 2056") def test_milvus_client_get_normal_string_output_fields(self): """ target: test get interface for string field method: create connection, collection, insert delete, and search expected: search/query successfully without deleted data """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, id_type="string", max_length=ct.default_length) # 2. insert rng = np.random.default_rng(seed=19530) rows = [ { default_primary_key_field_name: str(i), default_vector_field_name: list(rng.random((1, default_dim))[0]), default_float_field_name: i * 1.0, default_string_field_name: str(i), } for i in range(default_nb) ] self.insert(client, collection_name, rows) pks = [str(i) for i in range(default_nb)] # 3. get first primary key output_fields_array = [ default_primary_key_field_name, default_vector_field_name, default_float_field_name, default_string_field_name, ] first_pk_data = self.get(client, collection_name, ids=pks[0:1], output_fields=output_fields_array)[0] assert len(first_pk_data) == len(pks[0:1]) assert len(first_pk_data[0]) == len(output_fields_array) first_pk_data_1 = self.get(client, collection_name, ids="0", output_fields=output_fields_array)[0] assert first_pk_data == first_pk_data_1 assert len(first_pk_data_1[0]) == len(output_fields_array) self.drop_collection(client, collection_name) # Valid (index_type, json_cast_type) combinations for JSON path index # @pytest.fixture(scope="function", params=["DOUBLE", "VARCHAR", "json"", "bool", "ARRAY_DOUBLE"]) _json_path_index_params_query = [ ("INVERTED", "DOUBLE"), ("INVERTED", "ARRAY_DOUBLE"), ("STL_SORT", "DOUBLE"), ] class TestMilvusClientQueryJsonPathIndex(TestMilvusClientV2Base): """Test case of search interface""" @pytest.fixture(scope="function", params=_json_path_index_params_query, ids=[f"{t[0]}_{t[1]}" for t in _json_path_index_params_query]) def json_index_params(self, request): yield request.param @pytest.fixture(scope="function") def supported_varchar_scalar_index(self, json_index_params): yield json_index_params[0] @pytest.fixture(scope="function") def supported_json_cast_type(self, json_index_params): yield json_index_params[1] """ ****************************************************************** # The following are valid base cases ****************************************************************** """ @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("enable_dynamic_field", [True, False]) @pytest.mark.parametrize("is_flush", [True, False]) @pytest.mark.parametrize("is_release", [True, False]) @pytest.mark.parametrize("single_data_num", [50]) def test_milvus_client_search_json_path_index_all_expressions( self, enable_dynamic_field, supported_json_cast_type, supported_varchar_scalar_index, is_flush, is_release, single_data_num, ): """ target: test query after json path index with all supported basic expressions method: Query after json path index with all supported basic expressions step: 1. create collection 2. insert with different data distribution 3. flush if specified 4. query when there is no json path index under all expressions 5. release if specified 6. prepare index params with json path index 7. create json path index 8. create same json index twice 9. reload collection if released before to make sure the new index load successfully 10. sleep for 60s to make sure the new index load successfully without release and reload operations 11. query after there is json path index under all expressions which should get the same result with that without json path index expected: query successfully after there is json path index under all expressions which should get the same result with that without json path index """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection json_field_name = "json_field" schema = self.create_schema(client, enable_dynamic_field=enable_dynamic_field)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True, auto_id=False) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(default_string_field_name, DataType.VARCHAR, max_length=64) if not enable_dynamic_field: schema.add_field(json_field_name, DataType.JSON, nullable=True) index_params = self.prepare_index_params(client)[0] index_params.add_index(default_vector_field_name, metric_type="COSINE") self.create_collection(client, collection_name, schema=schema, index_params=index_params) # 2. insert with different data distribution vectors = cf.gen_vectors(default_nb + 60, default_dim) inserted_data_distribution = ct.get_all_kind_data_distribution nb_single = single_data_num for i in range(len(inserted_data_distribution)): rows = [ { default_primary_key_field_name: j, default_vector_field_name: vectors[j], default_string_field_name: f"{j}", json_field_name: inserted_data_distribution[i], } for j in range(i * nb_single, (i + 1) * nb_single) ] assert len(rows) == nb_single self.insert(client, collection_name=collection_name, data=rows) log.info(f"inserted {nb_single} {inserted_data_distribution[i]}") # 3. flush if specified if is_flush: self.flush(client, collection_name) # 4. query when there is no json path index under all expressions # skip negative expression for issue 40685 # "my_json['a'] != 1", "my_json['a'] != 1.0", "my_json['a'] != '1'", "my_json['a'] != 1.1", "my_json['a'] not in [1]" express_list = cf.gen_json_field_expressions_all_single_operator(supported_json_cast_type) compare_dict = {} for i in range(len(express_list)): json_list = [] id_list = [] log.info(f"query with filter {express_list[i]} before json path index is:") res = self.query( client, collection_name=collection_name, filter=express_list[i], output_fields=["count(*)"] )[0] count = res[0]["count(*)"] log.info(f"The count(*) after query with filter {express_list[i]} before json path index is: {count}") res = self.query( client, collection_name=collection_name, filter=express_list[i], output_fields=[f"{json_field_name}"] )[0] for single in res: id_list.append(single[f"{default_primary_key_field_name}"]) json_list.append(single[f"{json_field_name}"]) assert count == len(id_list) assert count == len(json_list) compare_dict.setdefault(f"{i}", {}) compare_dict[f"{i}"]["id_list"] = id_list compare_dict[f"{i}"]["json_list"] = json_list # 5. release if specified if is_release: self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 6. prepare index params with json path index index_name = "json_index" index_params = self.prepare_index_params(client)[0] if supported_json_cast_type == "ARRAY_DOUBLE": # For ARRAY_DOUBLE type, use array paths json_path_list = [f"{json_field_name}['a']"] else: json_path_list = [ f"{json_field_name}", f"{json_field_name}[0]", f"{json_field_name}[1]", f"{json_field_name}[6]", f"{json_field_name}['a']", f"{json_field_name}['a']['b']", f"{json_field_name}['a'][0]", f"{json_field_name}['a'][6]", f"{json_field_name}['a'][0]['b']", f"{json_field_name}['a']['b']['c']", f"{json_field_name}['a']['b'][0]['d']", f"{json_field_name}[10000]", f"{json_field_name}['a']['c'][0]['d']", ] index_params.add_index(field_name=default_vector_field_name, index_type="AUTOINDEX", metric_type="COSINE") for i in range(len(json_path_list)): index_params.add_index( field_name=json_field_name, index_name=index_name + f"{i}", index_type=supported_varchar_scalar_index, params={"json_cast_type": supported_json_cast_type, "json_path": json_path_list[i]}, ) # 7. create json path index self.create_index(client, collection_name, index_params) # 8. create same json index twice self.create_index(client, collection_name, index_params) # 9. reload collection if released before to make sure the new index load successfully if is_release: self.load_collection(client, collection_name) else: # 10. sleep for 60s to make sure the new index load successfully without release and reload operations time.sleep(60) # 11. query after there is json path index under all expressions which should get the same result # with that without json path index for i in range(len(express_list)): json_list = [] id_list = [] log.info(f"query with filter {express_list[i]} after json path index is:") count = self.query( client, collection_name=collection_name, filter=express_list[i], output_fields=["count(*)"] )[0] log.info(f"The count(*) after query with filter {express_list[i]} after json path index is: {count}") res = self.query( client, collection_name=collection_name, filter=express_list[i], output_fields=[f"{json_field_name}"] )[0] for single in res: id_list.append(single[f"{default_primary_key_field_name}"]) json_list.append(single[f"{json_field_name}"]) if len(json_list) != len(compare_dict[f"{i}"]["json_list"]): log.debug(f"json field after json path index under expression {express_list[i]} is:") log.debug(json_list) log.debug(f"json field before json path index to be compared under expression {express_list[i]} is:") log.debug(compare_dict[f"{i}"]["json_list"]) assert json_list == compare_dict[f"{i}"]["json_list"] if len(id_list) != len(compare_dict[f"{i}"]["id_list"]): log.debug(f"primary key field after json path index under expression {express_list[i]} is:") log.debug(id_list) log.debug( f"primary key field before json path index to be compared under expression {express_list[i]} is:" ) log.debug(compare_dict[f"{i}"]["id_list"]) assert id_list == compare_dict[f"{i}"]["id_list"] log.info(f"PASS with expression {express_list[i]}") @pytest.mark.xdist_group("TestQueryStringNonPrimaryShared") class TestQueryStringNonPrimaryShared(TestMilvusClientV2Base): """Read-only string query tests that share one int64-primary collection.""" @pytest.fixture(scope="module", autouse=True) def prepare_query_string_non_primary_collection(self, request): client = self._client() collection_name = QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION if client.has_collection(collection_name): client.drop_collection(collection_name) schema = build_query_string_non_primary_schema(client) client.create_collection(collection_name=collection_name, schema=schema, consistency_level="Strong") rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) client.insert(collection_name=collection_name, data=rows) client.flush(collection_name=collection_name) index_params = client.prepare_index_params() index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") client.create_index(collection_name=collection_name, index_params=index_params) client.load_collection(collection_name=collection_name) global QUERY_STRING_NON_PRIMARY_SHARED_ROWS QUERY_STRING_NON_PRIMARY_SHARED_ROWS = rows def teardown(): try: if client.has_collection(QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION): client.drop_collection(QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION) except Exception: pass request.addfinalizer(teardown) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_string_is_not_primary(self): """ target: test query data with string field is not primary method: create collection and insert data, query with string expr expected: query successfully """ client = self._client() expected = [] for row in QUERY_STRING_NON_PRIMARY_SHARED_ROWS: if row[default_string_field_name] >= "0": expected.append( { default_string_field_name: row[default_string_field_name], default_primary_key_field_name: row[default_primary_key_field_name], } ) self.query( client, QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION, filter=default_search_string_exp, output_fields=[default_string_field_name], check_task=CheckTasks.check_query_results, check_items={"exp_res": expected, "pk_name": default_primary_key_field_name}, ) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_with_invalid_string_expr(self): """ target: test query data method: query with invalid string expressions expected: raise exception """ client = self._client() for expression in cf.gen_invalid_string_expressions(): error = { ct.err_code: 1100, ct.err_msg: f"failed to create query plan: cannot parse expression: {expression}", } self.query( client, QUERY_STRING_NON_PRIMARY_SHARED_COLLECTION, filter=expression, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.xdist_group("TestQueryStringPrimaryShared") class TestQueryStringPrimaryShared(TestMilvusClientV2Base): """Read-only string query tests that share one varchar-primary collection.""" @pytest.fixture(scope="module", autouse=True) def prepare_query_string_primary_collection(self, request): client = self._client() collection_name = QUERY_STRING_PRIMARY_SHARED_COLLECTION if client.has_collection(collection_name): client.drop_collection(collection_name) schema = build_query_string_primary_schema(client) client.create_collection(collection_name=collection_name, schema=schema, consistency_level="Strong") rows = gen_query_string_primary_rows() client.insert(collection_name=collection_name, data=rows) client.flush(collection_name=collection_name) index_params = client.prepare_index_params() index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") client.create_index(collection_name=collection_name, index_params=index_params) client.load_collection(collection_name=collection_name) global QUERY_STRING_PRIMARY_SHARED_ROWS QUERY_STRING_PRIMARY_SHARED_ROWS = rows def teardown(): try: if client.has_collection(QUERY_STRING_PRIMARY_SHARED_COLLECTION): client.drop_collection(QUERY_STRING_PRIMARY_SHARED_COLLECTION) except Exception: pass request.addfinalizer(teardown) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_string_with_mix_expr(self): """ target: test query data method: query with mix expr in string field and int field expected: query successfully """ client = self._client() expected = [] for row in QUERY_STRING_PRIMARY_SHARED_ROWS: if row[ct.default_int64_field_name] >= 0 and row[default_string_field_name] >= "0": expected.append( { ct.default_float_field_name: row[ct.default_float_field_name], default_string_field_name: row[default_string_field_name], } ) output_fields = [ct.default_float_field_name, default_string_field_name] self.query( client, QUERY_STRING_PRIMARY_SHARED_COLLECTION, filter=default_search_mix_exp, output_fields=output_fields, check_task=CheckTasks.check_query_results, check_items={"exp_res": expected, "pk_name": default_string_field_name}, ) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_string_with_invalid_prefix_expr(self): """ target: test query with invalid prefix string expression method: use invalid prefix string expr on float field expected: raise error """ client = self._client() expression = 'float like "0%"' error = { ct.err_code: 65535, ct.err_msg: "like operation on non-string or no-json field is unsupported", } self.query( client, QUERY_STRING_PRIMARY_SHARED_COLLECTION, filter=expression, check_task=CheckTasks.err_res, check_items=error, ) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_compare_two_fields(self): """ target: test query with bool expression comparing two fields method: specify string primary field, compare two fields expected: verify query successfully """ client = self._client() expected = [] for row in QUERY_STRING_PRIMARY_SHARED_ROWS: expected.append( { ct.default_float_field_name: row[ct.default_float_field_name], default_string_field_name: row[default_string_field_name], ct.default_int64_field_name: row[ct.default_int64_field_name], } ) expression = "float <= int64" output_fields = [ct.default_int64_field_name, ct.default_float_field_name, ct.default_string_field_name] self.query( client, QUERY_STRING_PRIMARY_SHARED_COLLECTION, filter=expression, output_fields=output_fields, check_task=CheckTasks.check_query_results, check_items={"exp_res": expected, "pk_name": ct.default_string_field_name}, ) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_compare_invalid_fields(self): """ target: test query with invalid field comparison method: compare string and int field expected: raise error """ client = self._client() expression = "varchar == int64" error = { ct.err_code: 1100, ct.err_msg: "comparisons between VarChar and Int64 are not supported", } self.query( client, QUERY_STRING_PRIMARY_SHARED_COLLECTION, filter=expression, check_task=CheckTasks.err_res, check_items=error, ) class TestQueryString(TestMilvusClientV2Base): """ ****************************************************************** The following cases are used to test query with string ****************************************************************** """ @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("index_type", ["AUTOINDEX", "BITMAP"]) def test_milvus_client_query_string_expr_with_like_index(self, index_type): """ target: test query with like string expression and indexed varchar field expected: verify query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create indexes for varchar field index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name=ct.default_string_field_name, index_type=index_type) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with string index and get baseline result count expressions = [perfix_expr, suffix_expr, inner_match_expr] indexed_result_lens = {} for expression in expressions: result = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] indexed_result_lens[expression] = len(result) # 5. drop string index and verify query result remains same self.release_collection(client, collection_name) self.drop_index(client, collection_name, ct.default_string_field_name) self.load_collection(client, collection_name) for expression in expressions: result_without_index = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] assert len(result_without_index) == indexed_result_lens[expression] # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_query_string_field_pk_is_empty(self): """ target: test query with string expr and string field is primary method: create collection, string field is primary collection load and insert empty data with string field collection query uses string expr in string field expected: query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with string primary key schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100, is_primary=True) schema.add_field(ct.default_int64_field_name, DataType.INT64) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_json_field_name, DataType.JSON) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with empty string values for varchar field rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) for row in rows: row[ct.default_string_field_name] = "" self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="FLAT", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. verify entity count assert self.get_collection_stats(client, collection_name)[0]["row_count"] == default_nb # 5. query with string expression for empty strings string_exp = 'varchar >= ""' output_fields = [ct.default_int64_field_name, ct.default_float_field_name, ct.default_string_field_name] res = self.query(client, collection_name, filter=string_exp, output_fields=output_fields)[0] assert len(res) == 1 # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("index_type", ["AUTOINDEX", "BITMAP"]) def test_milvus_client_query_string_expr_with_match_auto_index_and_bitmap(self, index_type): """ target: test query with match string expression and indexed with auto index and bitmap expected: verify query successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field( ct.default_string_field_name, DataType.VARCHAR, max_length=65535, enable_match=True, enable_analyzer=True ) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with completely manual generation for precise TEXT_MATCH testing rows = [] for i in range(default_nb): row = { default_primary_key_field_name: i, default_vector_field_name: [random.random() for _ in range(default_dim)], } if i % 10 == 0: row[ct.default_string_field_name] = "test data 0" # Contains "test" and "0" elif i % 10 == 1: row[ct.default_string_field_name] = "content with 1" # Contains "1" elif i % 10 == 2: row[ct.default_string_field_name] = "this is test content" # Contains "test" else: row[ct.default_string_field_name] = f"sample document number {i} without keywords" rows.append(row) self.insert(client, collection_name, rows) expressions = [ ('TEXT_MATCH(varchar, "0")', default_nb // 10), ('TEXT_MATCH(varchar, "1")', default_nb // 10), ('TEXT_MATCH(varchar, "test")', (default_nb // 10) * 2), ] # 3. create indexes index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") index_params.add_index( field_name=ct.default_string_field_name, index_type=index_type, params={"tokenizer": "standard"} ) self.create_index(client, collection_name, index_params) # Wait for string field index to be ready before loading self.load_collection(client, collection_name) time.sleep(ct.default_graceful_time) # 4. query indexed_result_lens = {} for expression, exp_len in expressions: result = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] assert len(result) == exp_len indexed_result_lens[expression] = len(result) # 5. drop index and verify query result remains same self.release_collection(client, collection_name) self.drop_index(client, collection_name, ct.default_string_field_name) self.load_collection(client, collection_name) time.sleep(ct.default_graceful_time) for expression, exp_len in expressions: result_without_index = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] assert len(result_without_index) == indexed_result_lens[expression] == exp_len # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_bitmap_alter_offset_cache_param(self): """ target: test bitmap index with enable offset cache. expected: verify create index and load successfully """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(default_primary_key_field_name, DataType.INT64, is_primary=True) schema.add_field(default_vector_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_string_field_name, DataType.VARCHAR, max_length=100) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) self.insert(client, collection_name, rows) # 3. create indexes index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") index_params.add_index(field_name=ct.default_string_field_name, index_type="BITMAP") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with expression and get baseline result count expression = 'varchar like "0%"' result = self.query(client, collection_name, filter=expression, output_fields=[ct.default_string_field_name])[0] res_len = len(result) # 5. enable offset cache and verify result count remains same self.release_collection(client, collection_name) self.alter_index_properties( client, collection_name, ct.default_string_field_name, properties={"indexoffsetcache.enabled": True} ) self.load_collection(client, collection_name) result_with_cache = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] res_len_with_cache = len(result_with_cache) assert res_len_with_cache == res_len # 6. disable offset cache and verify result count remains same self.release_collection(client, collection_name) self.alter_index_properties( client, collection_name, ct.default_string_field_name, properties={"indexoffsetcache.enabled": False} ) self.load_collection(client, collection_name) result_without_cache = self.query( client, collection_name, filter=expression, output_fields=[ct.default_string_field_name] )[0] res_len_without_cache = len(result_without_cache) assert res_len_without_cache == res_len # 7. clean up self.drop_collection(client, collection_name) class TestQueryArray(TestMilvusClientV2Base): """ ****************************************************************** The following cases are used to test query with array ****************************************************************** """ @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("array_element_data_type", [DataType.INT64]) def test_milvus_client_query_array_with_inverted_index(self, array_element_data_type): """ target: test query with array fields and inverted index method: create collection with array fields, create inverted index, query with array expressions expected: verify query result """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with array fields additional_params = {"max_length": 1000} if array_element_data_type == DataType.VARCHAR else {} schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field("id", DataType.INT64, is_primary=True) schema.add_field( "contains", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params ) schema.add_field( "contains_any", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params ) schema.add_field( "contains_all", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params ) schema.add_field( "equals", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params ) schema.add_field( "array_length_field", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params, ) schema.add_field( "array_access", DataType.ARRAY, element_type=array_element_data_type, max_capacity=2000, **additional_params ) schema.add_field("emb", DataType.FLOAT_VECTOR, dim=128) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data train_data, query_expr = cf.prepare_array_test_data(3000, hit_rate=0.05) rows = train_data.to_dict("records") self.insert(client, collection_name, rows) # 3. create indexes index_params = self.prepare_index_params(client)[0] index_params.add_index( field_name="emb", index_type="HNSW", metric_type="L2", params={"M": 48, "efConstruction": 500} ) # Add INVERTED index for array fields for field in ["contains", "contains_any", "contains_all", "equals", "array_length_field", "array_access"]: index_params.add_index(field_name=field, index_type="INVERTED") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with array expressions and verify results for item in query_expr: expr = item["expr"] ground_truth = item["ground_truth"] res = self.query(client, collection_name, filter=expr, output_fields=["*"])[0] assert len(res) == len(ground_truth) # Verify the returned IDs match the ground truth returned_ids = [record["id"] for record in res] assert set(returned_ids) == set(ground_truth) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("use_index", [True, False]) @pytest.mark.parametrize("index_type", ["INVERTED", "BITMAP"]) def test_milvus_client_query_array_with_prefix_like(self, use_index, index_type): """ target: test query with array element using different like patterns method: create collection with string array field, query with various like patterns expected: verify query result for prefix, suffix, and inner match patterns """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with array fields schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_json_field_name, DataType.JSON, nullable=True) schema.add_field( ct.default_int32_array_field_name, DataType.ARRAY, element_type=DataType.INT32, max_capacity=ct.default_max_capacity, ) schema.add_field( ct.default_float_array_field_name, DataType.ARRAY, element_type=DataType.FLOAT, max_capacity=ct.default_max_capacity, ) schema.add_field( ct.default_string_array_field_name, DataType.ARRAY, element_type=DataType.VARCHAR, max_capacity=ct.default_max_capacity, max_length=100, nullable=True, ) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data # Generate string array data with specific pattern for testing string_field_value = [[str(j) for j in range(i, i + 3)] for i in range(default_nb)] # Convert array data to rows format rows = [] for i in range(default_nb): row = { ct.default_int64_field_name: i, ct.default_float_vec_field_name: [random.random() for _ in range(default_dim)], ct.default_json_field_name: {"number": i, "float": float(i)}, ct.default_int32_array_field_name: [np.int32(j) for j in range(i, i + ct.default_max_capacity)], ct.default_float_array_field_name: [np.float32(j) for j in range(i, i + ct.default_max_capacity)], ct.default_string_array_field_name: string_field_value[i], } rows.append(row) self.insert(client, collection_name, rows) # 3. create indexes index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") if use_index: index_params.add_index(field_name=ct.default_string_array_field_name, index_type=index_type) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with array element prefix like array_expressions = ['string_array[0] like "0%"', 'string_array[0] like "%0"', 'string_array[0] like "%0%"'] for array_expression in array_expressions: res = self.query(client, collection_name, filter=array_expression, limit=default_nb)[0] # 5. verify result based on expression type if 'like "0%"' in array_expression: # Prefix match: first element starts with "0" filter_data = [row for row in string_field_value if row[0].startswith("0")] def match_func(x): return x.startswith("0") elif 'like "%0"' in array_expression: # Suffix match: first element ends with "0" filter_data = [row for row in string_field_value if row[0].endswith("0")] def match_func(x): return x.endswith("0") elif 'like "%0%"' in array_expression: # Inner match: first element contains "0" filter_data = [row for row in string_field_value if "0" in row[0]] def match_func(x): return "0" in x assert len(res) == len(filter_data) # Verify the returned records have correct first elements for record in res: first_element = record[ct.default_string_array_field_name][0] assert match_func(first_element) self.drop_collection(client, collection_name) @pytest.mark.xdist_group("TestQueryCountDefaultShared") class TestQueryCountDefaultShared(TestMilvusClientV2Base): """Read-only count(*) tests that share one default-schema collection.""" @pytest.fixture(scope="module", autouse=True) def prepare_query_count_default_collection(self, request): client = self._client() collection_name = QUERY_COUNT_DEFAULT_SHARED_COLLECTION if client.has_collection(collection_name): client.drop_collection(collection_name) schema = build_query_count_default_schema(client) client.create_collection(collection_name=collection_name, schema=schema, consistency_level="Strong") rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) client.insert(collection_name=collection_name, data=rows) client.flush(collection_name=collection_name) index_params = client.prepare_index_params() index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") client.create_index(collection_name=collection_name, index_params=index_params) client.load_collection(collection_name=collection_name) def teardown(): try: if client.has_collection(QUERY_COUNT_DEFAULT_SHARED_COLLECTION): client.drop_collection(QUERY_COUNT_DEFAULT_SHARED_COLLECTION) except Exception: pass request.addfinalizer(teardown) @pytest.mark.tags(CaseLabel.L0) def test_milvus_client_count_with_expr(self): """ target: test count with expr method: count with expr expected: verify count """ client = self._client() self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb}], "pk_name": default_primary_key_field_name}, ) self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter="id in [0, 1]", output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 2}], "pk_name": default_primary_key_field_name}, ) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_invalid_output_field(self): """ target: test count with invalid output field method: query with invalid count output field expected: raise exception with field not exist error """ client = self._client() invalid_output_fields = ["count", "count(int64)", "count(**)"] for invalid_output_field in invalid_output_fields: if invalid_output_field == "count": err_msg = "field count not exist" else: err_msg = "for aggregation:count does not exist" self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter=default_search_exp, output_fields=[invalid_output_field], check_task=CheckTasks.err_res, check_items={"err_code": 1, "err_msg": err_msg}, ) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_with_pagination_param(self): """ target: test count with pagination params method: count with pagination params: offset, limit expected: exception """ client = self._client() res = self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter=default_search_exp, output_fields=["count(*)"], offset=10, )[0] assert res[0]["count(*)"] == default_nb self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter=default_search_exp, output_fields=["count(*)"], limit=10, check_task=CheckTasks.err_res, check_items={ct.err_code: 1, ct.err_msg: "count entities with pagination is not allowed"}, ) self.query( client, QUERY_COUNT_DEFAULT_SHARED_COLLECTION, filter=default_search_exp, output_fields=["count(*)"], offset=10, limit=10, check_task=CheckTasks.err_res, check_items={ct.err_code: 1, ct.err_msg: "count entities with pagination is not allowed"}, ) @pytest.mark.xdist_group("TestQueryCountExpressionShared") class TestQueryCountExpressionShared(TestMilvusClientV2Base): """Read-only count(*) expression tests that share one superset-schema collection.""" @pytest.fixture(scope="module", autouse=True) def prepare_query_count_expression_collection(self, request): client = self._client() collection_name = QUERY_COUNT_EXPR_SHARED_COLLECTION if client.has_collection(collection_name): client.drop_collection(collection_name) schema = build_query_count_expression_schema(client) client.create_collection(collection_name=collection_name, schema=schema, consistency_level="Strong") rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema) client.insert(collection_name=collection_name, data=rows) client.flush(collection_name=collection_name) index_params = client.prepare_index_params() index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") client.create_index(collection_name=collection_name, index_params=index_params) client.load_collection(collection_name=collection_name) global QUERY_COUNT_EXPR_SHARED_ROWS QUERY_COUNT_EXPR_SHARED_ROWS = rows def teardown(): try: if client.has_collection(QUERY_COUNT_EXPR_SHARED_COLLECTION): client.drop_collection(QUERY_COUNT_EXPR_SHARED_COLLECTION) except Exception: pass request.addfinalizer(teardown) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_expressions(self): """ target: test count with expr method: count with expr expected: verify count """ client = self._client() rows = QUERY_COUNT_EXPR_SHARED_ROWS for expressions in cf.gen_normal_expressions_and_templates(): log.debug(f"query with expression: {expressions}") expr = expressions[0].replace("&&", "and").replace("||", "or") filter_ids = [] for row in rows: int64 = row[ct.default_int64_field_name] # noqa: F841 - referenced by eval(expr) float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) if not expr or eval(expr): filter_ids.append(row[ct.default_int64_field_name]) expected_count = len(filter_ids) self.query( client, QUERY_COUNT_EXPR_SHARED_COLLECTION, filter=expr, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": expected_count}], "pk_name": ct.default_int64_field_name}, ) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_bool_expressions(self): """ target: test count with binary expr method: count with binary expr expected: verify count """ client = self._client() rows = QUERY_COUNT_EXPR_SHARED_ROWS for bool_type in [True, False, "true", "false"]: bool_type_cmp = bool_type if bool_type == "true": bool_type_cmp = True if bool_type == "false": bool_type_cmp = False filter_ids = [] for row in rows: if row[ct.default_bool_field_name] == bool_type_cmp: filter_ids.append(row[ct.default_int64_field_name]) expected_count = len(filter_ids) expression = f"{ct.default_bool_field_name} == {bool_type}" self.query( client, QUERY_COUNT_EXPR_SHARED_COLLECTION, filter=expression, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": expected_count}], "pk_name": ct.default_int64_field_name}, ) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_expression_auto_field(self): """ target: test count with expr method: count with expr expected: verify count """ client = self._client() rows = QUERY_COUNT_EXPR_SHARED_ROWS for expressions in cf.gen_normal_expressions_and_templates_field(ct.default_float_field_name): log.debug(f"query with expression: {expressions}") expr = expressions[0].replace("&&", "and").replace("||", "or") filter_ids = [] for row in rows: float = row[ct.default_float_field_name] # noqa: F841 - referenced by eval(expr) if not expr or eval(expr): filter_ids.append(row[ct.default_int64_field_name]) expected_count = len(filter_ids) expr = cf.get_expr_from_template(expressions[1]).replace("&&", "and").replace("||", "or") expr_params = cf.get_expr_params_from_template(expressions[1]) self.query( client, QUERY_COUNT_EXPR_SHARED_COLLECTION, filter=expr, filter_params=expr_params, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": expected_count}], "pk_name": ct.default_int64_field_name}, ) class TestQueryCount(TestMilvusClientV2Base): """ test query count(*) """ @pytest.mark.tags(CaseLabel.L1) @pytest.mark.parametrize("consistency_level", ["Bounded", "Strong", "Eventually"]) def test_milvus_client_count_consistency_level(self, consistency_level): """ target: test count(*) with different consistency levels method: 1. create collection with different consistency level 2. load collection 3. insert and count 4. verify count expected: expected count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with specified consistency level self.create_collection(client, collection_name, default_dim, consistency_level=consistency_level) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 3. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 4. wait based on consistency level if consistency_level == "Bounded": time.sleep(ct.default_graceful_time) elif consistency_level == "Strong": pass elif consistency_level == "Eventually": time.sleep(ct.default_graceful_time) # 5. query with count(*) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb}], "pk_name": ct.default_int64_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_duplicate_ids(self): """ target: test count duplicate ids method: 1. insert duplicate ids 2. count 3. delete duplicate ids 4. count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. create index and load collection index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 3. insert duplicate ids data tmp_nb = 100 schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info) # Set all primary keys to 0 (duplicate ids) for row in rows: row[default_primary_key_field_name] = 0 self.insert(client, collection_name, rows) # 4. query count self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb}], "pk_name": default_primary_key_field_name}, ) # 5. delete and verify count self.delete(client, collection_name, filter="id == 0") self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 0}], "pk_name": default_primary_key_field_name}, ) # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_multi_partitions(self): """ target: test count multi partitions method: 1. init partitions: p1, _default 2. count p1, _default, [p1, _default] 3. delete _default entities and count _default, [p1, _default] 4. drop p1 and count p1, [p1, _default] expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = cf.gen_unique_str("partition") half = default_nb // 2 # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert data into two partitions schema_info = self.describe_collection(client, collection_name)[0] # insert [0, half) into custom partition rows_partition = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=0) self.insert(client, collection_name, rows_partition, partition_name=partition_name) # insert [half, nb) into _default partition rows_default = cf.gen_row_data_by_schema(nb=half, schema=schema_info, start=half) self.insert(client, collection_name, rows_default) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_partitions(client, collection_name, [partition_name, "_default"]) # 4. query count for each partition for p_name in [partition_name, "_default"]: self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=[p_name], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": half}], "pk_name": default_primary_key_field_name}, )[0] # 5. delete entities from _default partition delete_expr = f"{default_primary_key_field_name} >= {half}" self.delete(client, collection_name, filter=delete_expr) # count _default partition after deletion self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=["_default"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 0}], "pk_name": default_primary_key_field_name}, )[0] # count both partitions after deletion self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=[partition_name, "_default"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": half}], "pk_name": default_primary_key_field_name}, )[0] # 6. drop partition and test error self.release_partitions(client, collection_name, [partition_name]) self.drop_partition(client, collection_name, partition_name) # query dropped partition should fail self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=[partition_name], check_task=CheckTasks.err_res, check_items={"err_code": 65535, "err_msg": f"partition name {partition_name} not found"}, ) # count remaining _default partition self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=["_default"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 0}], "pk_name": default_primary_key_field_name}, )[0] # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_partition_duplicate(self): """ target: test count from partitions which have duplicate ids method: 1. insert same ids into 2 partitions 2. count 3. delete some ids and count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() partition_name = "p1" # 1. create collection and partition self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.create_partition(client, collection_name, partition_name) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert same data into both partitions (duplicate ids) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows, partition_name="_default") self.insert(client, collection_name, rows, partition_name=partition_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. count total entities (should be default_nb * 2 due to duplicates) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb * 2}], "pk_name": default_primary_key_field_name}, )[0] # 5. delete some duplicate ids self.delete(client, collection_name, filter="id in [0, 1]") # 6. count remaining entities in partition p1 self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], partition_names=[partition_name], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb - 2}], "pk_name": default_primary_key_field_name}, )[0] # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_growing_sealed_segment(self): """ target: test count growing and sealed segment method: 1. insert -> index -> load 2. count 3. new insert 4. count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() tmp_nb = 100 # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert initial data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. create index and load (this creates sealed segment) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. count sealed segment data self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb}], "pk_name": default_primary_key_field_name}, )[0] # 5. new insert (this creates growing segment) new_rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info, start=tmp_nb) self.insert(client, collection_name, new_rows) # 6. count both sealed and growing segment data self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb * 2}], "pk_name": default_primary_key_field_name}, )[0] # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_during_handoff(self): """ target: test count during handoff method: 1. index -> load 2. insert 3. flush while count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection -> index -> load self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. insert data schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) # 3. flush while count (concurrent operations) def flush_collection(): self.flush(client, collection_name) def count_entities(): self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb}], "pk_name": default_primary_key_field_name}, ) t_flush = threading.Thread(target=flush_collection) t_count = threading.Thread(target=count_entities) t_flush.start() t_count.start() t_flush.join() t_count.join() # 4. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_delete_insert_duplicate_ids(self): """ target: test count after delete and re-insert same entities method: 1. insert and delete 2. count 3. re-insert deleted ids with different vectors 4. count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() tmp_nb = 100 # 1. create collection self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. insert initial data (sealed segment) [0, default_nb) schema_info = self.describe_collection(client, collection_name)[0] rows_initial = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows_initial) # 3. create index and load (make data into sealed segment) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. insert additional data (growing segment) [default_nb, default_nb + tmp_nb) rows_additional = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info, start=default_nb) self.insert(client, collection_name, rows_additional) # 5. delete sealed segment data [0, default_nb) -> count delete_expr = f"{default_primary_key_field_name} in {[i for i in range(default_nb)]}" self.delete(client, collection_name, filter=delete_expr) # Count should show only remaining growing segment data (tmp_nb records) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb}], "pk_name": default_primary_key_field_name}, ) # 6. re-insert deleted ids [0, default_nb) with different vectors rows_reinsert = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows_reinsert) # Count should show all data: tmp_nb (growing) + default_nb (re-inserted) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb + tmp_nb}], "pk_name": default_primary_key_field_name}, ) # 7. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_during_compact(self): """ target: test count during compact merge many small segments method: 1. init many small segments 2. compact while count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() tmp_nb = 100 # 1. create collection with one shard self.create_collection(client, collection_name, default_dim, consistency_level="Strong", shards_num=1) self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) # 2. init 10 small segments for i in range(10): schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info, start=i * tmp_nb) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. compact while count using threading def compact_collection(): self.compact(client, collection_name) def count_entities(): self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb * 10}], "pk_name": default_primary_key_field_name}, ) t_compact = threading.Thread(target=compact_collection) t_count = threading.Thread(target=count_entities) t_compact.start() t_count.start() t_count.join() t_compact.join() # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_query_count_expr_json(self): """ target: test query with part json key value method: 1. insert data and some entities doesn't have number key 2. query count with number expr filter expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with dynamic field enabled and json field schema = self.create_schema(client, enable_dynamic_field=True, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) schema.add_field(ct.default_json_field_name, DataType.JSON) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with partial json fields rows = cf.gen_default_rows_data(with_json=False) for i in range(default_nb): # Only odd numbered entities have the "number" field in json if i % 2 == 0: rows[i][ct.default_json_field_name] = {"string": str(i), "bool": bool(i)} else: rows[i][ct.default_json_field_name] = {"string": str(i), "bool": bool(i), "number": i} self.insert(client, collection_name, rows) time.sleep(0.4) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query count with json number field filter expression = f'{ct.default_json_field_name}["number"] < 100' self.query( client, collection_name, filter=expression, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 50}], "pk_name": default_primary_key_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_json_expr_on_search_n_query(self): """ target: verify more expressions of json object, json array and json texts are supported in search and query method: 1. insert data with vectors and different json format 2. verify insert successfully 3. build index and load 4. search and query with different expressions 5. verify search and query successfully expected: succeed """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with multiple json fields json_int = "json_int" json_float = "json_float" json_string = "json_string" json_bool = "json_bool" json_array = "json_array" json_embedded_object = "json_embedded_object" json_objects_array = "json_objects_array" dim = 16 schema = self.create_schema(client, enable_dynamic_field=False, auto_id=True)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=dim) schema.add_field(json_int, DataType.JSON) schema.add_field(json_float, DataType.JSON) schema.add_field(json_string, DataType.JSON) schema.add_field(json_bool, DataType.JSON) schema.add_field(json_array, DataType.JSON) schema.add_field(json_embedded_object, DataType.JSON) schema.add_field(json_objects_array, DataType.JSON) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. insert data with different json types nb = 1000 for i in range(10): # Generate vectors vectors = cf.gen_vectors(nb, dim) # Generate JSON data for each field using the same function as original test json_int_data = cf.gen_json_data_for_diff_json_types(nb=nb, start=i * nb, json_type=json_int) json_float_data = cf.gen_json_data_for_diff_json_types(nb=nb, start=i * nb, json_type=json_float) json_string_data = cf.gen_json_data_for_diff_json_types(nb=nb, start=i * nb, json_type=json_string) json_bool_data = cf.gen_json_data_for_diff_json_types(nb=nb, start=i * nb, json_type=json_bool) json_array_data = cf.gen_json_data_for_diff_json_types(nb=nb, start=i * nb, json_type=json_array) json_embedded_object_data = cf.gen_json_data_for_diff_json_types( nb=nb, start=i * nb, json_type=json_embedded_object ) json_objects_array_data = cf.gen_json_data_for_diff_json_types( nb=nb, start=i * nb, json_type=json_objects_array ) # Convert to rows format for MilvusClient V2 rows = [] for j in range(nb): row = { ct.default_float_vec_field_name: vectors[j], json_int: json_int_data[j], json_float: json_float_data[j], json_string: json_string_data[j], json_bool: json_bool_data[j], json_array: json_array_data[j], json_embedded_object: json_embedded_object_data[j], json_objects_array: json_objects_array_data[j], } rows.append(row) self.insert(client, collection_name, rows) time.sleep(0.4) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index( field_name=ct.default_float_vec_field_name, index_type="IVF_SQ8", metric_type="L2", params={"nlist": 64} ) self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. query with different json expressions - each should return 10 results query_exprs = [ f'json_contains_any({json_embedded_object}["{json_embedded_object}"]["level2"]["level2_array"], [1,3,5,7,9])', f'json_contains_any({json_embedded_object}["array"], [1,3,5,7,9])', f"{json_int} < 10", f"{json_float} <= 200.0 and {json_float} > 190.0", f'{json_string} in ["1","2","3","4","5","6","7","8","9","10"]', f"{json_bool} == true and {json_float} <= 10", f"{json_array} == [4001,4002,4003,4004,4005,4006,4007,4008,4009,4010] or {json_int} < 9", f'{json_embedded_object}["{json_embedded_object}"]["number"] < 10', f'{json_objects_array}[0]["level2"]["level2_str"] like "199%" and {json_objects_array}[1]["float"] >= 1990', ] search_vectors = cf.gen_vectors(2, dim) for expr in query_exprs: log.debug(f"query_expr: {expr}") # Query test res = self.query(client, collection_name, filter=expr, output_fields=["count(*)"]) assert res[0][0]["count(*)"] == 10, f"Query failed for expr: {expr}, got {res[0][0]['count(*)']}" # Search test search_res = self.search( client, collection_name, search_vectors, limit=10, filter=expr, output_fields=None, search_params={} ) assert len(search_res[0]) == 2, f"Search nq failed for expr: {expr}" for hits in search_res[0]: assert len(hits) == 10, f"Search limit failed for expr: {expr}" # 5. verify edge cases for issue #36718 edge_case_exprs = [ f'{json_embedded_object}["{json_embedded_object}"]["number"] in []', f'{json_embedded_object}["{json_embedded_object}"] in []', ] for expr in edge_case_exprs: log.debug(f"edge_case_expr: {expr}") # Query test - should return 0 results res = self.query(client, collection_name, filter=expr, output_fields=["count(*)"]) assert res[0][0]["count(*)"] == 0, f"Edge case query failed for expr: {expr}" # Search test - should return 0 results search_res = self.search( client, collection_name, search_vectors, limit=10, filter=expr, output_fields=None, search_params={} ) assert len(search_res[0]) == 2, f"Edge case search nq failed for expr: {expr}" for hits in search_res[0]: assert len(hits) == 0, f"Edge case search should return 0 results for expr: {expr}" # 6. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_alias_insert_delete_drop(self): """ target: test count after alias insert and load method: 1. init collection 2. alias insert more entities 3. count and alias count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() alias = cf.gen_unique_str("alias") # 1. create -> insert -> index -> load self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. create alias self.create_alias(client, collection_name, alias) # 3. new insert via alias - insert more entities additional_rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info, start=default_nb) self.insert(client, alias, additional_rows) # 4. count via alias - should see both original and new entities res = self.query(client, alias, filter=default_search_exp, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == default_nb * 2 # 5. delete via alias and count delete_expr = f"{default_primary_key_field_name} in {[i for i in range(default_nb)]}" self.delete(client, alias, filter=delete_expr) res = self.query(client, alias, filter=default_search_exp, output_fields=["count(*)"])[0] assert res[0]["count(*)"] == default_nb # 6. try to drop collection via alias (should fail) self.drop_collection( client, alias, check_task=CheckTasks.err_res, check_items={ct.err_code: 1, ct.err_msg: "cannot drop the collection via alias"}, ) # 7. clean up - drop alias and collection self.drop_alias(client, alias) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) @pytest.mark.parametrize("is_growing", [True, False]) def test_milvus_client_count_upsert_growing_sealed(self, is_growing): """ target: test count after upsert growing method: 1. create -> index -> load -> insert -> delete 2. upsert deleted id and count (+1) 3. upsert new id and count (+1) 4. upsert existed id and count (+0) expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # create collection and prepare data self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) if is_growing: # create -> index -> load -> insert -> delete (growing segments) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="FLAT", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) self.insert(client, collection_name, rows) # no flush - data stays in growing segments else: # create -> insert -> flush -> index -> load (sealed segments) self.insert(client, collection_name, rows) self.flush(client, collection_name) # flush to create sealed segments index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="FLAT", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # delete one entity (works for both growing and sealed) single_expr = f"{default_primary_key_field_name} in [0]" self.delete(client, collection_name, filter=single_expr) # upsert deleted id upsert_rows = cf.gen_row_data_by_schema(nb=1, schema=schema_info) # Ensure the primary key is 0 (the deleted id) upsert_rows[0][default_primary_key_field_name] = 0 self.upsert(client, collection_name, upsert_rows) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": ct.default_nb}], "pk_name": default_primary_key_field_name}, ) # upsert new id and count new_upsert_rows = cf.gen_row_data_by_schema(nb=1, schema=schema_info, start=default_nb) self.upsert(client, collection_name, new_upsert_rows) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": ct.default_nb + 1}], "pk_name": default_primary_key_field_name}, ) # upsert existed id and count existed_upsert_rows = cf.gen_row_data_by_schema(nb=1, schema=schema_info, start=10) self.upsert(client, collection_name, existed_upsert_rows) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": ct.default_nb + 1}], "pk_name": default_primary_key_field_name}, ) # clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_upsert_duplicate(self): """ target: test count after upsert duplicate method: 1. insert many duplicate ids 2. upsert id and count 3. delete id and count 4. upsert deleted id and count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. init collection and insert same ids tmp_nb = 100 self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) schema_info = self.describe_collection(client, collection_name)[0] # Generate data with all primary keys set to 0 (duplicate IDs) rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info) for row in rows: row[default_primary_key_field_name] = 0 self.insert(client, collection_name, rows) self.flush(client, collection_name) # Create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="FLAT", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. upsert id and count upsert_rows = cf.gen_row_data_by_schema(nb=tmp_nb, schema=schema_info, start=0) self.upsert(client, collection_name, upsert_rows) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb}], "pk_name": default_primary_key_field_name}, ) # 3. delete id and count self.delete(client, collection_name, filter="id in [0, 1]") delete_count = len([0, 1]) # delete_res.delete_count equivalent self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb - delete_count}], "pk_name": default_primary_key_field_name}, ) # 4. upsert deleted id and count deleted_upsert_rows = cf.gen_row_data_by_schema(nb=delete_count, schema=schema_info, start=0) self.upsert(client, collection_name, deleted_upsert_rows) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": tmp_nb}], "pk_name": default_primary_key_field_name}, ) # 5. clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_rename_collection(self): """ target: test count after rename collection method: 1. create -> insert -> index -> load 2. rename collection 3. count expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create -> insert -> index -> load self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=default_nb, schema=schema_info) self.insert(client, collection_name, rows) self.flush(client, collection_name) # Create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. rename collection new_name = cf.gen_unique_str("new_name") self.rename_collection(client, collection_name, new_name) # 3. count with new collection name self.query( client, new_name, filter=default_search_exp, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": default_nb}], "pk_name": default_primary_key_field_name}, ) # clean up self.drop_collection(client, new_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_ignore_growing(self): """ target: test count when disable growing segments method: 1. create -> index -> load -> insert 2. query count with ignore_growing expected: verify count 0 """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create -> index -> load self.create_collection(client, collection_name, default_dim, consistency_level="Strong") self.release_collection(client, collection_name) self.drop_index(client, collection_name, default_vector_field_name) index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=default_vector_field_name, index_type="FLAT", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 2. insert data after loading (will be in growing segment) schema_info = self.describe_collection(client, collection_name)[0] rows = cf.gen_row_data_by_schema(nb=100, schema=schema_info) self.insert(client, collection_name, rows) # 3. query count with ignore_growing=True (should return 0) self.query( client, collection_name, filter=default_search_exp, output_fields=["count(*)"], ignore_growing=True, check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 0}], "pk_name": default_primary_key_field_name}, ) # clean up self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L2) def test_milvus_client_count_expression_all_datatype(self): """ target: test count with expr method: count with expr expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with all data types schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field(ct.default_int64_field_name, DataType.INT64, is_primary=True) schema.add_field(ct.default_int32_field_name, DataType.INT32) schema.add_field(ct.default_int16_field_name, DataType.INT16) schema.add_field(ct.default_int8_field_name, DataType.INT8) schema.add_field(ct.default_float_field_name, DataType.FLOAT) schema.add_field(ct.default_double_field_name, DataType.DOUBLE) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. manually generate data to match the expression requirements rows = [] vectors = cf.gen_vectors(default_nb, default_dim) matched_id = default_nb // 2 for i in range(default_nb): row = { ct.default_int64_field_name: np.int64(i), ct.default_int32_field_name: np.int32(i), ct.default_int16_field_name: np.int16(i), ct.default_int8_field_name: 0 if i == matched_id else 1, ct.default_float_field_name: np.float32(i), ct.default_double_field_name: np.float64(i), ct.default_float_vec_field_name: vectors[i], } rows.append(row) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. count with expr using all data types expr = ( f"int64 >= 0 and int32 >= {matched_id} and int16 >= 0 and int8 <= 0 " f"and float <= {float(matched_id)} and double >= 0" ) self.query( client, collection_name, filter=expr, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": 1}], "pk_name": ct.default_int64_field_name}, ) self.drop_collection(client, collection_name) @pytest.mark.tags(CaseLabel.L1) def test_milvus_client_count_expression_comparative(self): """ target: test count with expr method: count with expr expected: verify count """ client = self._client() collection_name = cf.gen_collection_name_by_testcase_name() # 1. create collection with two int64 fields for comparison schema = self.create_schema(client, enable_dynamic_field=False, auto_id=False)[0] schema.add_field("int64_1", DataType.INT64, is_primary=True) schema.add_field("int64_2", DataType.INT64) schema.add_field(ct.default_float_vec_field_name, DataType.FLOAT_VECTOR, dim=default_dim) self.create_collection(client, collection_name, schema=schema, consistency_level="Strong") # 2. generate data nb = 10 int_values = [random.randint(0, nb) for _ in range(nb)] rows = [] vectors = cf.gen_vectors(nb, default_dim) for i in range(nb): row = { "int64_1": i, # primary key: 0, 1, 2, ..., 9 "int64_2": int_values[i], # random values from 0 to nb ct.default_float_vec_field_name: vectors[i], } rows.append(row) self.insert(client, collection_name, rows) self.flush(client, collection_name) # 3. create index and load index_params = self.prepare_index_params(client)[0] index_params.add_index(field_name=ct.default_float_vec_field_name, index_type="HNSW", metric_type="L2") self.create_index(client, collection_name, index_params) self.load_collection(client, collection_name) # 4. calculate expected result manually res = 0 for i in range(nb): if i >= int_values[i]: res += 1 # 5. count with comparative expression expression = "int64_1 >= int64_2" self.query( client, collection_name, filter=expression, output_fields=["count(*)"], check_task=CheckTasks.check_query_results, check_items={exp_res: [{"count(*)": res}], "pk_name": "int64_1"}, ) # clean up self.drop_collection(client, collection_name)