milvus-io
/
milvus
mirror of https://github.com/milvus-io/milvus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix: Fix using json index when iterative_filter is specified (#40946) 

issue: #40934 
pr: #40945

Signed-off-by: sunby <sunbingyi1992@gmail.com>
pull/40939/head
Bingyi Sun 2025-03-27 23:02:20 +08:00 committed by GitHub
parent b64bb63e77
commit f0346a149a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 78 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
internal/core/src/exec/expression/TermExpr.cpp
View File

@ -236,7 +236,7 @@ PhyTermFilterExpr::ExecVisitorImplTemplateJson(EvalCtx& context) {
if (expr_->is_in_field_) {
return ExecTermJsonVariableInField<ValueType>(context);
} else {
if (is_index_mode_) {
if (is_index_mode_ && !has_offset_input_) {
// we create double index for json int64 field for now
using GetType =
std::conditional_t<std::is_same_v<ValueType, int64_t>,

77
tests/python_client/testcases/test_search.py
View File

@ -3472,6 +3472,83 @@ class TestCollectionSearch(TestcaseBase):
for hits in search_res:
ids = hits.ids
assert set(ids).issubset(filter_ids_set)
# 6. search again with expression template and hint
search_params = default_search_params.copy()
search_params.update({"hints": "iterative_filter"})
search_res, _ = collection_w.search(search_vectors[:default_nq], default_search_field,
search_params,
limit=nb, expr=expr, expr_params=expr_params,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids))})
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
log.info("binbin2")
log.info(ids)
log.info(filter_ids_set)
assert set(ids).issubset(filter_ids_set)
# 7. create json index
default_json_path_index = {"index_type": "INVERTED", "params": {"json_cast_type": "double",
"json_path": f"{ct.default_json_field_name}['number']"}}
collection_w.create_index(ct.default_json_field_name, default_json_path_index, index_name = f"{ct.default_json_field_name}_0")
default_json_path_index = {"index_type": "INVERTED", "params": {"json_cast_type": "double",
"json_path": f"{ct.default_json_field_name}['float']"}}
collection_w.create_index(ct.default_json_field_name, default_json_path_index, index_name = f"{ct.default_json_field_name}_1")
# 8. release and load to make sure the new index is loaded
collection_w.release()
collection_w.load()
# 9. search expressions after json path index
expr = expressions[0].replace("&&", "and").replace("||", "or")
search_res, _ = collection_w.search(search_vectors[:default_nq], default_search_field,
default_search_params,
limit=nb, expr=expr,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids))})
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
assert set(ids).issubset(filter_ids_set)
# 10. search again with expression template after json path index
expr = cf.get_expr_from_template(expressions[1]).replace("&&", "and").replace("||", "or")
expr_params = cf.get_expr_params_from_template(expressions[1])
search_res, _ = collection_w.search(search_vectors[:default_nq], default_search_field,
default_search_params,
limit=nb, expr=expr, expr_params=expr_params,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids))})
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
assert set(ids).issubset(filter_ids_set)
# 11. search again with expression template and hint after json path index
search_params = default_search_params.copy()
search_params.update({"hints": "iterative_filter"})
log.info("binbin")
log.info(expr)
log.info(expr_params)
log.info(search_params)
search_res, _ = collection_w.search(search_vectors[:default_nq], default_search_field,
search_params,
limit=nb, expr=expr, expr_params=expr_params,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids))})
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
log.info(ids)
log.info(filter_ids_set)
assert set(ids).issubset(filter_ids_set)
@pytest.mark.tags(CaseLabel.L2)
def test_search_expression_all_data_type(self, nq, _async, null_data_percent):