Skip to content

Commit

Permalink
test: Add sparse invert index algo check tests (#39691)
Browse files Browse the repository at this point in the history 
related issue: #39332

---------

Signed-off-by: yanliang567 <[email protected]>
  • Loading branch information
@yanliang567
yanliang567 authored Feb 12, 2025
1 parent c12c4b4 commit 5fdc757
Show file tree
Hide file tree
Showing 5 changed files with 87 additions and 40 deletions.
4 changes: 3 additions & 1 deletion tests/python_client/common/common_type.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -220,10 +220,12 @@
"SPARSE_INVERTED_INDEX", "SPARSE_WAND",
"GPU_IVF_FLAT", "GPU_IVF_PQ"]

inverted_index_algo = ['TAAT_NAIVE', 'DAAT_WAND', 'DAAT_MAXSCORE']

default_all_indexes_params = [{}, {"nlist": 128}, {"nlist": 128}, {"nlist": 128, "m": 16, "nbits": 8},
{"M": 32, "efConstruction": 360}, {"nlist": 128}, {},
{}, {"nlist": 64},
{"drop_ratio_build": 0.2}, {"drop_ratio_build": 0.2},
{}, {"drop_ratio_build": 0.2},
{"nlist": 64}, {"nlist": 64, "m": 16, "nbits": 8}]

default_all_search_params_params = [{}, {"nprobe": 32}, {"nprobe": 32}, {"nprobe": 32},
Expand Down
10 changes: 7 additions & 3 deletions tests/python_client/testcases/test_full_text_search.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2315,9 +2315,10 @@ def test_full_text_search_default(
@pytest.mark.parametrize("expr", ["text_match"])
@pytest.mark.parametrize("offset", [10])
@pytest.mark.parametrize("tokenizer", ["jieba"])
@pytest.mark.parametrize("inverted_index_algo", ct.inverted_index_algo)
def test_full_text_search_with_jieba_tokenizer(
self, offset, tokenizer, expr, enable_inverted_index, enable_partition_key, empty_percent, index_type, nq
):
self, offset, tokenizer, expr, enable_inverted_index, enable_partition_key,
empty_percent, index_type, nq, inverted_index_algo):
"""
target: test full text search
method: 1. enable full text search with jieba tokenizer and insert data with varchar
Expand Down Expand Up @@ -2430,6 +2431,7 @@ def test_full_text_search_with_jieba_tokenizer(
"params": {
"bm25_k1": 1.5,
"bm25_b": 0.75,
"inverted_index_algo": inverted_index_algo
}
}
)
Expand Down Expand Up @@ -3302,8 +3304,9 @@ class TestHybridSearchWithFullTextSearch(TestcaseBase):
@pytest.mark.parametrize("enable_inverted_index", [True])
@pytest.mark.parametrize("index_type", ["SPARSE_INVERTED_INDEX"])
@pytest.mark.parametrize("tokenizer", ["standard"])
@pytest.mark.parametrize("inverted_index_algo", ct.inverted_index_algo)
def test_hybrid_search_with_full_text_search(
self, tokenizer, enable_inverted_index, enable_partition_key, empty_percent, index_type
self, tokenizer, enable_inverted_index, enable_partition_key, empty_percent, index_type, inverted_index_algo
):
"""
target: test full text search
Expand Down Expand Up @@ -3403,6 +3406,7 @@ def test_hybrid_search_with_full_text_search(
"params": {
"bm25_k1": 1.5,
"bm25_b": 0.75,
"inverted_index_algo": inverted_index_algo
}
}
)
Expand Down
22 changes: 22 additions & 0 deletions tests/python_client/testcases/test_index.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1485,6 +1485,28 @@ def test_invalid_sparse_ratio(self, ratio, index):
check_task=CheckTasks.err_res,
check_items=error)

@pytest.mark.tags(CaseLabel.L2)
@pytest.mark.parametrize("inverted_index_algo", ["INVALID_ALGO"])
@pytest.mark.parametrize("index ", ct.all_index_types[9:11])
def test_invalid_sparse_inverted_index_algo(self, inverted_index_algo, index):
"""
target: index creation for unsupported ratio parameter
method: indexing of unsupported ratio parameters
expected: raise exception
"""
c_name = cf.gen_unique_str(prefix)
schema = cf.gen_default_sparse_schema()
collection_w = self.init_collection_wrap(name=c_name, schema=schema)
data = cf.gen_default_list_sparse_data()
collection_w.insert(data=data)
params = {"index_type": index, "metric_type": "IP", "params": {"inverted_index_algo": inverted_index_algo}}
error = {ct.err_code: 999,
ct.err_msg: f"sparse inverted index algo {inverted_index_algo} not found or not supported, "
f"supported: [TAAT_NAIVE DAAT_WAND DAAT_MAXSCORE]"}
index, _ = self.index_wrap.init_index(collection_w.collection, ct.default_sparse_vec_field_name, params,
check_task=CheckTasks.err_res,
check_items=error)


@pytest.mark.tags(CaseLabel.GPU)
class TestNewIndexAsync(TestcaseBase):
Expand Down
85 changes: 52 additions & 33 deletions tests/python_client/testcases/test_search.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3104,24 +3104,23 @@ def test_search_with_expression(self, null_data_percent):
assert set(ids).issubset(filter_ids_set)

# 5. search again with expression template and search hints
if expr != "": # TODO: remove this when issue #39013 is fixed
search_param = default_search_params.copy()
search_param.update({"hints": "iterative_filter"})
search_res, _ = collection_w.search(vectors[:default_nq], default_search_field,
search_param, nb,
expr=expr, expr_params=expr_params, _async=_async,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids)),
"_async": _async})
if _async:
search_res.done()
search_res = search_res.result()
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
assert set(ids).issubset(filter_ids_set)
search_param = default_search_params.copy()
search_param.update({"hints": "iterative_filter"})
search_res, _ = collection_w.search(vectors[:default_nq], default_search_field,
search_param, nb,
expr=expr, expr_params=expr_params, _async=_async,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"ids": insert_ids,
"limit": min(nb, len(filter_ids)),
"_async": _async})
if _async:
search_res.done()
search_res = search_res.result()
filter_ids_set = set(filter_ids)
for hits in search_res:
ids = hits.ids
assert set(ids).issubset(filter_ids_set)

@pytest.mark.tags(CaseLabel.L2)
@pytest.mark.parametrize("bool_type", [True, False, "true", "false"])
Expand Down Expand Up @@ -12860,7 +12859,8 @@ class TestSparseSearch(TestcaseBase):

@pytest.mark.tags(CaseLabel.L1)
@pytest.mark.parametrize("index", ct.all_index_types[9:11])
def test_sparse_index_search(self, index):
@pytest.mark.parametrize("inverted_index_algo", ct.inverted_index_algo)
def test_sparse_index_search(self, index, inverted_index_algo):
"""
target: verify that sparse index for sparse vectors can be searched properly
method: create connection, collection, insert and search
Expand All @@ -12873,12 +12873,16 @@ def test_sparse_index_search(self, index):
data = cf.gen_default_list_sparse_data(nb=3000)
collection_w.insert(data)
params = cf.get_index_params_params(index)
params.update({"inverted_index_algo": inverted_index_algo})
index_params = {"index_type": index, "metric_type": "IP", "params": params}
collection_w.create_index(ct.default_sparse_vec_field_name, index_params, index_name=index)
collection_w.load()

_params = cf.get_search_params_params(index)
_params.update({"dim_max_score_ratio": 1.05})
search_params = {"params": _params}
collection_w.search(data[-1][0:default_nq], ct.default_sparse_vec_field_name,
ct.default_sparse_search_params, default_limit,
search_params, default_limit,
output_fields=[ct.default_sparse_vec_field_name],
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
Expand All @@ -12887,7 +12891,7 @@ def test_sparse_index_search(self, index):
"output_fields": [ct.default_sparse_vec_field_name]})
expr = "int64 < 100 "
collection_w.search(data[-1][0:default_nq], ct.default_sparse_vec_field_name,
ct.default_sparse_search_params, default_limit,
search_params, default_limit,
expr=expr, output_fields=[ct.default_sparse_vec_field_name],
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
Expand Down Expand Up @@ -12923,7 +12927,8 @@ def test_sparse_index_dim(self, index, dim):

@pytest.mark.tags(CaseLabel.L2)
@pytest.mark.parametrize("index", ct.all_index_types[9:11])
def test_sparse_index_enable_mmap_search(self, index):
@pytest.mark.parametrize("inverted_index_algo", ct.inverted_index_algo)
def test_sparse_index_enable_mmap_search(self, index, inverted_index_algo):
"""
target: verify that the sparse indexes of sparse vectors can be searched properly after turning on mmap
method: create connection, collection, enable mmap, insert and search
Expand All @@ -12939,6 +12944,7 @@ def test_sparse_index_enable_mmap_search(self, index):
collection_w.insert(data)

params = cf.get_index_params_params(index)
params.update({"inverted_index_algo": inverted_index_algo})
index_params = {"index_type": index, "metric_type": "IP", "params": params}
collection_w.create_index(ct.default_sparse_vec_field_name, index_params, index_name=index)

Expand Down Expand Up @@ -12968,9 +12974,9 @@ def test_sparse_index_enable_mmap_search(self, index):
assert len(res) == 4

@pytest.mark.tags(CaseLabel.L1)
@pytest.mark.parametrize("ratio", [0.01, 0.1, 0.5, 0.9])
@pytest.mark.parametrize("drop_ratio_build", [0.01])
@pytest.mark.parametrize("index", ct.all_index_types[9:11])
def test_search_sparse_ratio(self, ratio, index):
def test_search_sparse_ratio(self, drop_ratio_build, index):
"""
target: create a sparse index by adjusting the ratio parameter.
method: create a sparse index by adjusting the ratio parameter.
Expand All @@ -12982,16 +12988,28 @@ def test_search_sparse_ratio(self, ratio, index):
collection_w = self.init_collection_wrap(c_name, schema=schema)
data = cf.gen_default_list_sparse_data(nb=4000)
collection_w.insert(data)
params = {"index_type": index, "metric_type": "IP", "params": {"drop_ratio_build": ratio}}
params = {"index_type": index, "metric_type": "IP", "params": {"drop_ratio_build": drop_ratio_build}}
collection_w.create_index(ct.default_sparse_vec_field_name, params, index_name=index)
collection_w.load()
assert collection_w.has_index(index_name=index)[0] is True
search_params = {"metric_type": "IP", "params": {"drop_ratio_search": ratio}}
collection_w.search(data[-1][0:default_nq], ct.default_sparse_vec_field_name,
search_params, default_limit,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"limit": default_limit})
_params = {"drop_ratio_search": 0.2}
for dim_max_score_ratio in [0.5, 0.99, 1, 1.3]:
_params.update({"dim_max_score_ratio": dim_max_score_ratio})
search_params = {"metric_type": "IP", "params": _params}
collection_w.search(data[-1][0:default_nq], ct.default_sparse_vec_field_name,
search_params, default_limit,
check_task=CheckTasks.check_search_results,
check_items={"nq": default_nq,
"limit": default_limit})
error = {ct.err_code: 999,
ct.err_msg: "should be in range [0.500000, 1.300000]"}
for invalid_ratio in [0.49, 1.4]:
_params.update({"dim_max_score_ratio": invalid_ratio})
search_params = {"metric_type": "IP", "params": _params}
collection_w.search(data[-1][0:default_nq], ct.default_sparse_vec_field_name,
search_params, default_limit,
check_task=CheckTasks.err_res,
check_items=error)

@pytest.mark.tags(CaseLabel.L2)
@pytest.mark.parametrize("index", ct.all_index_types[9:11])
Expand Down Expand Up @@ -13024,8 +13042,8 @@ def test_sparse_vector_search_output_field(self, index):

@pytest.mark.tags(CaseLabel.L2)
@pytest.mark.parametrize("index", ct.all_index_types[9:11])
@pytest.mark.xfail(reason="issue #36174")
def test_sparse_vector_search_iterator(self, index):
@pytest.mark.parametrize("inverted_index_algo", ct.inverted_index_algo)
def test_sparse_vector_search_iterator(self, index, inverted_index_algo):
"""
target: create sparse vectors and search iterator
method: create sparse vectors and search iterator
Expand All @@ -13038,6 +13056,7 @@ def test_sparse_vector_search_iterator(self, index):
data = cf.gen_default_list_sparse_data(nb=4000)
collection_w.insert(data)
params = cf.get_index_params_params(index)
params.update({"inverted_index_algo": inverted_index_algo})
index_params = {"index_type": index, "metric_type": "IP", "params": params}
collection_w.create_index(ct.default_sparse_vec_field_name, index_params, index_name=index)

Expand Down
6 changes: 3 additions & 3 deletions tests/python_client/utils/util_pymilvus.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@
import numpy as np
import requests
from sklearn import preprocessing
from pymilvus import DataType
from pymilvus import MilvusClient, DataType
from utils.util_log import test_log as log
from utils.util_k8s import init_k8s_client_config

Expand Down Expand Up @@ -115,9 +115,9 @@ def get_milvus(host, port, uri=None, handler=None, **kwargs):
handler = "GRPC"
try_connect = kwargs.get("try_connect", True)
if uri is not None:
milvus = Milvus(uri=uri, handler=handler, try_connect=try_connect)
milvus = MilvusClient(uri=uri, handler=handler, try_connect=try_connect)
else:
milvus = Milvus(host=host, port=port, handler=handler, try_connect=try_connect)
milvus = MilvusClient(uri=f"http://{host}:{port}", handler=handler, try_connect=try_connect)
return milvus


Expand Down

0 comments on commit 5fdc757

Please sign in to comment.