e6161546-9e33-42e7-a2b6-f8fbe6aa8255

Committed 08 Sep 2023 02:22AM UTC coverage: 80.449% (-12.5%) from 92.929%

Build # e6161546-9e33-42e7-a2b6-f8fbe6aa8255

Build Type

push

circle-ci

Committed by

jiashenC

Commit Message

fix lint

Run Details

13 of 13 new or added lines in 8 files covered. (100.0%)

9398 of 11682 relevant lines covered (80.45%)

1.45 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

35.85

/evadb/executor/vector_index_scan_executor.py

# coding=utf-8
# Copyright 2018-2023 EvaDB
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Iterator

import pandas as pd

from evadb.catalog.sql_config import ROW_NUM_COLUMN
from evadb.database import EvaDBDatabase
from evadb.executor.abstract_executor import AbstractExecutor
from evadb.executor.executor_utils import handle_vector_store_params
from evadb.models.storage.batch import Batch
from evadb.plan_nodes.vector_index_scan_plan import VectorIndexScanPlan
from evadb.third_party.vector_stores.types import VectorIndexQuery
from evadb.third_party.vector_stores.utils import VectorStoreFactory
from evadb.utils.logging_manager import logger


# Helper function for getting row_num column alias.
def get_row_num_column_alias(column_list):
    for column in column_list:
        alias, col_name = column.split(".")
        if col_name == ROW_NUM_COLUMN:
            return alias


class VectorIndexScanExecutor(AbstractExecutor):
    def __init__(self, db: EvaDBDatabase, node: VectorIndexScanPlan):
        super().__init__(db, node)

        self.index_name = node.index_name
        self.limit_count = node.limit_count
        self.search_query_expr = node.search_query_expr

    def exec(self, *args, **kwargs) -> Iterator[Batch]:
        # Fetch the index from disk.
        index_catalog_entry = self.catalog().get_index_catalog_entry_by_name(
            self.index_name
        )
        self.index_path = index_catalog_entry.save_file_path
        self.index = VectorStoreFactory.init_vector_store(
            self.node.vector_store_type,
            self.index_name,
            **handle_vector_store_params(self.node.vector_store_type, self.index_path),
        )

        # Get the query feature vector. Create a dummy
        # batch to retreat a single file path.
        dummy_batch = Batch(
            frames=pd.DataFrame(
                {"0": [0]},
            )
        )
        search_batch = self.search_query_expr.evaluate(dummy_batch)

        # Scan index. The search batch comes from the Open call.
        feature_col_name = self.search_query_expr.output_objs[0].name
        search_batch.drop_column_alias()
        search_feat = search_batch.column_as_numpy_array(feature_col_name)[0]
        search_feat = search_feat.reshape(1, -1)
        index_result = self.index.query(
            VectorIndexQuery(search_feat, self.limit_count.value)
        )
        # todo support queries over distance as well
        # distance_list = index_result.similarities
        row_num_np = index_result.ids

        # Load projected columns from disk and join with search results.
        row_num_col_name = None

        # handle the case where the index_results are less than self.limit_count.value
        num_required_results = self.limit_count.value
        if len(index_result.ids) < self.limit_count.value:
            num_required_results = len(index_result.ids)
            logger.warning(
                f"The index {self.index_name} returned only {num_required_results} results, which is fewer than the required {self.limit_count.value}."
            )

        res_row_list = [None for _ in range(num_required_results)]
        for batch in self.children[0].exec(**kwargs):
            column_list = batch.columns
            if not row_num_col_name:
                row_num_alias = get_row_num_column_alias(column_list)
                row_num_col_name = "{}.{}".format(row_num_alias, ROW_NUM_COLUMN)

            # Nested join.
            for _, row in batch.frames.iterrows():
                for idx, row_num in enumerate(row_num_np):
                    if row_num == row[row_num_col_name]:
                        res_row = dict()
                        for col_name in column_list:
                            res_row[col_name] = row[col_name]
                        res_row_list[idx] = res_row

        yield Batch(pd.DataFrame(res_row_list))

1	# coding=utf-8
2	# Copyright 2018-2023 EvaDB
3	#
4	# Licensed under the Apache License, Version 2.0 (the "License");
5	# you may not use this file except in compliance with the License.
6	# You may obtain a copy of the License at
7	#
8	# http://www.apache.org/licenses/LICENSE-2.0
9	#
10	# Unless required by applicable law or agreed to in writing, software
11	# distributed under the License is distributed on an "AS IS" BASIS,
12	# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13	# See the License for the specific language governing permissions and
14	# limitations under the License.
15	from typing import Iterator	2✔
16
17	import pandas as pd	2✔
18
19	from evadb.catalog.sql_config import ROW_NUM_COLUMN	2✔
20	from evadb.database import EvaDBDatabase	2✔
21	from evadb.executor.abstract_executor import AbstractExecutor	2✔
22	from evadb.executor.executor_utils import handle_vector_store_params	2✔
23	from evadb.models.storage.batch import Batch	2✔
24	from evadb.plan_nodes.vector_index_scan_plan import VectorIndexScanPlan	2✔
25	from evadb.third_party.vector_stores.types import VectorIndexQuery	2✔
26	from evadb.third_party.vector_stores.utils import VectorStoreFactory	2✔
27	from evadb.utils.logging_manager import logger	2✔
28
29
30	# Helper function for getting row_num column alias.
31	def get_row_num_column_alias(column_list):	2✔
32	for column in column_list:	×
33	alias, col_name = column.split(".")	×
34	if col_name == ROW_NUM_COLUMN:	×
35	return alias	×
36
37
38	class VectorIndexScanExecutor(AbstractExecutor):	2✔
39	def __init__(self, db: EvaDBDatabase, node: VectorIndexScanPlan):	2✔
40	super().__init__(db, node)	1✔
41
42	self.index_name = node.index_name	1✔
43	self.limit_count = node.limit_count	1✔
44	self.search_query_expr = node.search_query_expr	1✔
45
46	def exec(self, args, *kwargs) -> Iterator[Batch]:	2✔
47	# Fetch the index from disk.
48	index_catalog_entry = self.catalog().get_index_catalog_entry_by_name(	×
49	self.index_name
50	)
51	self.index_path = index_catalog_entry.save_file_path	×
52	self.index = VectorStoreFactory.init_vector_store(	×
53	self.node.vector_store_type,
54	self.index_name,
55	**handle_vector_store_params(self.node.vector_store_type, self.index_path),
56	)
57
58	# Get the query feature vector. Create a dummy
59	# batch to retreat a single file path.
60	dummy_batch = Batch(	×
61	frames=pd.DataFrame(
62	{"0": [0]},
63	)
64	)
65	search_batch = self.search_query_expr.evaluate(dummy_batch)	×
66
67	# Scan index. The search batch comes from the Open call.
68	feature_col_name = self.search_query_expr.output_objs[0].name	×
69	search_batch.drop_column_alias()	×
70	search_feat = search_batch.column_as_numpy_array(feature_col_name)[0]	×
71	search_feat = search_feat.reshape(1, -1)	×
72	index_result = self.index.query(	×
73	VectorIndexQuery(search_feat, self.limit_count.value)
74	)
75	# todo support queries over distance as well
76	# distance_list = index_result.similarities
77	row_num_np = index_result.ids	×
78
79	# Load projected columns from disk and join with search results.
80	row_num_col_name = None	×
81
82	# handle the case where the index_results are less than self.limit_count.value
83	num_required_results = self.limit_count.value	×
84	if len(index_result.ids) < self.limit_count.value:	×
85	num_required_results = len(index_result.ids)	×
86	logger.warning(	×
87	f"The index {self.index_name} returned only {num_required_results} results, which is fewer than the required {self.limit_count.value}."
88	)
89
90	res_row_list = [None for _ in range(num_required_results)]	×
91	for batch in self.children[0].exec(**kwargs):	×
92	column_list = batch.columns	×
93	if not row_num_col_name:	×
94	row_num_alias = get_row_num_column_alias(column_list)	×
95	row_num_col_name = "{}.{}".format(row_num_alias, ROW_NUM_COLUMN)	×
96
97	# Nested join.
98	for _, row in batch.frames.iterrows():	×
99	for idx, row_num in enumerate(row_num_np):	×
100	if row_num == row[row_num_col_name]:	×
101	res_row = dict()	×
102	for col_name in column_list:	×
103	res_row[col_name] = row[col_name]	×
104	res_row_list[idx] = res_row	×
105
106	yield Batch(pd.DataFrame(res_row_list))	×

georgia-tech-db / eva / e6161546-9e33-42e7-a2b6-f8fbe6aa8255

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous