#754

Committed 04 Sep 2023 09:54PM UTC coverage: 74.807% (-5.5%) from 80.336%

Build # #754

Build Type

push

circle-ci

Committed by

jiashenC

Commit Message

update case

Run Details

8727 of 11666 relevant lines covered (74.81%)

0.75 hits per line

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

33.96

/evadb/executor/create_index_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 pathlib import Path

import pandas as pd

from evadb.catalog.sql_config import IDENTIFIER_COLUMN
from evadb.database import EvaDBDatabase
from evadb.executor.abstract_executor import AbstractExecutor
from evadb.executor.executor_utils import ExecutorError, handle_vector_store_params
from evadb.models.storage.batch import Batch
from evadb.plan_nodes.create_index_plan import CreateIndexPlan
from evadb.storage.storage_engine import StorageEngine
from evadb.third_party.vector_stores.types import FeaturePayload
from evadb.third_party.vector_stores.utils import VectorStoreFactory
from evadb.utils.logging_manager import logger


class CreateIndexExecutor(AbstractExecutor):
    def __init__(self, db: EvaDBDatabase, node: CreateIndexPlan):
        super().__init__(db, node)

    def exec(self, *args, **kwargs):
        if self.catalog().get_index_catalog_entry_by_name(self.node.name):
            msg = f"Index {self.node.name} already exists."
            logger.error(msg)
            raise ExecutorError(msg)

        self.index_path = self._get_index_save_path()
        self.index = None
        self._create_index()

        yield Batch(
            pd.DataFrame(
                [f"Index {self.node.name} successfully added to the database."]
            )
        )

    def _get_index_save_path(self) -> Path:
        index_dir = Path(self.config.get_value("storage", "index_dir"))
        if not index_dir.exists():
            index_dir.mkdir(parents=True, exist_ok=True)
        return str(
            index_dir
            / Path("{}_{}.index".format(self.node.vector_store_type, self.node.name))
        )

    def _create_index(self):
        try:
            # Get feature tables.
            feat_catalog_entry = self.node.table_ref.table.table_obj

            # Get feature column.
            feat_col_name = self.node.col_list[0].name
            feat_column = [
                col for col in feat_catalog_entry.columns if col.name == feat_col_name
            ][0]

            # Add features to index.
            # TODO: batch size is hardcoded for now.
            input_dim = -1
            storage_engine = StorageEngine.factory(self.db, feat_catalog_entry)
            for input_batch in storage_engine.read(feat_catalog_entry):
                if self.node.function:
                    # Create index through function expression.
                    # Function(input column) -> 2 dimension feature vector.
                    input_batch.modify_column_alias(feat_catalog_entry.name.lower())
                    feat_batch = self.node.function.evaluate(input_batch)
                    feat_batch.drop_column_alias()
                    input_batch.drop_column_alias()
                    feat = feat_batch.column_as_numpy_array("features")
                else:
                    # Create index on the feature table directly.
                    # Pandas wraps numpy array as an object inside a numpy
                    # array. Use zero index to get the actual numpy array.
                    feat = input_batch.column_as_numpy_array(feat_col_name)

                row_id = input_batch.column_as_numpy_array(IDENTIFIER_COLUMN)

                for i in range(len(input_batch)):
                    row_feat = feat[i].reshape(1, -1)
                    if self.index is None:
                        input_dim = row_feat.shape[1]
                        self.index = VectorStoreFactory.init_vector_store(
                            self.node.vector_store_type,
                            self.node.name,
                            **handle_vector_store_params(
                                self.node.vector_store_type, self.index_path
                            ),
                        )
                        self.index.create(input_dim)

                    # Row ID for mapping back to the row.
                    self.index.add([FeaturePayload(row_id[i], row_feat)])

            # Persist index.
            self.index.persist()

            # Save to catalog.
            self.catalog().insert_index_catalog_entry(
                self.node.name,
                self.index_path,
                self.node.vector_store_type,
                feat_column,
                self.node.function.signature() if self.node.function else None,
            )
        except Exception as e:
            # Delete index.
            if self.index:
                self.index.delete()

            # Throw exception back to user.
            raise ExecutorError(str(e))

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 pathlib import Path	1✔
16
17	import pandas as pd	1✔
18
19	from evadb.catalog.sql_config import IDENTIFIER_COLUMN	1✔
20	from evadb.database import EvaDBDatabase	1✔
21	from evadb.executor.abstract_executor import AbstractExecutor	1✔
22	from evadb.executor.executor_utils import ExecutorError, handle_vector_store_params	1✔
23	from evadb.models.storage.batch import Batch	1✔
24	from evadb.plan_nodes.create_index_plan import CreateIndexPlan	1✔
25	from evadb.storage.storage_engine import StorageEngine	1✔
26	from evadb.third_party.vector_stores.types import FeaturePayload	1✔
27	from evadb.third_party.vector_stores.utils import VectorStoreFactory	1✔
28	from evadb.utils.logging_manager import logger	1✔
29
30
31	class CreateIndexExecutor(AbstractExecutor):	1✔
32	def __init__(self, db: EvaDBDatabase, node: CreateIndexPlan):	1✔
33	super().__init__(db, node)	1✔
34
35	def exec(self, args, *kwargs):	1✔
36	if self.catalog().get_index_catalog_entry_by_name(self.node.name):	×
37	msg = f"Index {self.node.name} already exists."	×
38	logger.error(msg)	×
39	raise ExecutorError(msg)
40
41	self.index_path = self._get_index_save_path()	×
42	self.index = None	×
43	self._create_index()	×
44
45	yield Batch(	×
46	pd.DataFrame(
47	[f"Index {self.node.name} successfully added to the database."]
48	)
49	)
50
51	def _get_index_save_path(self) -> Path:	1✔
52	index_dir = Path(self.config.get_value("storage", "index_dir"))	×
53	if not index_dir.exists():	×
54	index_dir.mkdir(parents=True, exist_ok=True)	×
55	return str(	×
56	index_dir
57	/ Path("{}_{}.index".format(self.node.vector_store_type, self.node.name))
58	)
59
60	def _create_index(self):	1✔
61	try:	×
62	# Get feature tables.
63	feat_catalog_entry = self.node.table_ref.table.table_obj	×
64
65	# Get feature column.
66	feat_col_name = self.node.col_list[0].name	×
67	feat_column = [	×
68	col for col in feat_catalog_entry.columns if col.name == feat_col_name
69	][0]
70
71	# Add features to index.
72	# TODO: batch size is hardcoded for now.
73	input_dim = -1	×
74	storage_engine = StorageEngine.factory(self.db, feat_catalog_entry)	×
75	for input_batch in storage_engine.read(feat_catalog_entry):	×
76	if self.node.function:	×
77	# Create index through function expression.
78	# Function(input column) -> 2 dimension feature vector.
79	input_batch.modify_column_alias(feat_catalog_entry.name.lower())	×
80	feat_batch = self.node.function.evaluate(input_batch)	×
81	feat_batch.drop_column_alias()	×
82	input_batch.drop_column_alias()	×
83	feat = feat_batch.column_as_numpy_array("features")	×
84	else:
85	# Create index on the feature table directly.
86	# Pandas wraps numpy array as an object inside a numpy
87	# array. Use zero index to get the actual numpy array.
88	feat = input_batch.column_as_numpy_array(feat_col_name)	×
89
90	row_id = input_batch.column_as_numpy_array(IDENTIFIER_COLUMN)	×
91
92	for i in range(len(input_batch)):	×
93	row_feat = feat[i].reshape(1, -1)	×
94	if self.index is None:	×
95	input_dim = row_feat.shape[1]	×
96	self.index = VectorStoreFactory.init_vector_store(	×
97	self.node.vector_store_type,
98	self.node.name,
99	**handle_vector_store_params(
100	self.node.vector_store_type, self.index_path
101	),
102	)
103	self.index.create(input_dim)	×
104
105	# Row ID for mapping back to the row.
106	self.index.add([FeaturePayload(row_id[i], row_feat)])	×
107
108	# Persist index.
109	self.index.persist()	×
110
111	# Save to catalog.
112	self.catalog().insert_index_catalog_entry(	×
113	self.node.name,
114	self.index_path,
115	self.node.vector_store_type,
116	feat_column,
117	self.node.function.signature() if self.node.function else None,
118	)
119	except Exception as e:
120	# Delete index.
121	if self.index:
122	self.index.delete()
123
124	# Throw exception back to user.
125	raise ExecutorError(str(e))

georgia-tech-db / eva / #754

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