forked from lancedb/lancedb
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtest_query.py
More file actions
653 lines (542 loc) · 19.5 KB
/
test_query.py
File metadata and controls
653 lines (542 loc) · 19.5 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright The LanceDB Authors
import unittest.mock as mock
from datetime import timedelta
from pathlib import Path
import lancedb
from lancedb.index import IvfPq, FTS
from lancedb.rerankers.cross_encoder import CrossEncoderReranker
import numpy as np
import pandas.testing as tm
import pyarrow as pa
import pytest
import pytest_asyncio
from lancedb.pydantic import LanceModel, Vector
from lancedb.query import (
AsyncQueryBase,
LanceVectorQueryBuilder,
Query,
)
from lancedb.table import AsyncTable, LanceTable
@pytest.fixture(scope="module")
def table(tmpdir_factory) -> lancedb.table.Table:
tmp_path = str(tmpdir_factory.mktemp("data"))
db = lancedb.connect(tmp_path)
df = pa.table(
{
"vector": pa.array(
[[1, 2], [3, 4]], type=pa.list_(pa.float32(), list_size=2)
),
"id": pa.array([1, 2]),
"str_field": pa.array(["a", "b"]),
"float_field": pa.array([1.0, 2.0]),
}
)
return db.create_table("test", df)
@pytest_asyncio.fixture
async def table_async(tmp_path) -> AsyncTable:
conn = await lancedb.connect_async(
tmp_path, read_consistency_interval=timedelta(seconds=0)
)
data = pa.table(
{
"vector": pa.array(
[[1, 2], [3, 4]], type=pa.list_(pa.float32(), list_size=2)
),
"id": pa.array([1, 2]),
"str_field": pa.array(["a", "b"]),
"float_field": pa.array([1.0, 2.0]),
"text": pa.array(["a", "dog"]),
}
)
return await conn.create_table("test", data)
@pytest_asyncio.fixture
async def table_struct_async(tmp_path) -> AsyncTable:
conn = await lancedb.connect_async(
tmp_path, read_consistency_interval=timedelta(seconds=0)
)
struct = pa.array([{"n_legs": 2, "animals": "Parrot"}, {"year": 2022, "n_legs": 4}])
month = pa.array([4, 6])
table = pa.Table.from_arrays([struct, month], names=["a", "month"])
return await conn.create_table("test_struct", table)
@pytest.fixture
def multivec_table(vector_value_type=pa.float32()) -> lancedb.table.Table:
db = lancedb.connect("memory://")
# Generate 256 rows of data
num_rows = 256
# Generate data for each column
vector_data = [
[[i, i + 1], [i + 2, i + 3]] for i in range(num_rows)
] # Adjust to match nested structure
id_data = list(range(1, num_rows + 1))
float_field_data = [float(i) for i in range(1, num_rows + 1)]
# Create the Arrow table
df = pa.table(
{
"vector": pa.array(
vector_data, type=pa.list_(pa.list_(vector_value_type, list_size=2))
),
"id": pa.array(id_data),
"float_field": pa.array(float_field_data),
}
)
return db.create_table("test", df)
@pytest_asyncio.fixture
async def multivec_table_async(vector_value_type=pa.float32()) -> AsyncTable:
conn = await lancedb.connect_async(
"memory://", read_consistency_interval=timedelta(seconds=0)
)
# Generate 256 rows of data
num_rows = 256
# Generate data for each column
vector_data = [
[[i, i + 1], [i + 2, i + 3]] for i in range(num_rows)
] # Adjust to match nested structure
id_data = list(range(1, num_rows + 1))
float_field_data = [float(i) for i in range(1, num_rows + 1)]
# Create the Arrow table
df = pa.table(
{
"vector": pa.array(
vector_data, type=pa.list_(pa.list_(vector_value_type, list_size=2))
),
"id": pa.array(id_data),
"float_field": pa.array(float_field_data),
}
)
return await conn.create_table("test_async", df)
def test_cast(table):
class TestModel(LanceModel):
vector: Vector(2)
id: int
str_field: str
float_field: float
q = LanceVectorQueryBuilder(table, [0, 0], "vector").limit(1)
results = q.to_pydantic(TestModel)
assert len(results) == 1
r0 = results[0]
assert isinstance(r0, TestModel)
assert r0.id == 1
assert r0.vector == [1, 2]
assert r0.str_field == "a"
assert r0.float_field == 1.0
def test_offset(table):
results_without_offset = LanceVectorQueryBuilder(table, [0, 0], "vector")
assert len(results_without_offset.to_pandas()) == 2
results_with_offset = LanceVectorQueryBuilder(table, [0, 0], "vector").offset(1)
assert len(results_with_offset.to_pandas()) == 1
def test_query_builder(table):
rs = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.limit(1)
.select(["id", "vector"])
.to_list()
)
assert rs[0]["id"] == 1
assert all(np.array(rs[0]["vector"]) == [1, 2])
def test_with_row_id(table: lancedb.table.Table):
rs = table.search().with_row_id(True).to_arrow()
assert "_rowid" in rs.column_names
assert rs["_rowid"].to_pylist() == [0, 1]
def test_distance_range(table: lancedb.table.Table):
q = [0, 0]
rs = table.search(q).to_arrow()
dists = rs["_distance"].to_pylist()
min_dist = dists[0]
max_dist = dists[-1]
res = table.search(q).distance_range(upper_bound=min_dist).to_arrow()
assert len(res) == 0
res = table.search(q).distance_range(lower_bound=max_dist).to_arrow()
assert len(res) == 1
assert res["_distance"].to_pylist() == [max_dist]
res = table.search(q).distance_range(upper_bound=max_dist).to_arrow()
assert len(res) == 1
assert res["_distance"].to_pylist() == [min_dist]
res = table.search(q).distance_range(lower_bound=min_dist).to_arrow()
assert len(res) == 2
assert res["_distance"].to_pylist() == [min_dist, max_dist]
@pytest.mark.asyncio
async def test_distance_range_async(table_async: AsyncTable):
q = [0, 0]
rs = await table_async.query().nearest_to(q).to_arrow()
dists = rs["_distance"].to_pylist()
min_dist = dists[0]
max_dist = dists[-1]
res = (
await table_async.query()
.nearest_to(q)
.distance_range(upper_bound=min_dist)
.to_arrow()
)
assert len(res) == 0
res = (
await table_async.query()
.nearest_to(q)
.distance_range(lower_bound=max_dist)
.to_arrow()
)
assert len(res) == 1
assert res["_distance"].to_pylist() == [max_dist]
res = (
await table_async.query()
.nearest_to(q)
.distance_range(upper_bound=max_dist)
.to_arrow()
)
assert len(res) == 1
assert res["_distance"].to_pylist() == [min_dist]
res = (
await table_async.query()
.nearest_to(q)
.distance_range(lower_bound=min_dist)
.to_arrow()
)
assert len(res) == 2
assert res["_distance"].to_pylist() == [min_dist, max_dist]
@pytest.mark.parametrize(
"multivec_table", [pa.float16(), pa.float32(), pa.float64()], indirect=True
)
def test_multivector(multivec_table: lancedb.table.Table):
# create index on multivector
multivec_table.create_index(
metric="cosine",
vector_column_name="vector",
index_type="IVF_PQ",
num_partitions=1,
num_sub_vectors=2,
)
# query with single vector
q = [1, 2]
rs = multivec_table.search(q).to_arrow()
# query with multiple vectors
q = [[1, 2], [1, 2]]
rs2 = multivec_table.search(q).to_arrow()
assert len(rs2) == len(rs)
for i in range(2):
assert rs2["_distance"][i].as_py() == rs["_distance"][i].as_py() * 2
# can't query with vector that dim not matched
with pytest.raises(Exception):
multivec_table.search([1, 2, 3]).to_arrow()
# can't query with vector list that some dim not matched
with pytest.raises(Exception):
multivec_table.search([[1, 2], [1, 2, 3]]).to_arrow()
@pytest.mark.asyncio
@pytest.mark.parametrize(
"multivec_table_async", [pa.float16(), pa.float32(), pa.float64()], indirect=True
)
async def test_multivector_async(multivec_table_async: AsyncTable):
# create index on multivector
await multivec_table_async.create_index(
"vector",
config=IvfPq(distance_type="cosine", num_partitions=1, num_sub_vectors=2),
)
# query with single vector
q = [1, 2]
rs = await multivec_table_async.query().nearest_to(q).to_arrow()
# query with multiple vectors
q = [[1, 2], [1, 2]]
rs2 = await multivec_table_async.query().nearest_to(q).to_arrow()
assert len(rs2) == len(rs)
for i in range(2):
assert rs2["_distance"][i].as_py() == rs["_distance"][i].as_py() * 2
# can't query with vector that dim not matched
with pytest.raises(Exception):
await multivec_table_async.query().nearest_to([1, 2, 3]).to_arrow()
# can't query with vector list that some dim not matched
with pytest.raises(Exception):
await multivec_table_async.query().nearest_to([[1, 2], [1, 2, 3]]).to_arrow()
def test_vector_query_with_no_limit(table):
with pytest.raises(ValueError):
LanceVectorQueryBuilder(table, [0, 0], "vector").limit(0).select(
["id", "vector"]
).to_list()
with pytest.raises(ValueError):
LanceVectorQueryBuilder(table, [0, 0], "vector").limit(None).select(
["id", "vector"]
).to_list()
def test_query_builder_batches(table):
rs = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.limit(2)
.select(["id", "vector"])
.to_batches(1)
)
rs_list = []
for item in rs:
rs_list.append(item)
assert isinstance(item, pa.RecordBatch)
assert len(rs_list) == 1
assert len(rs_list[0]["id"]) == 2
assert all(rs_list[0].to_pandas()["vector"][0] == [1.0, 2.0])
assert rs_list[0].to_pandas()["id"][0] == 1
assert all(rs_list[0].to_pandas()["vector"][1] == [3.0, 4.0])
assert rs_list[0].to_pandas()["id"][1] == 2
def test_dynamic_projection(table):
rs = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.limit(1)
.select({"id": "id", "id2": "id * 2"})
.to_list()
)
assert rs[0]["id"] == 1
assert rs[0]["id2"] == 2
def test_query_builder_with_filter(table):
rs = LanceVectorQueryBuilder(table, [0, 0], "vector").where("id = 2").to_list()
assert rs[0]["id"] == 2
assert all(np.array(rs[0]["vector"]) == [3, 4])
def test_query_builder_with_prefilter(table):
df = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.where("id = 2", prefilter=True)
.limit(1)
.to_pandas()
)
assert df["id"].values[0] == 2
assert all(df["vector"].values[0] == [3, 4])
df = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.where("id = 2", prefilter=False)
.limit(1)
.to_pandas()
)
assert len(df) == 0
# ensure the default prefilter = True
df = (
LanceVectorQueryBuilder(table, [0, 0], "vector")
.where("id = 2")
.limit(1)
.to_pandas()
)
assert df["id"].values[0] == 2
assert all(df["vector"].values[0] == [3, 4])
def test_query_builder_with_metric(table):
query = [4, 8]
vector_column_name = "vector"
df_default = LanceVectorQueryBuilder(table, query, vector_column_name).to_pandas()
df_l2 = (
LanceVectorQueryBuilder(table, query, vector_column_name)
.metric("L2")
.to_pandas()
)
tm.assert_frame_equal(df_default, df_l2)
df_cosine = (
LanceVectorQueryBuilder(table, query, vector_column_name)
.metric("cosine")
.limit(1)
.to_pandas()
)
assert df_cosine._distance[0] == pytest.approx(
cosine_distance(query, df_cosine.vector[0]),
abs=1e-6,
)
assert 0 <= df_cosine._distance[0] <= 1
def test_query_builder_with_different_vector_column():
table = mock.MagicMock(spec=LanceTable)
query = [4, 8]
vector_column_name = "foo_vector"
builder = (
LanceVectorQueryBuilder(table, query, vector_column_name)
.metric("cosine")
.where("b < 10")
.select(["b"])
.limit(2)
)
ds = mock.Mock()
table.to_lance.return_value = ds
builder.to_arrow()
table._execute_query.assert_called_once_with(
Query(
vector=query,
filter="b < 10",
prefilter=True,
k=2,
metric="cosine",
columns=["b"],
nprobes=20,
refine_factor=None,
vector_column="foo_vector",
),
None,
)
def cosine_distance(vec1, vec2):
return 1 - np.dot(vec1, vec2) / (np.linalg.norm(vec1) * np.linalg.norm(vec2))
async def check_query(
query: AsyncQueryBase, *, expected_num_rows=None, expected_columns=None
):
num_rows = 0
results = await query.to_batches()
async for batch in results:
if expected_columns is not None:
assert batch.schema.names == expected_columns
num_rows += batch.num_rows
if expected_num_rows is not None:
assert num_rows == expected_num_rows
@pytest.mark.asyncio
async def test_query_async(table_async: AsyncTable):
await check_query(
table_async.query(),
expected_num_rows=2,
expected_columns=["vector", "id", "str_field", "float_field", "text"],
)
await check_query(table_async.query().where("id = 2"), expected_num_rows=1)
await check_query(
table_async.query().select(["id", "vector"]), expected_columns=["id", "vector"]
)
await check_query(
table_async.query().select({"foo": "id", "bar": "id + 1"}),
expected_columns=["foo", "bar"],
)
await check_query(table_async.query().limit(1), expected_num_rows=1)
await check_query(table_async.query().offset(1), expected_num_rows=1)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])), expected_num_rows=2
)
# Support different types of inputs for the vector query
for vector_query in [
[1, 2],
[1.0, 2.0],
np.array([1, 2]),
(1, 2),
]:
await check_query(
table_async.query().nearest_to(vector_query), expected_num_rows=2
)
# No easy way to check these vector query parameters are doing what they say. We
# just check that they don't raise exceptions and assume this is tested at a lower
# level.
await check_query(
table_async.query().where("id = 2").nearest_to(pa.array([1, 2])).postfilter(),
expected_num_rows=1,
)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])).refine_factor(1),
expected_num_rows=2,
)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])).nprobes(10),
expected_num_rows=2,
)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])).bypass_vector_index(),
expected_num_rows=2,
)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])).distance_type("dot"),
expected_num_rows=2,
)
await check_query(
table_async.query().nearest_to(pa.array([1, 2])).distance_type("DoT"),
expected_num_rows=2,
)
# Make sure we can use a vector query as a base query (e.g. call limit on it)
# Also make sure `vector_search` works
await check_query(table_async.vector_search([1, 2]).limit(1), expected_num_rows=1)
# Also check an empty query
await check_query(table_async.query().where("id < 0"), expected_num_rows=0)
# with row id
await check_query(
table_async.query().select(["id", "vector"]).with_row_id(),
expected_columns=["id", "vector", "_rowid"],
)
@pytest.mark.asyncio
@pytest.mark.slow
async def test_query_reranked_async(table_async: AsyncTable):
# FTS with rerank
await table_async.create_index("text", config=FTS(with_position=False))
await check_query(
table_async.query().nearest_to_text("dog").rerank(CrossEncoderReranker()),
expected_num_rows=1,
)
# Vector query with rerank
await check_query(
table_async.vector_search([1, 2]).rerank(
CrossEncoderReranker(), query_string="dog"
),
expected_num_rows=2,
)
@pytest.mark.asyncio
async def test_query_to_arrow_async(table_async: AsyncTable):
table = await table_async.to_arrow()
assert table.num_rows == 2
assert table.num_columns == 5
table = await table_async.query().to_arrow()
assert table.num_rows == 2
assert table.num_columns == 5
table = await table_async.query().where("id < 0").to_arrow()
assert table.num_rows == 0
assert table.num_columns == 5
@pytest.mark.asyncio
async def test_query_to_pandas_async(table_async: AsyncTable):
df = await table_async.to_pandas()
assert df.shape == (2, 5)
df = await table_async.query().to_pandas()
assert df.shape == (2, 5)
df = await table_async.query().where("id < 0").to_pandas()
assert df.shape == (0, 5)
@pytest.mark.asyncio
async def test_query_to_pandas_flatten_async(table_struct_async: AsyncTable):
df = await table_struct_async.query().to_pandas()
assert df.shape == (2, 2)
df = await table_struct_async.query().to_pandas(flatten=True)
assert df.shape == (2, 4)
@pytest.mark.asyncio
async def test_query_to_polars_async(table_async: AsyncTable):
schema = await table_async.schema()
num_columns = len(schema.names)
df = await table_async.query().to_polars()
assert df.shape == (2, num_columns)
df = await table_async.query().where("id < 0").to_polars()
assert df.shape == (0, num_columns)
@pytest.mark.asyncio
async def test_none_query(table_async: AsyncTable):
with pytest.raises(ValueError):
await table_async.query().nearest_to(None).to_arrow()
@pytest.mark.asyncio
async def test_fast_search_async(tmp_path):
db = await lancedb.connect_async(tmp_path)
vectors = pa.FixedShapeTensorArray.from_numpy_ndarray(
np.random.rand(256, 32)
).storage
table = await db.create_table("test", pa.table({"vector": vectors}))
await table.create_index(
"vector", config=IvfPq(num_partitions=1, num_sub_vectors=1)
)
await table.add(pa.table({"vector": vectors}))
q = [1.0] * 32
plan = await table.query().nearest_to(q).explain_plan(True)
assert "LanceScan" in plan
plan = await table.query().nearest_to(q).fast_search().explain_plan(True)
assert "LanceScan" not in plan
def test_explain_plan(table):
q = LanceVectorQueryBuilder(table, [0, 0], "vector")
plan = q.explain_plan(verbose=True)
assert "KNN" in plan
@pytest.mark.asyncio
async def test_explain_plan_async(table_async: AsyncTable):
plan = await table_async.query().nearest_to(pa.array([1, 2])).explain_plan(True)
assert "KNN" in plan
@pytest.mark.asyncio
async def test_query_camelcase_async(tmp_path):
db = await lancedb.connect_async(tmp_path)
table = await db.create_table("test", pa.table({"camelCase": pa.array([1, 2])}))
result = await table.query().select(["camelCase"]).to_arrow()
assert result == pa.table({"camelCase": pa.array([1, 2])})
@pytest.mark.asyncio
async def test_query_to_list_async(table_async: AsyncTable):
list = await table_async.query().to_list()
assert len(list) == 2
assert list[0]["vector"] == [1, 2]
assert list[1]["vector"] == [3, 4]
@pytest.mark.asyncio
async def test_query_with_f16(tmp_path: Path):
db = await lancedb.connect_async(tmp_path)
f16_arr = np.array([1.0, 2.0, 3.0, 4.0], dtype=np.float16)
df = pa.table(
{
"vector": pa.FixedSizeListArray.from_arrays(f16_arr, 2),
"id": pa.array([1, 2]),
}
)
tbl = await db.create_table("test", df)
results = await tbl.vector_search([np.float16(1), np.float16(2)]).to_pandas()
assert len(results) == 2