forked from googleapis/google-cloud-python
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathdataframe.py
More file actions
5141 lines (4514 loc) · 188 KB
/
Copy pathdataframe.py
File metadata and controls
5141 lines (4514 loc) · 188 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
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# Copyright 2023 Google LLC
#
# 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.
"""DataFrame is a two dimensional data structure."""
from __future__ import annotations
import datetime
import inspect
import itertools
import re
import sys
import textwrap
import typing
import warnings
from typing import (
TYPE_CHECKING,
Any,
Callable,
Dict,
Hashable,
Iterable,
List,
Literal,
Mapping,
Optional,
Sequence,
Tuple,
TypeVar,
Union,
cast,
overload,
)
import bigframes_vendored.constants as constants
import bigframes_vendored.pandas.core.frame as vendored_pandas_frame
import bigframes_vendored.pandas.pandas._typing as vendored_pandas_typing
import google.api_core.exceptions
import google.cloud.bigquery as bigquery
import google.cloud.bigquery.job
import google.cloud.bigquery.table
import numpy
import pandas
import pandas.io.formats.format
import pyarrow
import tabulate
from pandas.api import extensions as pd_ext
import bigframes.constants
import bigframes.core
import bigframes.core.block_transforms as block_ops
import bigframes.core.blocks as blocks
import bigframes.core.col
import bigframes.core.convert
import bigframes.core.explode
import bigframes.core.expression as ex
import bigframes.core.groupby as groupby
import bigframes.core.guid
import bigframes.core.indexers as indexers
import bigframes.core.indexes as indexes
import bigframes.core.interchange
import bigframes.core.ordering as order
import bigframes.core.utils as utils
import bigframes.core.validations as validations
import bigframes.core.window
import bigframes.core.window_spec as windows
import bigframes.dtypes
import bigframes.exceptions as bfe
import bigframes.formatting_helpers as formatter
import bigframes.functions
import bigframes.operations as ops
import bigframes.operations.aggregations as agg_ops
import bigframes.operations.ai
import bigframes.operations.plotting as plotting
import bigframes.operations.semantics
import bigframes.operations.structs
import bigframes.series
import bigframes.session._io.bigquery
import bigframes.session.execution_spec as ex_spec
from bigframes._tools import docs
from bigframes.core import agg_expressions
from bigframes.core.logging import log_adapter
from bigframes.core.window import rolling
from bigframes.functions import function_typing
if TYPE_CHECKING:
from _typeshed import SupportsRichComparison
import bigframes.extensions.bigframes.dataframe_accessor as bigquery_accessor
import bigframes.session
SingleItemValue = Union[
bigframes.series.Series,
int,
float,
str,
pandas.Timedelta,
Callable,
bigframes.core.col.Expression,
]
MultiItemValue = Union[
"DataFrame", Sequence[int | float | str | pandas.Timedelta | Callable]
]
U = TypeVar("U")
LevelType = typing.Hashable
LevelsType = typing.Union[LevelType, typing.Sequence[LevelType]]
ERROR_IO_ONLY_GS_PATHS = f"Only Google Cloud Storage (gs://...) paths are supported. {constants.FEEDBACK_LINK}"
ERROR_IO_REQUIRES_WILDCARD = (
"Google Cloud Storage path must contain a wildcard '*' character. See: "
"https://cloud.google.com/bigquery/docs/reference/standard-sql/other-statements#export_data_statement"
f"{constants.FEEDBACK_LINK}"
)
# Inherits from pandas DataFrame so that we can use the same docstrings.
@log_adapter.class_logger
@docs.inherit_docs(vendored_pandas_frame.DataFrame)
class DataFrame:
# internal flag to disable cache at all
_disable_cache_override: bool = False
# Must be above 5000 for pandas to delegate to bigframes for binops
__pandas_priority__ = 15000
def __init__(
self,
data=None,
index: vendored_pandas_typing.Axes | None = None,
columns: vendored_pandas_typing.Axes | None = None,
dtype: typing.Optional[
bigframes.dtypes.DtypeString | bigframes.dtypes.Dtype
] = None,
copy: typing.Optional[bool] = None,
*,
session: typing.Optional[bigframes.session.Session] = None,
):
global bigframes
self._query_job: Optional[google.cloud.bigquery.job.QueryJob] = None
if copy is not None and not copy:
raise ValueError(
f"DataFrame constructor only supports copy=True. {constants.FEEDBACK_LINK}"
)
# Ignore object dtype if provided, as it provides no additional
# information about what BigQuery type to use.
if dtype is not None and bigframes.dtypes.is_object_like(dtype):
dtype = None
# Check to see if constructing from BigQuery-backed objects before
# falling back to pandas constructor
block = None
if isinstance(data, blocks.Block):
block = data
elif isinstance(data, DataFrame):
block = data._get_block()
# Dict of Series
elif (
utils.is_dict_like(data)
and len(data) >= 1
and any(
isinstance(data[key], bigframes.series.Series) for key in data.keys()
)
):
if not all(
isinstance(data[key], bigframes.series.Series) for key in data.keys()
):
# TODO(tbergeron): Support local list/series data by converting to memtable.
raise NotImplementedError(
f"Cannot mix Series with other types. {constants.FEEDBACK_LINK}"
)
keys = list(data.keys())
first_label, first_series = keys[0], data[keys[0]]
block = (
typing.cast(bigframes.series.Series, first_series)
._get_block()
.with_column_labels([first_label])
)
for key in keys[1:]:
other = typing.cast(bigframes.series.Series, data[key])
other_block = other._block.with_column_labels([key])
# Pandas will keep original sorting if all indices are aligned.
# We cannot detect this easily however, and so always sort on index
block, _ = block.join( # type:ignore
other_block, how="outer", sort=True
)
if block:
if index is not None:
bf_index = indexes.Index(index)
idx_block = bf_index._block
idx_cols = idx_block.index_columns
block, (_, r_mapping) = block.reset_index().join(
bf_index._block.reset_index(), how="inner"
)
block = block.set_index([r_mapping[idx_col] for idx_col in idx_cols])
if columns:
column_ids = [
block.resolve_label_exact_or_error(label) for label in list(columns)
]
block = block.select_columns(column_ids) # type:ignore
if dtype:
bf_dtype = bigframes.dtypes.bigframes_type(dtype)
block = block.multi_apply_unary_op(ops.AsTypeOp(to_type=bf_dtype))
else:
if isinstance(dtype, str) and dtype.lower() == "json":
dtype = bigframes.dtypes.JSON_DTYPE
import bigframes.pandas
pd_dataframe = pandas.DataFrame(
data=data,
index=index, # type:ignore
columns=columns, # type:ignore
dtype=dtype, # type:ignore
)
if session:
block = session.read_pandas(pd_dataframe)._get_block()
else:
block = bigframes.pandas.read_pandas(pd_dataframe)._get_block()
# We use _block as an indicator in __getattr__ and __setattr__ to see
# if the object is fully initialized, so make sure we set the _block
# attribute last.
self._block = block
self._block.session._register_object(self)
def __dir__(self):
return dir(type(self)) + [
label
for label in self._block.column_labels
if label and isinstance(label, str)
]
def _ipython_key_completions_(self) -> List[str]:
return list(
[
label
for label in self._block.column_labels
if label and isinstance(label, str)
]
)
def _find_indices(
self,
columns: Union[blocks.Label, Sequence[blocks.Label]],
tolerance: bool = False,
) -> Sequence[int]:
"""Find corresponding indices in df._block.column_labels for column name(s).
Order is kept the same as input names order.
Args:
columns: column name(s)
tolerance: True to pass through columns not found. False to raise
ValueError.
"""
col_ids = self._sql_names(columns, tolerance)
return [self._block.value_columns.index(col_id) for col_id in col_ids]
def _resolve_label_exact(self, label) -> Optional[str]:
return self._block.resolve_label_exact(label)
def _sql_names(
self,
columns: Union[blocks.Label, Sequence[blocks.Label], pandas.Index],
tolerance: bool = False,
) -> Sequence[str]:
"""Retrieve sql name (column name in BQ schema) of column(s)."""
labels = (
columns
if utils.is_list_like(columns) and not isinstance(columns, tuple)
else [columns]
) # type:ignore
results: Sequence[str] = []
for label in labels:
col_ids = self._block.label_to_col_id.get(label, [])
if not tolerance and len(col_ids) == 0:
raise ValueError(f"Column name {label} doesn't exist")
results = (*results, *col_ids)
return results
@property
@validations.requires_index
def index(
self,
) -> indexes.Index:
return indexes.Index.from_frame(self)
@index.setter
def index(self, value):
# TODO: Handle assigning MultiIndex
result = self._assign_single_item("_new_bf_index", value).set_index(
"_new_bf_index"
)
self._set_block(result._get_block())
self.index.name = value.name if hasattr(value, "name") else None
@property
@validations.requires_index
def loc(self) -> indexers.LocDataFrameIndexer:
return indexers.LocDataFrameIndexer(self)
@property
@validations.requires_ordering()
def iloc(self) -> indexers.ILocDataFrameIndexer:
return indexers.ILocDataFrameIndexer(self)
@property
@validations.requires_ordering()
def iat(self) -> indexers.IatDataFrameIndexer:
return indexers.IatDataFrameIndexer(self)
@property
@validations.requires_index
def at(self) -> indexers.AtDataFrameIndexer:
return indexers.AtDataFrameIndexer(self)
@property
def dtypes(self) -> pandas.Series:
dtypes = self._block.dtypes
bigframes.dtypes.warn_on_db_dtypes_json_dtype(dtypes)
return pandas.Series(data=dtypes, index=self._block.column_labels)
@property
def columns(self) -> pandas.Index:
return self._block.column_labels
@columns.setter
def columns(self, labels: pandas.Index):
new_block = self._block.with_column_labels(labels)
self._set_block(new_block)
@property
def shape(self) -> Tuple[int, int]:
return self._block.shape
@property
def axes(self) -> list:
return [self.index, self.columns]
@property
def size(self) -> int:
rows, cols = self.shape
return rows * cols
@property
def ndim(self) -> int:
return 2
@property
def empty(self) -> bool:
return self.size == 0
@property
def values(self) -> numpy.ndarray:
return self.to_numpy()
@property
def bqclient(self) -> bigframes.Session:
"""BigQuery REST API Client the DataFrame uses for operations."""
return self._session.bqclient
@property
def _session(self) -> bigframes.Session:
return self._get_block().expr.session
@property
def bigquery(
self,
) -> bigquery_accessor.BigframesBigQueryDataFrameAccessor:
"""
Accessor for BigQuery functionality.
Returns:
bigframes.extensions.core.dataframe_accessor.BigQueryDataFrameAccessor:
Accessor that exposes BigQuery functionality on a DataFrame,
with method names closer to SQL.
"""
# Import the accessor here to avoid circular imports.
import bigframes.extensions.bigframes.dataframe_accessor
return bigframes.extensions.bigframes.dataframe_accessor.BigframesBigQueryDataFrameAccessor(
self
)
@property
def _has_index(self) -> bool:
return len(self._block.index_columns) > 0
@property
@validations.requires_ordering()
def T(self) -> DataFrame:
return DataFrame(self._get_block().transpose())
@validations.requires_index
@validations.requires_ordering()
def transpose(self) -> DataFrame:
return self.T
def __len__(self):
rows, _ = self.shape
return rows
def __bool__(self):
raise ValueError(
"Cannot convert dataframe into bool. Consider using .empty(), .any(), or .all() methods."
)
def __iter__(self):
return iter(self.columns)
def __contains__(self, key) -> bool:
return key in self.columns
def astype(
self,
dtype: Union[
bigframes.dtypes.DtypeString,
bigframes.dtypes.Dtype,
type,
dict[str, Union[bigframes.dtypes.DtypeString, bigframes.dtypes.Dtype]],
],
*,
errors: Literal["raise", "null"] = "raise",
) -> DataFrame:
if errors not in ["raise", "null"]:
raise ValueError("Arg 'error' must be one of 'raise' or 'null'")
safe_cast = errors == "null"
if isinstance(dtype, dict):
result = self.copy()
for col, to_type in dtype.items():
result[col] = result[col].astype(to_type)
return result
dtype = bigframes.dtypes.bigframes_type(dtype)
return self._apply_unary_op(ops.AsTypeOp(dtype, safe_cast))
def _should_sql_have_index(self) -> bool:
"""Should the SQL we pass to BQML and other I/O include the index?"""
return self._has_index and (
self.index.name is not None or len(self.index.names) > 1
)
def _to_placeholder_table(
self, dry_run: bool = False
) -> google.cloud.bigquery.table.TableReference:
"""Compiles this DataFrame's expression tree to SQL and saves it to a
(temporary) view or table (in the case of a dry run).
"""
return self._block.to_placeholder_table(
include_index=self._should_sql_have_index(), dry_run=dry_run
)
def _to_sql_query(
self, include_index: bool, enable_cache: bool = True
) -> Tuple[str, list[str], list[blocks.Label]]:
"""Compiles this DataFrame's expression tree to SQL, optionally
including index columns.
Args:
include_index (bool):
whether to include index columns.
Returns:
Tuple[sql_string, index_column_id_list, index_column_label_list]:
If include_index is set to False, index_column_id_list and index_column_label_list
return empty lists.
"""
return self._block.to_sql_query(include_index, enable_cache=enable_cache)
@property
def sql(self) -> str:
"""Compiles this DataFrame's expression tree to SQL.
Returns:
str:
string representing the compiled SQL.
"""
try:
include_index = self._should_sql_have_index()
sql, _, _ = self._to_sql_query(include_index=include_index)
return sql
except AttributeError as e:
# Workaround for a development-mode debugging issue:
# An `AttributeError` originating *inside* this @property getter (e.g., due to
# a typo or referencing a non-existent attribute) can be mistakenly intercepted
# by the class's __getattr__ method if one is defined.
# We catch the AttributeError and raise SyntaxError instead to make it clear
# the error originates *here* in the property implementation.
# See: https://stackoverflow.com/questions/50542177/correct-handling-of-attributeerror-in-getattr-when-using-property
raise SyntaxError(
"AttributeError encountered. Please check the implementation for incorrect attribute access."
) from e
@property
def query_job(self) -> Optional[google.cloud.bigquery.job.QueryJob]:
"""BigQuery job metadata for the most recent query.
Returns:
None or google.cloud.bigquery.job.QueryJob:
The most recent `QueryJob
<https://cloud.google.com/python/docs/reference/bigquery/latest/google.cloud.bigquery.job.QueryJob>`_.
"""
if self._query_job is None:
self._set_internal_query_job(self._compute_dry_run())
return self._query_job
def memory_usage(self, index: bool = True):
n_rows, _ = self.shape
# like pandas, treat all variable-size objects as just 8-byte pointers, ignoring actual object
column_sizes = self.dtypes.map(
lambda dtype: bigframes.dtypes.DTYPE_BYTE_SIZES.get(dtype, 8) * n_rows
)
if index and self._has_index:
index_size = pandas.Series([self.index._memory_usage()], index=["Index"])
column_sizes = pandas.concat([index_size, column_sizes])
return column_sizes
def info(
self,
verbose: Optional[bool] = None,
buf=None,
max_cols: Optional[int] = None,
memory_usage: Optional[bool] = None,
show_counts: Optional[bool] = None,
):
obuf = buf or sys.stdout
n_rows, n_columns = self.shape
max_cols = (
max_cols
if max_cols is not None
else bigframes.options.display.max_info_columns
)
show_all_columns = verbose if verbose is not None else (n_columns < max_cols)
obuf.write(f"{type(self)}\n")
if self._block.has_index:
index_type = "MultiIndex" if self.index.nlevels > 1 else "Index"
index_stats = f"{n_rows} entries"
if n_rows > 0:
# These accessses are kind of expensive, maybe should try to skip?
first_indice = self.index[0]
last_indice = self.index[-1]
index_stats += f", {first_indice} to {last_indice}"
obuf.write(f"{index_type}: {index_stats}\n")
else:
obuf.write("NullIndex\n")
if n_columns == 0:
# We don't display any more information if the dataframe has no columns
obuf.write("Empty DataFrame\n")
return
dtype_strings = self.dtypes.astype("string")
if show_all_columns:
obuf.write(f"Data columns (total {n_columns} columns):\n")
column_info = self.columns.to_frame(name="Column")
max_rows = bigframes.options.display.max_info_rows
too_many_rows = n_rows > max_rows if max_rows is not None else False
if show_counts if show_counts is not None else (not too_many_rows):
non_null_counts = self.count().to_pandas()
column_info["Non-Null Count"] = non_null_counts.map(
lambda x: f"{int(x)} non-null"
)
column_info["Dtype"] = dtype_strings
column_info = column_info.reset_index(drop=True)
column_info.index.name = "#"
column_info_formatted = tabulate.tabulate(column_info, headers="keys") # type: ignore
obuf.write(column_info_formatted)
obuf.write("\n")
else: # Just number of columns and first, last
obuf.write(
f"Columns: {n_columns} entries, {self.columns[0]} to {self.columns[-1]}\n"
)
dtype_counts = dtype_strings.value_counts().sort_index(ascending=True).items()
dtype_counts_formatted = ", ".join(
f"{dtype}({count})" for dtype, count in dtype_counts
)
obuf.write(f"dtypes: {dtype_counts_formatted}\n")
show_memory = (
memory_usage
if memory_usage is not None
else bigframes.options.display.memory_usage
)
if show_memory:
# TODO: Convert to different units (kb, mb, etc.)
obuf.write(f"memory usage: {self.memory_usage().sum()} bytes\n")
def select_dtypes(self, include=None, exclude=None) -> DataFrame:
# Create empty pandas dataframe with same schema and then leverage actual pandas implementation
as_pandas = pandas.DataFrame(
{
col_id: pandas.Series([], dtype=dtype)
for col_id, dtype in zip(self._block.value_columns, self._block.dtypes)
}
)
selected_columns = tuple(
as_pandas.select_dtypes(include=include, exclude=exclude).columns
)
return DataFrame(self._block.select_columns(selected_columns))
def _set_internal_query_job(
self, query_job: Optional[google.cloud.bigquery.job.QueryJob]
):
self._query_job = query_job
def __getitem__(
self,
key: Union[
blocks.Label,
List[str],
List[blocks.Label],
# Index of column labels can be treated the same as a sequence of column labels.
pandas.Index,
bigframes.series.Series,
slice,
],
): # No return type annotations (like pandas) as type cannot always be determined statically
# NOTE: This implements the operations described in
# https://pandas.pydata.org/docs/getting_started/intro_tutorials/03_subset_data.html
import bigframes.core.col
import bigframes.pandas
if isinstance(key, bigframes.pandas.Series):
return self._getitem_bool_series(key)
if isinstance(key, slice):
return self.iloc[key]
if isinstance(key, bigframes.core.col.Expression):
return self.loc[key]
# TODO(tswast): Fix this pylance warning: Class overlaps "Hashable"
# unsafely and could produce a match at runtime
if isinstance(key, blocks.Label):
return self._getitem_label(key)
if utils.is_list_like(key):
return self._getitem_columns(key)
else:
# TODO(tswast): What case is this supposed to be handling?
return self._getitem_columns([cast(Hashable, key)])
def _getitem_columns(self, key: Sequence[blocks.Label]) -> DataFrame:
selected_ids: Tuple[str, ...] = ()
for label in key:
col_ids = self._block.label_to_col_id[label]
selected_ids = (*selected_ids, *col_ids)
return DataFrame(self._block.select_columns(selected_ids))
def _getitem_label(self, key: blocks.Label):
col_ids = self._block.cols_matching_label(key)
if len(col_ids) == 0:
raise KeyError(
f"{key} not found in DataFrame columns: {self._block.column_labels}"
)
block = self._block.select_columns(col_ids)
if isinstance(self.columns, pandas.MultiIndex):
# Multiindex should drop-level if not selecting entire
key_levels = len(key) if isinstance(key, tuple) else 1
index_levels = self.columns.nlevels
if key_levels < index_levels:
block = block.with_column_labels(
block.column_labels.droplevel(list(range(key_levels)))
)
# Force return DataFrame in this case, even if only single column
return DataFrame(block)
if len(col_ids) == 1:
return bigframes.series.Series(block)
return DataFrame(block)
# Bool Series selects rows
def _getitem_bool_series(self, key: bigframes.series.Series) -> DataFrame:
if not key.dtype == pandas.BooleanDtype():
raise NotImplementedError(
f"Only boolean series currently supported for indexing. {constants.FEEDBACK_LINK}"
)
# TODO: enforce stricter alignment
(
combined_index,
(
get_column_left,
get_column_right,
),
) = self._block.join(key._block, how="left")
block = combined_index
filter_col_id = get_column_right[key._value_column]
block = block.filter_by_id(filter_col_id)
block = block.drop_columns([filter_col_id])
return DataFrame(block)
def __getattr__(self, key: str):
# To allow subclasses to set private attributes before the class is
# fully initialized, protect against recursion errors with
# uninitialized DataFrame objects. Note: this comes at the downside
# that columns with a leading `_` won't be treated as columns.
#
# See:
# https://github.com/googleapis/python-bigquery-dataframes/issues/728
# and
# https://nedbatchelder.com/blog/201010/surprising_getattr_recursion.html
if key == "_block":
raise AttributeError(key)
if key in self._block.column_labels:
return self.__getitem__(key)
if hasattr(pandas.DataFrame, key):
log_adapter.submit_pandas_labels(
self._block.expr.session.bqclient, self.__class__.__name__, key
)
raise AttributeError(
textwrap.dedent(
f"""
BigQuery DataFrames has not yet implemented an equivalent to
'pandas.DataFrame.{key}'. {constants.FEEDBACK_LINK}
"""
)
)
raise AttributeError(key)
def __setattr__(self, key: str, value):
if key == "_block":
object.__setattr__(self, key, value)
return
# To allow subclasses to set private attributes before the class is
# fully initialized, assume anything set before `_block` is initialized
# is a regular attribute.
if not hasattr(self, "_block"):
object.__setattr__(self, key, value)
return
# If someone has a column named the same as a normal attribute
# (e.g. index), we want to set the normal attribute, not the column.
# To do that, check if there is a normal attribute by using
# __getattribute__ (not __getattr__, because that includes columns).
# If that returns a value without raising, then we know this is a
# normal attribute and we should prefer that.
try:
object.__getattribute__(self, key)
return object.__setattr__(self, key, value)
except AttributeError:
pass
# If we made it here, then we know that it's not a regular attribute
# already, so it might be a column to update. Note: we don't allow
# adding new columns using __setattr__, only __setitem__, that way we
# can still add regular new attributes.
if key in self._block.column_labels:
self[key] = value
else:
object.__setattr__(self, key, value)
def __repr__(self) -> str:
"""Converts a DataFrame to a string. Calls to_pandas.
Only represents the first `bigframes.options.display.max_rows`.
"""
# Protect against errors with uninitialized DataFrame. See:
# https://github.com/googleapis/python-bigquery-dataframes/issues/728
if not hasattr(self, "_block"):
return object.__repr__(self)
opts = bigframes.options.display
max_results = opts.max_rows
if opts.repr_mode == "deferred":
return formatter.repr_query_job(self._compute_dry_run())
# TODO(swast): pass max_columns and get the true column count back. Maybe
# get 1 more column than we have requested so that pandas can add the
# ... for us?
pandas_df, row_count, query_job = self._block.retrieve_repr_request_results(
max_results
)
self._set_internal_query_job(query_job)
from bigframes.display import plaintext
return plaintext.create_text_representation(
pandas_df,
row_count,
is_series=False,
has_index=self._has_index,
column_count=len(self.columns),
)
def _get_display_df_and_blob_cols(self) -> tuple[DataFrame, list[str]]:
"""Process ObjectRef columns for display."""
df = self
blob_cols = []
if bigframes.options.display.blob_display:
blob_cols = [
series_name
for series_name, series in self.items()
if series.dtype == bigframes.dtypes.OBJ_REF_DTYPE
]
if blob_cols:
df = self.copy()
for col in blob_cols:
# TODO(garrettwu): Not necessary to get access urls for all the rows. Update when having a to get URLs from local data.
df[col] = df[col].blob._get_runtime(mode="R", with_metadata=True)
return df, blob_cols
def _repr_mimebundle_(self, include=None, exclude=None):
"""
Custom display method for IPython/Jupyter environments.
This is called by IPython's display system when the object is displayed.
"""
# TODO(b/467647693): Anywidget integration has been tested in Jupyter, VS Code, and
# BQ Studio, but there is a known compatibility issue with Marimo that needs to be addressed.
from bigframes.display import html
return html.repr_mimebundle(self, include=include, exclude=exclude)
def __delitem__(self, key: str):
df = self.drop(columns=[key])
self._set_block(df._get_block())
def __setitem__(
self,
key: str | list[str] | pandas.Index,
value: SingleItemValue | MultiItemValue,
):
if isinstance(key, (list, pandas.Index)):
df = self._assign_multi_items(key, value)
else:
df = self._assign_single_item(key, value)
self._set_block(df._get_block())
def _apply_binop(
self,
other: float | int | bigframes.series.Series | DataFrame,
op,
axis: str | int = "columns",
how: str = "outer",
reverse: bool = False,
):
if isinstance(other, bigframes.dtypes.LOCAL_SCALAR_TYPES):
return self._apply_scalar_binop(other, op, reverse=reverse)
elif isinstance(other, DataFrame):
return self._apply_dataframe_binop(other, op, how=how, reverse=reverse)
elif isinstance(other, pandas.DataFrame):
return self._apply_dataframe_binop(
DataFrame(other), op, how=how, reverse=reverse
)
elif utils.get_axis_number(axis) == 0:
return self._apply_series_binop_axis_0(other, op, how, reverse)
elif utils.get_axis_number(axis) == 1:
return self._apply_series_binop_axis_1(other, op, how, reverse)
raise NotImplementedError(
f"binary operation is not implemented on the second operand of type {type(other).__name__}."
f"{constants.FEEDBACK_LINK}"
)
def _apply_scalar_binop(
self,
other: bigframes.dtypes.LOCAL_SCALAR_TYPE,
op: ops.BinaryOp,
reverse: bool = False,
) -> DataFrame:
if reverse:
expr = op.as_expr(
left_input=ex.const(other),
right_input=ex.free_var("var1"),
)
else:
expr = op.as_expr(
left_input=ex.free_var("var1"),
right_input=ex.const(other),
)
return DataFrame(self._block.multi_apply_unary_op(expr))
def _apply_series_binop_axis_0(
self,
other,
op: ops.BinaryOp,
how: str = "outer",
reverse: bool = False,
) -> DataFrame:
bf_series = bigframes.core.convert.to_bf_series(
other, self.index if self._has_index else None, self._session
)
aligned_block, columns, expr_pairs = self._block._align_axis_0(
bf_series._block, how=how
)
result = aligned_block._apply_binop(
op, inputs=expr_pairs, labels=columns, reverse=reverse
)
return DataFrame(result)
def _apply_series_binop_axis_1(
self,
other,
op: ops.BinaryOp,
how: str = "outer",
reverse: bool = False,
) -> DataFrame:
"""Align dataframe with pandas series by inlining series values as literals."""
# If we already know the transposed schema (from the transpose cache), we don't need to materialize rows from other
# Instead, can fully defer execution (as a cross-join)
if (
isinstance(other, bigframes.series.Series)
and other._block._transpose_cache is not None
):
aligned_block, columns, expr_pairs = self._block._align_series_block_axis_1(
other._block, how=how
)
else:
# Fallback path, materialize `other` locally
pd_series = bigframes.core.convert.to_pd_series(other, self.columns)
aligned_block, columns, expr_pairs = self._block._align_pd_series_axis_1(
pd_series, how=how
)
result = aligned_block._apply_binop(
op, inputs=expr_pairs, labels=columns, reverse=reverse
)
return DataFrame(result)
def _apply_dataframe_binop(
self,
other: DataFrame,
op: ops.BinaryOp,
how: str = "outer",
reverse: bool = False,
) -> DataFrame:
aligned_block, columns, expr_pairs = self._block._align_both_axes(
other._block, how=how
)
result = aligned_block._apply_binop(
op, inputs=expr_pairs, labels=columns, reverse=reverse
)
return DataFrame(result)
def eq(self, other: typing.Any, axis: str | int = "columns") -> DataFrame:
return self._apply_binop(other, ops.eq_op, axis=axis)
def __eq__(self, other) -> DataFrame: # type: ignore
return self.eq(other)
def ne(self, other: typing.Any, axis: str | int = "columns") -> DataFrame:
return self._apply_binop(other, ops.ne_op, axis=axis)
def __ne__(self, other) -> DataFrame: # type: ignore
return self.ne(other)
def __invert__(self) -> DataFrame:
return self._apply_unary_op(ops.invert_op)
def le(self, other: typing.Any, axis: str | int = "columns") -> DataFrame:
return self._apply_binop(other, ops.le_op, axis=axis)
def __le__(self, other) -> DataFrame:
return self.le(other)
def lt(self, other: typing.Any, axis: str | int = "columns") -> DataFrame:
return self._apply_binop(other, ops.lt_op, axis=axis)
def __lt__(self, other) -> DataFrame:
return self.lt(other)
def ge(self, other: typing.Any, axis: str | int = "columns") -> DataFrame:
return self._apply_binop(other, ops.ge_op, axis=axis)
def __ge__(self, other) -> DataFrame:
return self.ge(other)