/
pandas.py
1178 lines (1001 loc) · 45.9 KB
/
pandas.py
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
from __future__ import annotations
import io
import os
import typing as t
import logging
import functools
from enum import Enum
from typing import TYPE_CHECKING
from concurrent.futures import ThreadPoolExecutor
from starlette.requests import Request
from starlette.responses import Response
from .base import set_sample
from .base import IODescriptor
from ..types import LazyType
from ..utils.pkg import find_spec
from ..utils.http import set_cookies
from ...exceptions import BadInput
from ...exceptions import InvalidArgument
from ...exceptions import UnprocessableEntity
from ...exceptions import MissingDependencyException
from ..service.openapi import SUCCESS_DESCRIPTION
from ..utils.lazy_loader import LazyLoader
from ..service.openapi.specification import Schema
from ..service.openapi.specification import MediaType
EXC_MSG = "pandas' is required to use PandasDataFrame or PandasSeries. Install with 'pip install bentoml[io-pandas]'"
if TYPE_CHECKING:
import numpy as np
import pandas as pd
from typing_extensions import Self
from bentoml.grpc.v1alpha1 import service_pb2 as pb
from .. import external_typing as ext
from .base import OpenAPIResponse
from ..types import PathType
from ..context import InferenceApiContext as Context
else:
from bentoml.grpc.utils import import_generated_stubs
pb, _ = import_generated_stubs()
pd = LazyLoader("pd", globals(), "pandas", exc_msg=EXC_MSG)
np = LazyLoader("np", globals(), "numpy")
logger = logging.getLogger(__name__)
# Check for parquet support
@functools.lru_cache(maxsize=1)
def get_parquet_engine() -> str:
if find_spec("pyarrow") is not None:
return "pyarrow"
elif find_spec("fastparquet") is not None:
return "fastparquet"
else:
logger.warning(
"Neither pyarrow nor fastparquet packages found. Parquet de/serialization will not be available."
)
raise MissingDependencyException(
"Parquet serialization is not available. Try installing pyarrow or fastparquet first."
)
def _openapi_types(item: str) -> str: # pragma: no cover
# convert pandas types to OpenAPI types
if item.startswith("int"):
return "integer"
elif item.startswith("float") or item.startswith("double"):
return "number"
elif item.startswith("str") or item.startswith("date"):
return "string"
elif item.startswith("bool"):
return "boolean"
else:
return "object"
def _dataframe_openapi_schema(
dtype: bool | ext.PdDTypeArg | None,
orient: ext.DataFrameOrient = None,
) -> Schema: # pragma: no cover
if isinstance(dtype, dict):
if orient == "records":
return Schema(
type="array",
items=Schema(
type="object",
properties={
k: Schema(type=_openapi_types(v)) for k, v in dtype.items()
},
),
)
if orient == "index":
return Schema(
type="object",
additionalProperties=Schema(
type="object",
properties={
k: Schema(type=_openapi_types(v)) for k, v in dtype.items()
},
),
)
if orient == "columns":
return Schema(
type="object",
properties={
k: Schema(
type="object",
additionalProperties=Schema(type=_openapi_types(v)),
)
for k, v in dtype.items()
},
)
return Schema(
type="object",
properties={
k: Schema(type="array", items=Schema(type=_openapi_types(v)))
for k, v in dtype.items()
},
)
else:
return Schema(type="object")
def _series_openapi_schema(
dtype: bool | ext.PdDTypeArg | None, orient: ext.SeriesOrient = None
) -> Schema: # pragma: no cover
if isinstance(dtype, str):
if orient in ["index", "values"]:
return Schema(
type="object", additionalProperties=Schema(type=_openapi_types(dtype))
)
if orient in ["records", "columns"]:
return Schema(type="array", items=Schema(type=_openapi_types(dtype)))
return Schema(type="object")
class SerializationFormat(Enum):
JSON = "application/json"
PARQUET = "application/octet-stream"
CSV = "text/csv"
def __init__(self, mime_type: str):
self.mime_type = mime_type
def __str__(self) -> str:
if self == SerializationFormat.JSON:
return "json"
elif self == SerializationFormat.PARQUET:
return "parquet"
elif self == SerializationFormat.CSV:
return "csv"
else:
raise ValueError(f"Unknown serialization format: {self}")
def _infer_serialization_format_from_request(
request: Request, default_format: SerializationFormat
) -> SerializationFormat:
"""Determine the serialization format from the request's headers['content-type']"""
content_type = request.headers.get("content-type")
if content_type == "application/json":
return SerializationFormat.JSON
elif content_type == "application/octet-stream":
return SerializationFormat.PARQUET
elif content_type == "text/csv":
return SerializationFormat.CSV
elif content_type:
logger.debug(
"Unknown Content-Type ('%s'), falling back to '%s' serialization format.",
content_type,
default_format,
)
return default_format
else:
logger.debug(
"Content-Type not specified, falling back to '%s' serialization format.",
default_format,
)
return default_format
def _validate_serialization_format(serialization_format: SerializationFormat):
if (
serialization_format is SerializationFormat.PARQUET
and get_parquet_engine() is None
):
raise MissingDependencyException(
"Parquet serialization is not available. Try installing pyarrow or fastparquet first."
)
class PandasDataFrame(
IODescriptor["ext.PdDataFrame"], descriptor_id="bentoml.io.PandasDataFrame"
):
"""
:obj:`PandasDataFrame` defines API specification for the inputs/outputs of a Service,
where either inputs will be converted to or outputs will be converted from type
:code:`pd.DataFrame` as specified in your API function signature.
A sample service implementation:
.. code-block:: python
:caption: `service.py`
from __future__ import annotations
import bentoml
import pandas as pd
import numpy as np
from bentoml.io import PandasDataFrame
input_spec = PandasDataFrame.from_sample(pd.DataFrame(np.array([[5,4,3,2]])))
runner = bentoml.sklearn.get("sklearn_model_clf").to_runner()
svc = bentoml.Service("iris-classifier", runners=[runner])
@svc.api(input=input_spec, output=PandasDataFrame())
def predict(input_arr):
res = runner.run(input_arr)
return pd.DataFrame(res)
Users then can then serve this service with :code:`bentoml serve`:
.. code-block:: bash
% bentoml serve ./service.py:svc --reload
Users can then send requests to the newly started services with any client:
.. tab-set::
.. tab-item:: Bash
.. code-block:: bash
% curl -X POST -H "Content-Type: application/json" \\
--data '[{"0":5,"1":4,"2":3,"3":2}]' http://0.0.0.0:3000/predict
# [{"0": 1}]%
.. tab-item:: Python
.. code-block:: python
:caption: `request.py`
import requests
requests.post(
"http://0.0.0.0:3000/predict",
headers={"content-type": "application/json"},
data='[{"0":5,"1":4,"2":3,"3":2}]'
).text
Args:
orient: Indication of expected JSON string format. Compatible JSON strings can be
produced by :func:`pandas.io.json.to_json()` with a corresponding orient value.
Possible orients are:
- :obj:`split` - :code:`dict[str, Any]` ↦ {``idx`` ↠ ``[idx]``, ``columns`` ↠ ``[columns]``, ``data`` ↠ ``[values]``}
- :obj:`records` - :code:`list[Any]` ↦ [{``column`` ↠ ``value``}, ..., {``column`` ↠ ``value``}]
- :obj:`index` - :code:`dict[str, Any]` ↦ {``idx`` ↠ {``column`` ↠ ``value``}}
- :obj:`columns` - :code:`dict[str, Any]` ↦ {``column`` ↠ {``index`` ↠ ``value``}}
- :obj:`values` - :code:`dict[str, Any]` ↦ Values arrays
columns: List of columns name that users wish to update.
apply_column_names: Whether to update incoming DataFrame columns. If :code:`apply_column_names=True`,
then ``columns`` must be specified.
dtype: Data type users wish to convert their inputs/outputs to. If it is a boolean,
then pandas will infer dtypes. Else if it is a dictionary of column to
``dtype``, then applies those to incoming dataframes. If ``False``, then don't
infer dtypes at all (only applies to the data). This is not applicable for :code:`orient='table'`.
enforce_dtype: Whether to enforce a certain data type. if :code:`enforce_dtype=True` then :code:`dtype` must be specified.
shape: Optional shape check that users can specify for their incoming HTTP
requests. We will only check the number of columns you specified for your
given shape:
.. code-block:: python
:caption: `service.py`
import pandas as pd
from bentoml.io import PandasDataFrame
df = pd.DataFrame([[1, 2, 3]]) # shape (1,3)
inp = PandasDataFrame.from_sample(df)
@svc.api(
input=PandasDataFrame(shape=(51, 10),
enforce_shape=True),
output=PandasDataFrame()
)
def predict(input_df: pd.DataFrame) -> pd.DataFrame:
# if input_df have shape (40,9),
# it will throw out errors
...
enforce_shape: Whether to enforce a certain shape. If ``enforce_shape=True`` then ``shape`` must be specified.
default_format: The default serialization format to use if the request does not specify a ``Content-Type`` Headers.
It is also the serialization format used for the response. Possible values are:
- :obj:`json` - JSON text format (inferred from content-type ``"application/json"``)
- :obj:`parquet` - Parquet binary format (inferred from content-type ``"application/octet-stream"``)
- :obj:`csv` - CSV text format (inferred from content-type ``"text/csv"``)
Returns:
:obj:`PandasDataFrame`: IO Descriptor that represents a :code:`pd.DataFrame`.
"""
_proto_fields = ("dataframe",)
def __init__(
self,
orient: ext.DataFrameOrient = "records",
columns: list[str] | None = None,
apply_column_names: bool = False,
dtype: bool | ext.PdDTypeArg | None = None,
enforce_dtype: bool = False,
shape: tuple[int, ...] | None = None,
enforce_shape: bool = False,
default_format: t.Literal["json", "parquet", "csv"] = "json",
):
self._orient: ext.DataFrameOrient = orient
self._columns = columns
self._apply_column_names = apply_column_names
# TODO: convert dtype to numpy dtype
self._dtype = dtype
self._enforce_dtype = enforce_dtype
self._shape = shape
self._enforce_shape = enforce_shape
self._default_format = SerializationFormat[default_format.upper()]
_validate_serialization_format(self._default_format)
self._mime_type = self._default_format.mime_type
@classmethod
def from_sample(
cls,
sample: ext.PdDataFrame | PathType | ext.NpNDArray,
*,
orient: ext.DataFrameOrient = "records",
apply_column_names: bool = True,
enforce_shape: bool = True,
enforce_dtype: bool = True,
default_format: t.Literal["json", "parquet", "csv"] = "json",
) -> Self:
"""
Create a :obj:`PandasDataFrame` IO Descriptor from given inputs.
Args:
sample: Given sample ``pd.DataFrame`` data
orient: Indication of expected JSON string format. Compatible JSON strings can be
produced by :func:`pandas.io.json.to_json()` with a corresponding orient value.
Possible orients are:
- :obj:`split` - :code:`dict[str, Any]` ↦ {``idx`` ↠ ``[idx]``, ``columns`` ↠ ``[columns]``, ``data`` ↠ ``[values]``}
- :obj:`records` - :code:`list[Any]` ↦ [{``column`` ↠ ``value``}, ..., {``column`` ↠ ``value``}]
- :obj:`index` - :code:`dict[str, Any]` ↦ {``idx`` ↠ {``column`` ↠ ``value``}}
- :obj:`columns` - :code:`dict[str, Any]` ↦ {``column`` ↠ {``index`` ↠ ``value``}}
- :obj:`values` - :code:`dict[str, Any]` ↦ Values arrays
- :obj:`table` - :code:`dict[str, Any]` ↦ {``schema``: { schema }, ``data``: { data }}
apply_column_names: Update incoming DataFrame columns. ``columns`` must be specified at
function signature. If you don't want to enforce a specific columns
name then change ``apply_column_names=False``.
enforce_dtype: Enforce a certain data type. `dtype` must be specified at function
signature. If you don't want to enforce a specific dtype then change
``enforce_dtype=False``.
enforce_shape: Enforce a certain shape. ``shape`` must be specified at function
signature. If you don't want to enforce a specific shape then change
``enforce_shape=False``.
default_format: The default serialization format to use if the request does not specify a ``Content-Type`` Headers.
It is also the serialization format used for the response. Possible values are:
- :obj:`json` - JSON text format (inferred from content-type ``"application/json"``)
- :obj:`parquet` - Parquet binary format (inferred from content-type ``"application/octet-stream"``)
- :obj:`csv` - CSV text format (inferred from content-type ``"text/csv"``)
Returns:
:obj:`PandasDataFrame`: :code:`PandasDataFrame` IODescriptor from given users inputs.
Example:
.. code-block:: python
:caption: `service.py`
import pandas as pd
from bentoml.io import PandasDataFrame
arr = [[1,2,3]]
input_spec = PandasDataFrame.from_sample(pd.DataFrame(arr))
@svc.api(input=input_spec, output=PandasDataFrame())
def predict(inputs: pd.DataFrame) -> pd.DataFrame: ...
"""
if LazyType["ext.NpNDArray"]("numpy", "ndarray").isinstance(sample):
@set_sample.register(np.ndarray)
def _(cls: Self, sample: ext.NpNDArray):
if isinstance(cls, PandasDataFrame):
__ = pd.DataFrame(sample)
cls.sample = __
cls._shape = __.shape
cls._columns = [str(i) for i in range(sample.shape[1])]
return super().from_sample(
sample,
dtype=True, # set to True to infer from given input
orient=orient,
enforce_shape=enforce_shape,
enforce_dtype=enforce_dtype,
apply_column_names=apply_column_names,
default_format=default_format,
)
@set_sample.register(pd.DataFrame)
def _(cls, sample: pd.DataFrame):
if isinstance(cls, PandasDataFrame):
cls.sample = sample
cls._shape = sample.shape
cls._columns = [str(x) for x in list(sample.columns)]
@set_sample.register(str)
@set_sample.register(os.PathLike)
def _(cls, sample: str):
if isinstance(cls, PandasDataFrame):
try:
if os.path.exists(sample):
try:
ext = os.path.splitext(sample)[-1].strip(".")
__ = getattr(
pd,
{
"json": "read_json",
"csv": "read_csv",
"html": "read_html",
"xls": "read_excel",
"xlsx": "read_excel",
"hdf5": "read_hdf",
"parquet": "read_parquet",
"pickle": "read_pickle",
"sql": "read_sql",
}[ext],
)(sample)
cls.sample = __
cls._shape = __.shape
cls._columns = [str(x) for x in list(__.columns)]
except KeyError:
raise InvalidArgument(f"Unsupported sample '{sample}' format.")
else:
__ = pd.read_json(sample)
cls.sample = __
cls._shape = __.shape
cls._columns = [str(x) for x in list(__.columns)]
except ValueError as e:
raise InvalidArgument(
f"Failed to create a 'pd.DataFrame' from sample {sample}: {e}"
) from None
def _convert_dtype(
self, value: ext.PdDTypeArg | None
) -> str | dict[str, t.Any] | None:
# TODO: support extension dtypes
if LazyType["ext.NpNDArray"]("numpy", "ndarray").isinstance(value):
return str(value.dtype)
elif isinstance(value, bool):
return str(value)
elif isinstance(value, dict):
return {str(k): self._convert_dtype(v) for k, v in value.items()}
elif value is None:
return "null"
else:
logger.warning(f"{type(value)} is not yet supported.")
return None
def to_spec(self) -> dict[str, t.Any]:
# TODO: support extension dtypes
dtype = None
if self._dtype is not None:
if isinstance(self._dtype, bool):
dtype = self._dtype
else:
dtype = self._dtype.name
return {
"id": self.descriptor_id,
"args": {
"orient": self._orient,
"columns": self._columns,
"dtype": self._convert_dtype(dtype),
"shape": self._shape,
"enforce_dtype": self._enforce_dtype,
"enforce_shape": self._enforce_shape,
"default_format": str(self._default_format),
},
}
@classmethod
def from_spec(cls, spec: dict[str, t.Any]) -> Self:
if "args" not in spec:
raise InvalidArgument(f"Missing args key in PandasDataFrame spec: {spec}")
res = PandasDataFrame(**spec["args"])
return res
def input_type(self) -> LazyType[ext.PdDataFrame]:
return LazyType("pandas", "DataFrame")
def openapi_schema(self) -> Schema:
return _dataframe_openapi_schema(self._dtype, self._orient)
def openapi_components(self) -> dict[str, t.Any] | None:
pass
def openapi_example(self):
if self.sample is not None:
return self.sample.to_json(orient=self._orient)
def openapi_request_body(self) -> dict[str, t.Any]:
return {
"content": {
self._mime_type: MediaType(
schema=self.openapi_schema(), example=self.openapi_example()
)
},
"required": True,
"x-bentoml-io-descriptor": self.to_spec(),
}
def openapi_responses(self) -> OpenAPIResponse:
return {
"description": SUCCESS_DESCRIPTION,
"content": {
self._mime_type: MediaType(
schema=self.openapi_schema(), example=self.openapi_example()
)
},
"x-bentoml-io-descriptor": self.to_spec(),
}
async def from_http_request(self, request: Request) -> ext.PdDataFrame:
"""
Process incoming requests and convert incoming
objects to `pd.DataFrame`
Args:
request (`starlette.requests.Requests`):
Incoming Requests
Returns:
a `pd.DataFrame` object. This can then be used
inside users defined logics.
Raises:
BadInput:
Raised when the incoming requests are bad formatted.
"""
serialization_format = _infer_serialization_format_from_request(
request, self._default_format
)
_validate_serialization_format(serialization_format)
obj = await request.body()
if serialization_format is SerializationFormat.JSON:
assert not isinstance(self._dtype, bool)
res = pd.read_json(io.BytesIO(obj), dtype=self._dtype, orient=self._orient)
elif serialization_format is SerializationFormat.PARQUET:
res = pd.read_parquet(io.BytesIO(obj), engine=get_parquet_engine())
elif serialization_format is SerializationFormat.CSV:
assert not isinstance(self._dtype, bool)
res: ext.PdDataFrame = pd.read_csv(io.BytesIO(obj), dtype=self._dtype)
else:
raise InvalidArgument(
f"Unknown serialization format ({serialization_format})."
) from None
assert isinstance(res, pd.DataFrame)
return self.validate_dataframe(res)
async def to_http_response(
self, obj: ext.PdDataFrame, ctx: Context | None = None
) -> Response:
"""
Process given objects and convert it to HTTP response.
Args:
obj (`pd.DataFrame`):
`pd.DataFrame` that will be serialized to JSON or parquet
Returns:
HTTP Response of type `starlette.responses.Response`. This can
be accessed via cURL or any external web traffic.
"""
obj = self.validate_dataframe(obj)
# For the response it doesn't make sense to enforce the same serialization format as specified
# by the request's headers['content-type']. Instead we simply use the _default_format.
serialization_format = self._default_format
if not LazyType["ext.PdDataFrame"](pd.DataFrame).isinstance(obj):
raise InvalidArgument(
f"return object is not of type `pd.DataFrame`, got type {type(obj)} instead"
)
if serialization_format is SerializationFormat.JSON:
resp = obj.to_json(orient=self._orient)
elif serialization_format is SerializationFormat.PARQUET:
resp = obj.to_parquet(engine=get_parquet_engine())
elif serialization_format is SerializationFormat.CSV:
resp = obj.to_csv()
else:
raise InvalidArgument(
f"Unknown serialization format ({serialization_format})."
) from None
if ctx is not None:
res = Response(
resp,
media_type=serialization_format.mime_type,
headers=ctx.response.headers, # type: ignore (bad starlette types)
status_code=ctx.response.status_code,
)
set_cookies(res, ctx.response.cookies)
return res
else:
return Response(resp, media_type=serialization_format.mime_type)
def validate_dataframe(
self, dataframe: ext.PdDataFrame, exception_cls: t.Type[Exception] = BadInput
) -> ext.PdDataFrame:
if not LazyType["ext.PdDataFrame"]("pandas.core.frame.DataFrame").isinstance(
dataframe
):
raise InvalidArgument(
f"return object is not of type 'pd.DataFrame', got type '{type(dataframe)}' instead"
) from None
# TODO: dtype check
# if self._dtype is not None and self._dtype != dataframe.dtypes:
# msg = f'{self.__class__.__name__}: Expecting DataFrame of dtype "{self._dtype}", but "{dataframe.dtypes}" was received.'
# if self._enforce_dtype:
# raise exception_cls(msg) from None
if self._columns is not None and len(self._columns) != dataframe.shape[1]:
msg = "length of 'columns' (%d) does not match the # of columns of incoming data."
if self._apply_column_names:
raise BadInput(msg % len(self._columns)) from None
else:
logger.debug(msg, len(self._columns))
dataframe.columns = pd.Index(self._columns)
# TODO: convert from wide to long format (melt())
if self._shape is not None and self._shape != dataframe.shape:
msg = f'{self.__class__.__name__}: Expecting DataFrame of shape "{self._shape}", but "{dataframe.shape}" was received.'
if self._enforce_shape and not all(
left == right
for left, right in zip(self._shape, dataframe.shape)
if left != -1 and right != -1
):
raise exception_cls(msg) from None
return dataframe
async def from_proto(self, field: pb.DataFrame | bytes) -> ext.PdDataFrame:
"""
Process incoming protobuf request and convert it to ``pandas.DataFrame``
Args:
request: Incoming RPC request message.
context: grpc.ServicerContext
Returns:
a ``pandas.DataFrame`` object. This can then be used
inside users defined logics.
"""
# TODO: support different serialization format
if isinstance(field, bytes):
# TODO: handle serialized_bytes for dataframe
raise NotImplementedError(
'Currently not yet implemented. Use "dataframe" instead.'
)
else:
# note that there is a current bug where we don't check for
# dtype of given fields per Series to match with types of a given
# columns, hence, this would result in a wrong DataFrame that is not
# expected by our users.
assert isinstance(field, pb.DataFrame)
# columns orient: { column_name : {index : columns.series._value}}
if self._orient != "columns":
raise BadInput(
f"'dataframe' field currently only supports 'columns' orient. Make sure to set 'orient=columns' in {self.__class__.__name__}."
) from None
data: list[t.Any] = []
def process_columns_contents(content: pb.Series) -> dict[str, t.Any]:
# To be use inside a ThreadPoolExecutor to handle
# large tabular data
if len(content.ListFields()) != 1:
raise BadInput(
f"Array contents can only be one of given values key. Use one of '{list(map(lambda f: f[0].name,content.ListFields()))}' instead."
) from None
return {str(i): c for i, c in enumerate(content.ListFields()[0][1])}
with ThreadPoolExecutor(max_workers=10) as executor:
futures = executor.map(process_columns_contents, field.columns)
data.extend([i for i in list(futures)])
dataframe = pd.DataFrame(
dict(zip(field.column_names, data)),
columns=t.cast(t.List[str], field.column_names),
)
return self.validate_dataframe(dataframe)
async def to_proto(self, obj: ext.PdDataFrame) -> pb.DataFrame:
"""
Process given objects and convert it to grpc protobuf response.
Args:
obj: ``pandas.DataFrame`` that will be serialized to protobuf
context: grpc.aio.ServicerContext from grpc.aio.Server
Returns:
``service_pb2.Response``:
Protobuf representation of given ``pandas.DataFrame``
"""
from .numpy import npdtype_to_fieldpb_map
# TODO: support different serialization format
obj = self.validate_dataframe(obj)
mapping = npdtype_to_fieldpb_map()
# note that this is not safe, since we are not checking the dtype of the series
# FIXME(aarnphm): validate and handle mix columns dtype
# currently we don't support ExtensionDtype
columns_name: list[str] = list(map(str, obj.columns))
not_supported: list[ext.PdDType] = list(
filter(
lambda x: x not in mapping,
map(lambda x: t.cast("ext.PdSeries", obj[x]).dtype, columns_name),
)
)
if len(not_supported) > 0:
raise UnprocessableEntity(
f'dtype in column "{obj.columns}" is not currently supported.'
) from None
return pb.DataFrame(
column_names=columns_name,
columns=[
pb.Series(
**{mapping[t.cast("ext.NpDTypeLike", obj[col].dtype)]: obj[col]}
)
for col in columns_name
],
)
class PandasSeries(
IODescriptor["ext.PdSeries"], descriptor_id="bentoml.io.PandasSeries"
):
"""
:code:`PandasSeries` defines API specification for the inputs/outputs of a Service, where
either inputs will be converted to or outputs will be converted from type
:code:`pd.Series` as specified in your API function signature.
A sample service implementation:
.. code-block:: python
:caption: `service.py`
import bentoml
import pandas as pd
import numpy as np
from bentoml.io import PandasSeries
runner = bentoml.sklearn.get("sklearn_model_clf").to_runner()
svc = bentoml.Service("iris-classifier", runners=[runner])
@svc.api(input=PandasSeries(), output=PandasSeries())
def predict(input_arr):
res = runner.run(input_arr) # type: np.ndarray
return pd.Series(res)
Users then can then serve this service with :code:`bentoml serve`:
.. code-block:: bash
% bentoml serve ./service.py:svc --reload
Users can then send requests to the newly started services with any client:
.. tab-set::
.. tab-item:: Bash
.. code-block:: bash
% curl -X POST -H "Content-Type: application/json" \\
--data '[{"0":5,"1":4,"2":3,"3":2}]' http://0.0.0.0:3000/predict
# [{"0": 1}]%
.. tab-item:: Python
.. code-block:: python
:caption: `request.py`
import requests
requests.post(
"http://0.0.0.0:3000/predict",
headers={"content-type": "application/json"},
data='[{"0":5,"1":4,"2":3,"3":2}]'
).text
Args:
orient: Indication of expected JSON string format. Compatible JSON strings can be
produced by :func:`pandas.io.json.to_json()` with a corresponding orient value.
Possible orients are:
- :obj:`split` - :code:`dict[str, Any]` ↦ {``idx`` ↠ ``[idx]``, ``columns`` ↠ ``[columns]``, ``data`` ↠ ``[values]``}
- :obj:`records` - :code:`list[Any]` ↦ [{``column`` ↠ ``value``}, ..., {``column`` ↠ ``value``}]
- :obj:`index` - :code:`dict[str, Any]` ↦ {``idx`` ↠ {``column`` ↠ ``value``}}
- :obj:`columns` - :code:`dict[str, Any]` ↦ {``column`` ↠ {``index`` ↠ ``value``}}
- :obj:`values` - :code:`dict[str, Any]` ↦ Values arrays
columns: List of columns name that users wish to update.
apply_column_names (`bool`, `optional`, default to :code:`False`):
apply_column_names: Whether to update incoming DataFrame columns. If :code:`apply_column_names=True`,
then ``columns`` must be specified.
dtype: Data type users wish to convert their inputs/outputs to. If it is a boolean,
then pandas will infer dtypes. Else if it is a dictionary of column to
``dtype``, then applies those to incoming dataframes. If ``False``, then don't
infer dtypes at all (only applies to the data). This is not applicable for :code:`orient='table'`.
enforce_dtype: Whether to enforce a certain data type. if :code:`enforce_dtype=True` then :code:`dtype` must be specified.
shape: Optional shape check that users can specify for their incoming HTTP
requests. We will only check the number of columns you specified for your
given shape:
.. code-block:: python
:caption: `service.py`
import pandas as pd
from bentoml.io import PandasSeries
@svc.api(input=PandasSeries(shape=(51,), enforce_shape=True), output=PandasSeries())
def infer(input_series: pd.Series) -> pd.Series:
# if input_series has shape (40,), it will error
...
enforce_shape: Whether to enforce a certain shape. If ``enforce_shape=True`` then ``shape`` must be specified.
Returns:
:obj:`PandasSeries`: IO Descriptor that represents a :code:`pd.Series`.
"""
_proto_fields = ("series",)
_mime_type = "application/json"
def __init__(
self,
orient: ext.SeriesOrient = "records",
dtype: ext.PdDTypeArg | None = None,
enforce_dtype: bool = False,
shape: tuple[int, ...] | None = None,
enforce_shape: bool = False,
):
self._orient: ext.SeriesOrient = orient
self._dtype = dtype
self._enforce_dtype = enforce_dtype
self._shape = shape
self._enforce_shape = enforce_shape
@classmethod
def from_sample(
cls,
sample: ext.PdSeries | t.Sequence[t.Any],
*,
orient: ext.SeriesOrient = "records",
enforce_shape: bool = True,
enforce_dtype: bool = True,
) -> Self:
"""
Create a :obj:`PandasSeries` IO Descriptor from given inputs.
Args:
sample_input: Given sample ``pd.DataFrame`` data
orient: Indication of expected JSON string format. Compatible JSON strings can be
produced by :func:`pandas.io.json.to_json()` with a corresponding orient value.
Possible orients are:
- :obj:`split` - :code:`dict[str, Any]` ↦ {``idx`` ↠ ``[idx]``, ``columns`` ↠ ``[columns]``, ``data`` ↠ ``[values]``}
- :obj:`records` - :code:`list[Any]` ↦ [{``column`` ↠ ``value``}, ..., {``column`` ↠ ``value``}]
- :obj:`index` - :code:`dict[str, Any]` ↦ {``idx`` ↠ {``column`` ↠ ``value``}}
- :obj:`table` - :code:`dict[str, Any]` ↦ {``schema``: { schema }, ``data``: { data }}
enforce_dtype: Enforce a certain data type. `dtype` must be specified at function
signature. If you don't want to enforce a specific dtype then change
``enforce_dtype=False``.
enforce_shape: Enforce a certain shape. ``shape`` must be specified at function
signature. If you don't want to enforce a specific shape then change
``enforce_shape=False``.
Returns:
:obj:`PandasSeries`: :code:`PandasSeries` IODescriptor from given users inputs.
Example:
.. code-block:: python
:caption: `service.py`
import pandas as pd
from bentoml.io import PandasSeries
arr = [1,2,3]
input_spec = PandasSeries.from_sample(pd.DataFrame(arr))
@svc.api(input=input_spec, output=PandasSeries())
def predict(inputs: pd.Series) -> pd.Series: ...
"""
if LazyType["ext.NpNDArray"]("numpy", "ndarray").isinstance(sample):
@set_sample.register(np.ndarray)
def _(cls: Self, sample: ext.NpNDArray):
if isinstance(cls, PandasSeries):
__ = pd.Series(sample)
cls.sample = __
cls._dtype = __.dtype
cls._shape = __.shape
return super().from_sample(
sample,
orient=orient,
enforce_dtype=enforce_dtype,
enforce_shape=enforce_shape,
)
@set_sample.register(pd.Series)
def _(cls, sample: ext.PdSeries):
if isinstance(cls, PandasSeries):
cls.sample = sample
cls._dtype = sample.dtype
cls._shape = sample.shape
@set_sample.register(list)
@set_sample.register(tuple)
@set_sample.register(set)
def _(cls, sample: t.Sequence[t.Any]):
if isinstance(cls, PandasSeries):
__ = pd.Series(sample)
cls.sample = __
cls._dtype = __.dtype
cls._shape = __.shape
def input_type(self) -> LazyType[ext.PdSeries]:
return LazyType("pandas", "Series")
def _convert_dtype(
self, value: ext.PdDTypeArg | None
) -> str | dict[str, t.Any] | None:
# TODO: support extension dtypes
if LazyType["ext.NpNDArray"]("numpy", "ndarray").isinstance(value):
return str(value.dtype)
elif isinstance(value, bool):
return str(value)
elif isinstance(value, dict):
return {str(k): self._convert_dtype(v) for k, v in value.items()}
elif value is None:
return "null"
else:
logger.warning(f"{type(value)} is not yet supported.")
return None
def to_spec(self) -> dict[str, t.Any]:
return {
"id": self.descriptor_id,
"args": {
"orient": self._orient,
"dtype": self._convert_dtype(self._dtype),
"shape": self._shape,
"enforce_dtype": self._enforce_dtype,
"enforce_shape": self._enforce_shape,
},
}
@classmethod
def from_spec(cls, spec: dict[str, t.Any]) -> Self:
if "args" not in spec:
raise InvalidArgument(f"Missing args key in PandasSeries spec: {spec}")
res = PandasSeries(**spec["args"])
return res
def openapi_schema(self) -> Schema:
return _series_openapi_schema(self._dtype, self._orient)
def openapi_components(self) -> dict[str, t.Any] | None:
pass
def openapi_example(self):
if self.sample is not None:
return self.sample.to_json(orient=self._orient)
def openapi_request_body(self) -> dict[str, t.Any]:
return {
"content": {