"docs/source/vscode:/vscode.git/clone" did not exist on "415e9a0980b00ef230d850bff7ecf0021c52640d"
testing_utils.py 33.1 KB
Newer Older
Sylvain Gugger's avatar
Sylvain Gugger committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# 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.

15
import inspect
16
import logging
17
import os
18
import re
19
import shutil
20
import sys
21
import tempfile
Aymeric Augustin's avatar
Aymeric Augustin committed
22
import unittest
23
from distutils.util import strtobool
24
from io import StringIO
25
from pathlib import Path
26

27
28
29
from .file_utils import (
    _datasets_available,
    _faiss_available,
30
    _flax_available,
NielsRogge's avatar
NielsRogge committed
31
32
    _pandas_available,
    _scatter_available,
33
34
35
36
37
38
    _sentencepiece_available,
    _tf_available,
    _tokenizers_available,
    _torch_available,
    _torch_tpu_available,
)
39
from .integrations import _has_optuna, _has_ray
40
41


Julien Chaumond's avatar
Julien Chaumond committed
42
SMALL_MODEL_IDENTIFIER = "julien-c/bert-xsmall-dummy"
Julien Chaumond's avatar
Julien Chaumond committed
43
DUMMY_UNKWOWN_IDENTIFIER = "julien-c/dummy-unknown"
44
DUMMY_DIFF_TOKENIZER_IDENTIFIER = "julien-c/dummy-diff-tokenizer"
Julien Chaumond's avatar
Julien Chaumond committed
45
# Used to test Auto{Config, Model, Tokenizer} model_type detection.
Julien Chaumond's avatar
Julien Chaumond committed
46
47


48
def parse_flag_from_env(key, default=False):
49
    try:
50
51
52
53
54
55
56
57
58
59
60
61
62
        value = os.environ[key]
    except KeyError:
        # KEY isn't set, default to `default`.
        _value = default
    else:
        # KEY is set, convert it to True or False.
        try:
            _value = strtobool(value)
        except ValueError:
            # More values are supported, but let's keep the message simple.
            raise ValueError("If set, {} must be yes or no.".format(key))
    return _value

63

Julien Chaumond's avatar
Julien Chaumond committed
64
65
66
67
68
69
70
71
72
73
74
75
76
def parse_int_from_env(key, default=None):
    try:
        value = os.environ[key]
    except KeyError:
        _value = default
    else:
        try:
            _value = int(value)
        except ValueError:
            raise ValueError("If set, {} must be a int.".format(key))
    return _value


77
_run_slow_tests = parse_flag_from_env("RUN_SLOW", default=False)
78
_run_pt_tf_cross_tests = parse_flag_from_env("RUN_PT_TF_CROSS_TESTS", default=False)
79
_run_custom_tokenizers = parse_flag_from_env("RUN_CUSTOM_TOKENIZERS", default=False)
80
_run_pipeline_tests = parse_flag_from_env("RUN_PIPELINE_TESTS", default=False)
81
_run_git_lfs_tests = parse_flag_from_env("RUN_GIT_LFS_TESTS", default=False)
Julien Chaumond's avatar
Julien Chaumond committed
82
_tf_gpu_memory_limit = parse_int_from_env("TF_GPU_MEMORY_LIMIT", default=None)
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
def is_pt_tf_cross_test(test_case):
    """
    Decorator marking a test as a test that control interactions between PyTorch and TensorFlow.

    PT+TF tests are skipped by default and we can run only them by setting RUN_PT_TF_CROSS_TESTS environment variable
    to a truthy value and selecting the is_pt_tf_cross_test pytest mark.

    """
    if not _run_pt_tf_cross_tests or not _torch_available or not _tf_available:
        return unittest.skip("test is PT+TF test")(test_case)
    else:
        try:
            import pytest  # We don't need a hard dependency on pytest in the main library
        except ImportError:
            return test_case
        else:
            return pytest.mark.is_pt_tf_cross_test()(test_case)


def is_pipeline_test(test_case):
    """
    Decorator marking a test as a pipeline test.

108
109
    Pipeline tests are skipped by default and we can run only them by setting RUN_PIPELINE_TESTS environment variable
    to a truthy value and selecting the is_pipeline_test pytest mark.
110
111
112
113
114
115
116
117
118
119
120
121
122

    """
    if not _run_pipeline_tests:
        return unittest.skip("test is pipeline test")(test_case)
    else:
        try:
            import pytest  # We don't need a hard dependency on pytest in the main library
        except ImportError:
            return test_case
        else:
            return pytest.mark.is_pipeline_test()(test_case)


123
124
125
126
def slow(test_case):
    """
    Decorator marking a test as slow.

Sylvain Gugger's avatar
Sylvain Gugger committed
127
    Slow tests are skipped by default. Set the RUN_SLOW environment variable to a truthy value to run them.
128
129
130

    """
    if not _run_slow_tests:
131
132
133
        return unittest.skip("test is slow")(test_case)
    else:
        return test_case
134
135


136
137
138
139
def custom_tokenizers(test_case):
    """
    Decorator marking a test for a custom tokenizer.

Sylvain Gugger's avatar
Sylvain Gugger committed
140
141
    Custom tokenizers require additional dependencies, and are skipped by default. Set the RUN_CUSTOM_TOKENIZERS
    environment variable to a truthy value to run them.
142
143
    """
    if not _run_custom_tokenizers:
144
145
146
        return unittest.skip("test of custom tokenizers")(test_case)
    else:
        return test_case
147
148


149
150
151
152
153
154
155
156
157
158
159
160
161
def require_git_lfs(test_case):
    """
    Decorator marking a test that requires git-lfs.

    git-lfs requires additional dependencies, and tests are skipped by default. Set the RUN_GIT_LFS_TESTS environment
    variable to a truthy value to run them.
    """
    if not _run_git_lfs_tests:
        return unittest.skip("test of git lfs workflow")(test_case)
    else:
        return test_case


162
163
164
165
166
167
168
169
def require_torch(test_case):
    """
    Decorator marking a test that requires PyTorch.

    These tests are skipped when PyTorch isn't installed.

    """
    if not _torch_available:
170
171
172
        return unittest.skip("test requires PyTorch")(test_case)
    else:
        return test_case
173
174
175
176
177
178
179
180
181
182


def require_tf(test_case):
    """
    Decorator marking a test that requires TensorFlow.

    These tests are skipped when TensorFlow isn't installed.

    """
    if not _tf_available:
183
184
185
        return unittest.skip("test requires TensorFlow")(test_case)
    else:
        return test_case
186
187


188
189
190
191
192
193
194
195
196
197
198
199
def require_flax(test_case):
    """
    Decorator marking a test that requires JAX & Flax

    These tests are skipped when one / both are not installed

    """
    if not _flax_available:
        test_case = unittest.skip("test requires JAX & Flax")(test_case)
    return test_case


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
def require_sentencepiece(test_case):
    """
    Decorator marking a test that requires SentencePiece.

    These tests are skipped when SentencePiece isn't installed.

    """
    if not _sentencepiece_available:
        return unittest.skip("test requires SentencePiece")(test_case)
    else:
        return test_case


def require_tokenizers(test_case):
    """
    Decorator marking a test that requires 🤗 Tokenizers.

    These tests are skipped when 🤗 Tokenizers isn't installed.

    """
    if not _tokenizers_available:
        return unittest.skip("test requires tokenizers")(test_case)
    else:
        return test_case


NielsRogge's avatar
NielsRogge committed
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
def require_pandas(test_case):
    """
    Decorator marking a test that requires pandas. These tests are skipped when pandas isn't installed.
    """
    if not _pandas_available:
        return unittest.skip("test requires pandas")(test_case)
    else:
        return test_case


def require_scatter(test_case):
    """
    Decorator marking a test that requires PyTorch Scatter. These tests are skipped when PyTorch Scatter isn't
    installed.
    """
    if not _scatter_available:
        return unittest.skip("test requires PyTorch Scatter")(test_case)
    else:
        return test_case


247
def require_torch_multi_gpu(test_case):
248
249
250
251
252
    """
    Decorator marking a test that requires a multi-GPU setup (in PyTorch).

    These tests are skipped on a machine without multiple GPUs.

253
    To run *only* the multi_gpu tests, assuming all test names contain multi_gpu: $ pytest -sv ./tests -k "multi_gpu"
254
255
256
257
258
259
260
261
    """
    if not _torch_available:
        return unittest.skip("test requires PyTorch")(test_case)

    import torch

    if torch.cuda.device_count() < 2:
        return unittest.skip("test requires multiple GPUs")(test_case)
262
263
    else:
        return test_case
264
265


266
def require_torch_non_multi_gpu(test_case):
267
268
269
270
271
272
273
274
275
276
    """
    Decorator marking a test that requires 0 or 1 GPU setup (in PyTorch).
    """
    if not _torch_available:
        return unittest.skip("test requires PyTorch")(test_case)

    import torch

    if torch.cuda.device_count() > 1:
        return unittest.skip("test requires 0 or 1 GPU")(test_case)
277
278
    else:
        return test_case
279
280


281
# this is a decorator identical to require_torch_non_multi_gpu, but is used as a quick band-aid to
282
283
# allow all of examples to be run multi-gpu CI and it reminds us that tests decorated with this one
# need to be ported and aren't so by design.
284
require_torch_non_multi_gpu_but_fix_me = require_torch_non_multi_gpu
285
286


Lysandre Debut's avatar
Lysandre Debut committed
287
288
289
290
291
292
def require_torch_tpu(test_case):
    """
    Decorator marking a test that requires a TPU (in PyTorch).
    """
    if not _torch_tpu_available:
        return unittest.skip("test requires PyTorch TPU")
293
294
    else:
        return test_case
Lysandre Debut's avatar
Lysandre Debut committed
295
296


297
if _torch_available:
Stas Bekman's avatar
Stas Bekman committed
298
299
300
301
    # Set env var CUDA_VISIBLE_DEVICES="" to force cpu-mode
    import torch

    torch_device = "cuda" if torch.cuda.is_available() else "cpu"
302
303
else:
    torch_device = None
304
305


306
def require_torch_gpu(test_case):
307
    """Decorator marking a test that requires CUDA and PyTorch. """
308
    if torch_device != "cuda":
309
        return unittest.skip("test requires CUDA")(test_case)
310
311
    else:
        return test_case
312
313


Ola Piktus's avatar
Ola Piktus committed
314
315
316
317
def require_datasets(test_case):
    """Decorator marking a test that requires datasets."""

    if not _datasets_available:
318
319
320
        return unittest.skip("test requires `datasets`")(test_case)
    else:
        return test_case
Ola Piktus's avatar
Ola Piktus committed
321
322
323
324
325


def require_faiss(test_case):
    """Decorator marking a test that requires faiss."""
    if not _faiss_available:
326
327
328
        return unittest.skip("test requires `faiss`")(test_case)
    else:
        return test_case
Ola Piktus's avatar
Ola Piktus committed
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
def require_optuna(test_case):
    """
    Decorator marking a test that requires optuna.

    These tests are skipped when optuna isn't installed.

    """
    if not _has_optuna:
        return unittest.skip("test requires optuna")(test_case)
    else:
        return test_case


def require_ray(test_case):
    """
    Decorator marking a test that requires Ray/tune.

    These tests are skipped when Ray/tune isn't installed.

    """
    if not _has_ray:
        return unittest.skip("test requires Ray/tune")(test_case)
    else:
        return test_case


357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
def get_gpu_count():
    """
    Return the number of available gpus (regardless of whether torch or tf is used)
    """
    if _torch_available:
        import torch

        return torch.cuda.device_count()
    elif _tf_available:
        import tensorflow as tf

        return len(tf.config.list_physical_devices("GPU"))
    else:
        return 0


373
def get_tests_dir(append_path=None):
374
    """
375
376
377
378
    Args:
        append_path: optional path to append to the tests dir path

    Return:
Sylvain Gugger's avatar
Sylvain Gugger committed
379
380
        The full path to the `tests` dir, so that the tests can be invoked from anywhere. Optionally `append_path` is
        joined after the `tests` dir the former is provided.
381

382
383
384
    """
    # this function caller's __file__
    caller__file__ = inspect.stack()[1][1]
385
386
387
388
389
    tests_dir = os.path.abspath(os.path.dirname(caller__file__))
    if append_path:
        return os.path.join(tests_dir, append_path)
    else:
        return tests_dir
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
#
# Helper functions for dealing with testing text outputs
# The original code came from:
# https://github.com/fastai/fastai/blob/master/tests/utils/text.py

# When any function contains print() calls that get overwritten, like progress bars,
# a special care needs to be applied, since under pytest -s captured output (capsys
# or contextlib.redirect_stdout) contains any temporary printed strings, followed by
# \r's. This helper function ensures that the buffer will contain the same output
# with and without -s in pytest, by turning:
# foo bar\r tar mar\r final message
# into:
# final message
# it can handle a single string or a multiline buffer
def apply_print_resets(buf):
    return re.sub(r"^.*\r", "", buf, 0, re.M)


def assert_screenout(out, what):
    out_pr = apply_print_resets(out).lower()
    match_str = out_pr.find(what.lower())
    assert match_str != -1, f"expecting to find {what} in output: f{out_pr}"


class CaptureStd:
Sylvain Gugger's avatar
Sylvain Gugger committed
417
418
419
    """
    Context manager to capture:
        stdout, clean it up and make it available via obj.out stderr, and make it available via obj.err
420

Sylvain Gugger's avatar
Sylvain Gugger committed
421
422
        init arguments: - out - capture stdout: True/False, default True - err - capture stdout: True/False, default
        True
423

Sylvain Gugger's avatar
Sylvain Gugger committed
424
        Examples::
425

Sylvain Gugger's avatar
Sylvain Gugger committed
426
427
428
            with CaptureStdout() as cs:
                print("Secret message")
            print(f"captured: {cs.out}")
429

Sylvain Gugger's avatar
Sylvain Gugger committed
430
431
432
433
            import sys
            with CaptureStderr() as cs:
                print("Warning: ", file=sys.stderr)
            print(f"captured: {cs.err}")
434

Sylvain Gugger's avatar
Sylvain Gugger committed
435
436
437
438
439
            # to capture just one of the streams, but not the other
            with CaptureStd(err=False) as cs:
                print("Secret message")
            print(f"captured: {cs.out}")
            # but best use the stream-specific subclasses
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

    """

    def __init__(self, out=True, err=True):
        if out:
            self.out_buf = StringIO()
            self.out = "error: CaptureStd context is unfinished yet, called too early"
        else:
            self.out_buf = None
            self.out = "not capturing stdout"

        if err:
            self.err_buf = StringIO()
            self.err = "error: CaptureStd context is unfinished yet, called too early"
        else:
            self.err_buf = None
            self.err = "not capturing stderr"

    def __enter__(self):
        if self.out_buf:
            self.out_old = sys.stdout
            sys.stdout = self.out_buf

        if self.err_buf:
            self.err_old = sys.stderr
            sys.stderr = self.err_buf

        return self

    def __exit__(self, *exc):
        if self.out_buf:
            sys.stdout = self.out_old
            self.out = apply_print_resets(self.out_buf.getvalue())

        if self.err_buf:
            sys.stderr = self.err_old
            self.err = self.err_buf.getvalue()

    def __repr__(self):
        msg = ""
        if self.out_buf:
            msg += f"stdout: {self.out}\n"
        if self.err_buf:
            msg += f"stderr: {self.err}\n"
        return msg


# in tests it's the best to capture only the stream that's wanted, otherwise
# it's easy to miss things, so unless you need to capture both streams, use the
# subclasses below (less typing). Or alternatively, configure `CaptureStd` to
# disable the stream you don't need to test.


class CaptureStdout(CaptureStd):
    """ Same as CaptureStd but captures only stdout """

    def __init__(self):
        super().__init__(err=False)


class CaptureStderr(CaptureStd):
    """ Same as CaptureStd but captures only stderr """

    def __init__(self):
        super().__init__(out=False)
505
506


507
class CaptureLogger:
Sylvain Gugger's avatar
Sylvain Gugger committed
508
509
    """
    Context manager to capture `logging` streams
510
511
512
513
514
515
516

    Args:
    - logger: 'logging` logger object

    Results:
        The captured output is available via `self.out`

517
    Example::
518

519
520
        >>> from transformers import logging
        >>> from transformers.testing_utils import CaptureLogger
521

522
523
        >>> msg = "Testing 1, 2, 3"
        >>> logging.set_verbosity_info()
524
        >>> logger = logging.get_logger("transformers.models.bart.tokenization_bart")
525
526
527
        >>> with CaptureLogger(logger) as cl:
        ...     logger.info(msg)
        >>> assert cl.out, msg+"\n"
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
    """

    def __init__(self, logger):
        self.logger = logger
        self.io = StringIO()
        self.sh = logging.StreamHandler(self.io)
        self.out = ""

    def __enter__(self):
        self.logger.addHandler(self.sh)
        return self

    def __exit__(self, *exc):
        self.logger.removeHandler(self.sh)
        self.out = self.io.getvalue()

    def __repr__(self):
        return f"captured: {self.out}\n"


548
class TestCasePlus(unittest.TestCase):
Sylvain Gugger's avatar
Sylvain Gugger committed
549
550
    """
    This class extends `unittest.TestCase` with additional features.
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
    Feature 1: A set of fully resolved important file and dir path accessors.

    In tests often we need to know where things are relative to the current test file, and it's not trivial since the
    test could be invoked from more than one directory or could reside in sub-directories with different depths. This
    class solves this problem by sorting out all the basic paths and provides easy accessors to them:

    * ``pathlib`` objects (all fully resolved):

       - ``test_file_path`` - the current test file path (=``__file__``)
       - ``test_file_dir`` - the directory containing the current test file
       - ``tests_dir`` - the directory of the ``tests`` test suite
       - ``examples_dir`` - the directory of the ``examples`` test suite
       - ``repo_root_dir`` - the directory of the repository
       - ``src_dir`` - the directory of ``src`` (i.e. where the ``transformers`` sub-dir resides)

    * stringified paths---same as above but these return paths as strings, rather than ``pathlib`` objects:

       - ``test_file_path_str``
       - ``test_file_dir_str``
       - ``tests_dir_str``
       - ``examples_dir_str``
       - ``repo_root_dir_str``
       - ``src_dir_str``

576
    Feature 2: Flexible auto-removable temporary dirs which are guaranteed to get removed at the end of test.
577

578
    1. Create a unique temporary dir:
579
580
581
582
583

    ::

        def test_whatever(self):
            tmp_dir = self.get_auto_remove_tmp_dir()
584

585
586
587
588
589
590
    ``tmp_dir`` will contain the path to the created temporary dir. It will be automatically removed at the end of the
    test.


    2. Create a temporary dir of my choice, ensure it's empty before the test starts and don't
    empty it after the test.
591
592
593
594

    ::

        def test_whatever(self):
595
            tmp_dir = self.get_auto_remove_tmp_dir("./xxx")
596

597
598
    This is useful for debug when you want to monitor a specific directory and want to make sure the previous tests
    didn't leave any data in there.
599

600
601
    3. You can override the first two options by directly overriding the ``before`` and ``after`` args, leading to the
       following behavior:
602

603
    ``before=True``: the temporary dir will always be cleared at the beginning of the test.
604

605
    ``before=False``: if the temporary dir already existed, any existing files will remain there.
606

607
608
609
    ``after=True``: the temporary dir will always be deleted at the end of the test.

    ``after=False``: the temporary dir will always be left intact at the end of the test.
610

611
612
613
    Note 1: In order to run the equivalent of ``rm -r`` safely, only subdirs of the project repository checkout are
    allowed if an explicit ``tmp_dir`` is used, so that by mistake no ``/tmp`` or similar important part of the
    filesystem will get nuked. i.e. please always pass paths that start with ``./``
614

615
616
    Note 2: Each test can register multiple temporary dirs and they all will get auto-removed, unless requested
    otherwise.
617

618
619
620
621
622
623
624
625
    Feature 3: Get a copy of the ``os.environ`` object that sets up ``PYTHONPATH`` specific to the current test suite.
    This is useful for invoking external programs from the test suite - e.g. distributed training.


    ::
        def test_whatever(self):
            env = self.get_env()

626
627
628
    """

    def setUp(self):
629
        # get_auto_remove_tmp_dir feature:
630
631
        self.teardown_tmp_dirs = []

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
        # figure out the resolved paths for repo_root, tests, examples, etc.
        self._test_file_path = inspect.getfile(self.__class__)
        path = Path(self._test_file_path).resolve()
        self._test_file_dir = path.parents[0]
        for up in [1, 2, 3]:
            tmp_dir = path.parents[up]
            if (tmp_dir / "src").is_dir() and (tmp_dir / "tests").is_dir():
                break
        if tmp_dir:
            self._repo_root_dir = tmp_dir
        else:
            raise ValueError(f"can't figure out the root of the repo from {self._test_file_path}")
        self._tests_dir = self._repo_root_dir / "tests"
        self._examples_dir = self._repo_root_dir / "examples"
        self._src_dir = self._repo_root_dir / "src"

    @property
    def test_file_path(self):
        return self._test_file_path

    @property
    def test_file_path_str(self):
        return str(self._test_file_path)

    @property
    def test_file_dir(self):
        return self._test_file_dir

    @property
    def test_file_dir_str(self):
        return str(self._test_file_dir)

    @property
    def tests_dir(self):
        return self._tests_dir

    @property
    def tests_dir_str(self):
        return str(self._tests_dir)

    @property
    def examples_dir(self):
        return self._examples_dir

    @property
    def examples_dir_str(self):
        return str(self._examples_dir)

    @property
    def repo_root_dir(self):
        return self._repo_root_dir

    @property
    def repo_root_dir_str(self):
        return str(self._repo_root_dir)

    @property
    def src_dir(self):
        return self._src_dir

    @property
    def src_dir_str(self):
        return str(self._src_dir)

    def get_env(self):
        """
        Return a copy of the ``os.environ`` object that sets up ``PYTHONPATH`` correctly, depending on the test suite
        it's invoked from. This is useful for invoking external programs from the test suite - e.g. distributed
        training.

        It always inserts ``./src`` first, then ``./tests`` or ``./examples`` depending on the test suite type and
        finally the preset ``PYTHONPATH`` if any (all full resolved paths).

        """
        env = os.environ.copy()
        paths = [self.src_dir_str]
        if "/examples" in self.test_file_dir_str:
            paths.append(self.examples_dir_str)
        else:
            paths.append(self.tests_dir_str)
        paths.append(env.get("PYTHONPATH", ""))

        env["PYTHONPATH"] = ":".join(paths)
        return env

717
    def get_auto_remove_tmp_dir(self, tmp_dir=None, before=None, after=None):
718
719
        """
        Args:
720
            tmp_dir (:obj:`string`, `optional`):
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
                if :obj:`None`:

                   - a unique temporary path will be created
                   - sets ``before=True`` if ``before`` is :obj:`None`
                   - sets ``after=True`` if ``after`` is :obj:`None`
                else:

                   - :obj:`tmp_dir` will be created
                   - sets ``before=True`` if ``before`` is :obj:`None`
                   - sets ``after=False`` if ``after`` is :obj:`None`
            before (:obj:`bool`, `optional`):
                If :obj:`True` and the :obj:`tmp_dir` already exists, make sure to empty it right away if :obj:`False`
                and the :obj:`tmp_dir` already exists, any existing files will remain there.
            after (:obj:`bool`, `optional`):
                If :obj:`True`, delete the :obj:`tmp_dir` at the end of the test if :obj:`False`, leave the
                :obj:`tmp_dir` and its contents intact at the end of the test.
737
738

        Returns:
739
            tmp_dir(:obj:`string`): either the same value as passed via `tmp_dir` or the path to the auto-selected tmp
Sylvain Gugger's avatar
Sylvain Gugger committed
740
            dir
741
742
        """
        if tmp_dir is not None:
743
744
745
746
747
748
749
750
751
752

            # defining the most likely desired behavior for when a custom path is provided.
            # this most likely indicates the debug mode where we want an easily locatable dir that:
            # 1. gets cleared out before the test (if it already exists)
            # 2. is left intact after the test
            if before is None:
                before = True
            if after is None:
                after = False

753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
            # using provided path
            path = Path(tmp_dir).resolve()

            # to avoid nuking parts of the filesystem, only relative paths are allowed
            if not tmp_dir.startswith("./"):
                raise ValueError(
                    f"`tmp_dir` can only be a relative path, i.e. `./some/path`, but received `{tmp_dir}`"
                )

            # ensure the dir is empty to start with
            if before is True and path.exists():
                shutil.rmtree(tmp_dir, ignore_errors=True)

            path.mkdir(parents=True, exist_ok=True)

        else:
769
770
771
772
773
774
775
776
777
            # defining the most likely desired behavior for when a unique tmp path is auto generated
            # (not a debug mode), here we require a unique tmp dir that:
            # 1. is empty before the test (it will be empty in this situation anyway)
            # 2. gets fully removed after the test
            if before is None:
                before = True
            if after is None:
                after = True

778
779
780
781
782
783
784
785
786
787
            # using unique tmp dir (always empty, regardless of `before`)
            tmp_dir = tempfile.mkdtemp()

        if after is True:
            # register for deletion
            self.teardown_tmp_dirs.append(tmp_dir)

        return tmp_dir

    def tearDown(self):
788
789

        # get_auto_remove_tmp_dir feature: remove registered temp dirs
790
791
792
        for path in self.teardown_tmp_dirs:
            shutil.rmtree(path, ignore_errors=True)
        self.teardown_tmp_dirs = []
793
794
795


def mockenv(**kwargs):
Sylvain Gugger's avatar
Sylvain Gugger committed
796
797
    """
    this is a convenience wrapper, that allows this:
798

Sylvain Gugger's avatar
Sylvain Gugger committed
799
800
    @mockenv(RUN_SLOW=True, USE_TF=False) def test_something(): run_slow = os.getenv("RUN_SLOW", False) use_tf =
    os.getenv("USE_TF", False)
801
802
    """
    return unittest.mock.patch.dict(os.environ, kwargs)
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
# --- pytest conf functions --- #

# to avoid multiple invocation from tests/conftest.py and examples/conftest.py - make sure it's called only once
pytest_opt_registered = {}


def pytest_addoption_shared(parser):
    """
    This function is to be called from `conftest.py` via `pytest_addoption` wrapper that has to be defined there.

    It allows loading both `conftest.py` files at once without causing a failure due to adding the same `pytest`
    option.

    """
    option = "--make-reports"
    if option not in pytest_opt_registered:
        parser.addoption(
            option,
            action="store",
            default=False,
            help="generate report files. The value of this option is used as a prefix to report names",
        )
        pytest_opt_registered[option] = 1


830
831
def pytest_terminal_summary_main(tr, id):
    """
Sylvain Gugger's avatar
Sylvain Gugger committed
832
833
    Generate multiple reports at the end of test suite run - each report goes into a dedicated file in the current
    directory. The report files are prefixed with the test suite name.
834
835
836

    This function emulates --duration and -rA pytest arguments.

Sylvain Gugger's avatar
Sylvain Gugger committed
837
838
    This function is to be called from `conftest.py` via `pytest_terminal_summary` wrapper that has to be defined
    there.
839
840
841
842

    Args:
    - tr: `terminalreporter` passed from `conftest.py`
    - id: unique id like `tests` or `examples` that will be incorporated into the final reports
Sylvain Gugger's avatar
Sylvain Gugger committed
843
      filenames - this is needed as some jobs have multiple runs of pytest, so we can't have them overwrite each other.
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861

    NB: this functions taps into a private _pytest API and while unlikely, it could break should
    pytest do internal changes - also it calls default internal methods of terminalreporter which
    can be hijacked by various `pytest-` plugins and interfere.

    """
    from _pytest.config import create_terminal_writer

    if not len(id):
        id = "tests"

    config = tr.config
    orig_writer = config.get_terminal_writer()
    orig_tbstyle = config.option.tbstyle
    orig_reportchars = tr.reportchars

    dir = "reports"
    Path(dir).mkdir(parents=True, exist_ok=True)
Stas Bekman's avatar
Stas Bekman committed
862
    report_files = {
863
        k: f"{dir}/{id}_{k}.txt"
Stas Bekman's avatar
Stas Bekman committed
864
865
866
867
868
869
870
871
872
873
874
875
        for k in [
            "durations",
            "errors",
            "failures_long",
            "failures_short",
            "failures_line",
            "passes",
            "stats",
            "summary_short",
            "warnings",
        ]
    }
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895

    # custom durations report
    # note: there is no need to call pytest --durations=XX to get this separate report
    # adapted from https://github.com/pytest-dev/pytest/blob/897f151e/src/_pytest/runner.py#L66
    dlist = []
    for replist in tr.stats.values():
        for rep in replist:
            if hasattr(rep, "duration"):
                dlist.append(rep)
    if dlist:
        dlist.sort(key=lambda x: x.duration, reverse=True)
        with open(report_files["durations"], "w") as f:
            durations_min = 0.05  # sec
            f.write("slowest durations\n")
            for i, rep in enumerate(dlist):
                if rep.duration < durations_min:
                    f.write(f"{len(dlist)-i} durations < {durations_min} secs were omitted")
                    break
                f.write(f"{rep.duration:02.2f}s {rep.when:<8} {rep.nodeid}\n")

Stas Bekman's avatar
Stas Bekman committed
896
897
898
899
900
901
902
903
904
905
906
907
908
909
    def summary_failures_short(tr):
        # expecting that the reports were --tb=long (default) so we chop them off here to the last frame
        reports = tr.getreports("failed")
        if not reports:
            return
        tr.write_sep("=", "FAILURES SHORT STACK")
        for rep in reports:
            msg = tr._getfailureheadline(rep)
            tr.write_sep("_", msg, red=True, bold=True)
            # chop off the optional leading extra frames, leaving only the last one
            longrepr = re.sub(r".*_ _ _ (_ ){10,}_ _ ", "", rep.longreprtext, 0, re.M | re.S)
            tr._tw.line(longrepr)
            # note: not printing out any rep.sections to keep the report short

910
911
912
913
    # use ready-made report funcs, we are just hijacking the filehandle to log to a dedicated file each
    # adapted from https://github.com/pytest-dev/pytest/blob/897f151e/src/_pytest/terminal.py#L814
    # note: some pytest plugins may interfere by hijacking the default `terminalreporter` (e.g.
    # pytest-instafail does that)
Stas Bekman's avatar
Stas Bekman committed
914
915
916
917

    # report failures with line/short/long styles
    config.option.tbstyle = "auto"  # full tb
    with open(report_files["failures_long"], "w") as f:
918
919
920
        tr._tw = create_terminal_writer(config, f)
        tr.summary_failures()

Stas Bekman's avatar
Stas Bekman committed
921
922
923
924
925
926
927
928
929
930
931
    # config.option.tbstyle = "short" # short tb
    with open(report_files["failures_short"], "w") as f:
        tr._tw = create_terminal_writer(config, f)
        summary_failures_short(tr)

    config.option.tbstyle = "line"  # one line per error
    with open(report_files["failures_line"], "w") as f:
        tr._tw = create_terminal_writer(config, f)
        tr.summary_failures()

    with open(report_files["errors"], "w") as f:
932
933
934
        tr._tw = create_terminal_writer(config, f)
        tr.summary_errors()

Stas Bekman's avatar
Stas Bekman committed
935
    with open(report_files["warnings"], "w") as f:
936
937
938
939
        tr._tw = create_terminal_writer(config, f)
        tr.summary_warnings()  # normal warnings
        tr.summary_warnings()  # final warnings

Stas Bekman's avatar
Stas Bekman committed
940
941
    tr.reportchars = "wPpsxXEf"  # emulate -rA (used in summary_passes() and short_test_summary())
    with open(report_files["passes"], "w") as f:
942
943
944
        tr._tw = create_terminal_writer(config, f)
        tr.summary_passes()

Stas Bekman's avatar
Stas Bekman committed
945
    with open(report_files["summary_short"], "w") as f:
946
947
948
        tr._tw = create_terminal_writer(config, f)
        tr.short_test_summary()

Stas Bekman's avatar
Stas Bekman committed
949
    with open(report_files["stats"], "w") as f:
950
951
952
953
954
955
956
        tr._tw = create_terminal_writer(config, f)
        tr.summary_stats()

    # restore:
    tr._tw = orig_writer
    tr.reportchars = orig_reportchars
    config.option.tbstyle = orig_tbstyle
957
958


959
# --- distributed testing functions --- #
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013

# adapted from https://stackoverflow.com/a/59041913/9201239
import asyncio  # noqa


class _RunOutput:
    def __init__(self, returncode, stdout, stderr):
        self.returncode = returncode
        self.stdout = stdout
        self.stderr = stderr


async def _read_stream(stream, callback):
    while True:
        line = await stream.readline()
        if line:
            callback(line)
        else:
            break


async def _stream_subprocess(cmd, env=None, stdin=None, timeout=None, quiet=False, echo=False) -> _RunOutput:
    if echo:
        print("\nRunning: ", " ".join(cmd))

    p = await asyncio.create_subprocess_exec(
        cmd[0],
        *cmd[1:],
        stdin=stdin,
        stdout=asyncio.subprocess.PIPE,
        stderr=asyncio.subprocess.PIPE,
        env=env,
    )

    # note: there is a warning for a possible deadlock when using `wait` with huge amounts of data in the pipe
    # https://docs.python.org/3/library/asyncio-subprocess.html#asyncio.asyncio.subprocess.Process.wait
    #
    # If it starts hanging, will need to switch to the following code. The problem is that no data
    # will be seen until it's done and if it hangs for example there will be no debug info.
    # out, err = await p.communicate()
    # return _RunOutput(p.returncode, out, err)

    out = []
    err = []

    def tee(line, sink, pipe, label=""):
        line = line.decode("utf-8").rstrip()
        sink.append(line)
        if not quiet:
            print(label, line, file=pipe)

    # XXX: the timeout doesn't seem to make any difference here
    await asyncio.wait(
        [
Stas Bekman's avatar
Stas Bekman committed
1014
            _read_stream(p.stdout, lambda l: tee(l, out, sys.stdout, label="stdout:")),
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
            _read_stream(p.stderr, lambda l: tee(l, err, sys.stderr, label="stderr:")),
        ],
        timeout=timeout,
    )
    return _RunOutput(await p.wait(), out, err)


def execute_subprocess_async(cmd, env=None, stdin=None, timeout=180, quiet=False, echo=True) -> _RunOutput:

    loop = asyncio.get_event_loop()
    result = loop.run_until_complete(
        _stream_subprocess(cmd, env=env, stdin=stdin, timeout=timeout, quiet=quiet, echo=echo)
    )

    cmd_str = " ".join(cmd)
    if result.returncode > 0:
1031
        stderr = "\n".join(result.stderr)
1032
        raise RuntimeError(
1033
1034
            f"'{cmd_str}' failed with returncode {result.returncode}\n\n"
            f"The combined stderr from workers follows:\n{stderr}"
1035
        )
Stas Bekman's avatar
Stas Bekman committed
1036
1037
1038
1039

    # check that the subprocess actually did run and produced some output, should the test rely on
    # the remote side to do the testing
    if not result.stdout and not result.stderr:
1040
1041
1042
        raise RuntimeError(f"'{cmd_str}' produced no output.")

    return result