/
data.py
2584 lines (2190 loc) · 90.6 KB
/
data.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
# This file is part of Hypothesis, which may be found at
# https://github.com/HypothesisWorks/hypothesis/
#
# Copyright the Hypothesis Authors.
# Individual contributors are listed in AUTHORS.rst and the git log.
#
# This Source Code Form is subject to the terms of the Mozilla Public License,
# v. 2.0. If a copy of the MPL was not distributed with this file, You can
# obtain one at https://mozilla.org/MPL/2.0/.
import abc
import contextlib
import math
import time
from collections import defaultdict
from enum import IntEnum
from random import Random
from sys import float_info
from typing import (
TYPE_CHECKING,
Any,
Callable,
DefaultDict,
Dict,
FrozenSet,
Iterable,
Iterator,
List,
Literal,
NoReturn,
Optional,
Sequence,
Set,
Tuple,
Type,
TypedDict,
TypeVar,
Union,
)
import attr
from hypothesis.errors import Frozen, InvalidArgument, StopTest
from hypothesis.internal.cache import LRUReusedCache
from hypothesis.internal.compat import add_note, floor, int_from_bytes, int_to_bytes
from hypothesis.internal.conjecture.floats import float_to_lex, lex_to_float
from hypothesis.internal.conjecture.junkdrawer import IntList, uniform
from hypothesis.internal.conjecture.utils import (
INT_SIZES,
INT_SIZES_SAMPLER,
Sampler,
calc_label_from_name,
many,
)
from hypothesis.internal.floats import (
SIGNALING_NAN,
SMALLEST_SUBNORMAL,
float_to_int,
int_to_float,
make_float_clamper,
next_down,
next_up,
sign_aware_lte,
)
from hypothesis.internal.intervalsets import IntervalSet
if TYPE_CHECKING:
from typing import TypeAlias
from typing_extensions import dataclass_transform
from hypothesis.strategies import SearchStrategy
from hypothesis.strategies._internal.strategies import Ex
else:
TypeAlias = object
def dataclass_transform():
def wrapper(tp):
return tp
return wrapper
ONE_BOUND_INTEGERS_LABEL = calc_label_from_name("trying a one-bound int allowing 0")
INTEGER_RANGE_DRAW_LABEL = calc_label_from_name("another draw in integer_range()")
BIASED_COIN_LABEL = calc_label_from_name("biased_coin()")
TOP_LABEL = calc_label_from_name("top")
DRAW_BYTES_LABEL = calc_label_from_name("draw_bytes() in ConjectureData")
DRAW_FLOAT_LABEL = calc_label_from_name("drawing a float")
FLOAT_STRATEGY_DO_DRAW_LABEL = calc_label_from_name(
"getting another float in FloatStrategy"
)
INTEGER_WEIGHTED_DISTRIBUTION = calc_label_from_name(
"drawing from a weighted distribution in integers"
)
InterestingOrigin = Tuple[
Type[BaseException], str, int, Tuple[Any, ...], Tuple[Tuple[Any, ...], ...]
]
TargetObservations = Dict[Optional[str], Union[int, float]]
T = TypeVar("T")
class IntegerKWargs(TypedDict):
min_value: Optional[int]
max_value: Optional[int]
weights: Optional[Sequence[float]]
shrink_towards: int
class FloatKWargs(TypedDict):
min_value: float
max_value: float
allow_nan: bool
smallest_nonzero_magnitude: float
class StringKWargs(TypedDict):
intervals: IntervalSet
min_size: int
max_size: Optional[int]
class BytesKWargs(TypedDict):
size: int
class BooleanKWargs(TypedDict):
p: float
IRType: TypeAlias = Union[int, str, bool, float, bytes]
IRKWargsType: TypeAlias = Union[
IntegerKWargs, FloatKWargs, StringKWargs, BytesKWargs, BooleanKWargs
]
IRTypeName: TypeAlias = Literal["integer", "string", "boolean", "float", "bytes"]
class ExtraInformation:
"""A class for holding shared state on a ``ConjectureData`` that should
be added to the final ``ConjectureResult``."""
def __repr__(self) -> str:
return "ExtraInformation({})".format(
", ".join(f"{k}={v!r}" for k, v in self.__dict__.items()),
)
def has_information(self) -> bool:
return bool(self.__dict__)
class Status(IntEnum):
OVERRUN = 0
INVALID = 1
VALID = 2
INTERESTING = 3
def __repr__(self) -> str:
return f"Status.{self.name}"
@dataclass_transform()
@attr.s(frozen=True, slots=True, auto_attribs=True)
class StructuralCoverageTag:
label: int
STRUCTURAL_COVERAGE_CACHE: Dict[int, StructuralCoverageTag] = {}
def structural_coverage(label: int) -> StructuralCoverageTag:
try:
return STRUCTURAL_COVERAGE_CACHE[label]
except KeyError:
return STRUCTURAL_COVERAGE_CACHE.setdefault(label, StructuralCoverageTag(label))
NASTY_FLOATS = sorted(
[
0.0,
0.5,
1.1,
1.5,
1.9,
1.0 / 3,
10e6,
10e-6,
1.175494351e-38,
next_up(0.0),
float_info.min,
float_info.max,
3.402823466e38,
9007199254740992,
1 - 10e-6,
2 + 10e-6,
1.192092896e-07,
2.2204460492503131e-016,
]
+ [2.0**-n for n in (24, 14, 149, 126)] # minimum (sub)normals for float16,32
+ [float_info.min / n for n in (2, 10, 1000, 100_000)] # subnormal in float64
+ [math.inf, math.nan] * 5
+ [SIGNALING_NAN],
key=float_to_lex,
)
NASTY_FLOATS = list(map(float, NASTY_FLOATS))
NASTY_FLOATS.extend([-x for x in NASTY_FLOATS])
FLOAT_INIT_LOGIC_CACHE = LRUReusedCache(4096)
POOLED_KWARGS_CACHE = LRUReusedCache(4096)
DRAW_STRING_DEFAULT_MAX_SIZE = 10**10 # "arbitrarily large"
class Example:
"""Examples track the hierarchical structure of draws from the byte stream,
within a single test run.
Examples are created to mark regions of the byte stream that might be
useful to the shrinker, such as:
- The bytes used by a single draw from a strategy.
- Useful groupings within a strategy, such as individual list elements.
- Strategy-like helper functions that aren't first-class strategies.
- Each lowest-level draw of bits or bytes from the byte stream.
- A single top-level example that spans the entire input.
Example-tracking allows the shrinker to try "high-level" transformations,
such as rearranging or deleting the elements of a list, without having
to understand their exact representation in the byte stream.
Rather than store each ``Example`` as a rich object, it is actually
just an index into the ``Examples`` class defined below. This has two
purposes: Firstly, for most properties of examples we will never need
to allocate storage at all, because most properties are not used on
most examples. Secondly, by storing the properties as compact lists
of integers, we save a considerable amount of space compared to
Python's normal object size.
This does have the downside that it increases the amount of allocation
we do, and slows things down as a result, in some usage patterns because
we repeatedly allocate the same Example or int objects, but it will
often dramatically reduce our memory usage, so is worth it.
"""
__slots__ = ("owner", "index")
def __init__(self, owner: "Examples", index: int) -> None:
self.owner = owner
self.index = index
def __eq__(self, other: object) -> bool:
if self is other:
return True
if not isinstance(other, Example):
return NotImplemented
return (self.owner is other.owner) and (self.index == other.index)
def __ne__(self, other: object) -> bool:
if self is other:
return False
if not isinstance(other, Example):
return NotImplemented
return (self.owner is not other.owner) or (self.index != other.index)
def __repr__(self) -> str:
return f"examples[{self.index}]"
@property
def label(self) -> int:
"""A label is an opaque value that associates each example with its
approximate origin, such as a particular strategy class or a particular
kind of draw."""
return self.owner.labels[self.owner.label_indices[self.index]]
@property
def parent(self):
"""The index of the example that this one is nested directly within."""
if self.index == 0:
return None
return self.owner.parentage[self.index]
@property
def start(self) -> int:
"""The position of the start of this example in the byte stream."""
return self.owner.starts[self.index]
@property
def end(self) -> int:
"""The position directly after the last byte in this byte stream.
i.e. the example corresponds to the half open region [start, end).
"""
return self.owner.ends[self.index]
@property
def ir_start(self) -> int:
return self.owner.ir_starts[self.index]
@property
def ir_end(self) -> int:
return self.owner.ir_ends[self.index]
@property
def depth(self):
"""Depth of this example in the example tree. The top-level example has a
depth of 0."""
return self.owner.depths[self.index]
@property
def trivial(self):
"""An example is "trivial" if it only contains forced bytes and zero bytes.
All examples start out as trivial, and then get marked non-trivial when
we see a byte that is neither forced nor zero."""
return self.index in self.owner.trivial
@property
def discarded(self) -> bool:
"""True if this is example's ``stop_example`` call had ``discard`` set to
``True``. This means we believe that the shrinker should be able to delete
this example completely, without affecting the value produced by its enclosing
strategy. Typically set when a rejection sampler decides to reject a
generated value and try again."""
return self.index in self.owner.discarded
@property
def length(self) -> int:
"""The number of bytes in this example."""
return self.end - self.start
@property
def ir_length(self) -> int:
"""The number of ir nodes in this example."""
return self.ir_end - self.ir_start
@property
def children(self) -> "List[Example]":
"""The list of all examples with this as a parent, in increasing index
order."""
return [self.owner[i] for i in self.owner.children[self.index]]
class ExampleProperty:
"""There are many properties of examples that we calculate by
essentially rerunning the test case multiple times based on the
calls which we record in ExampleRecord.
This class defines a visitor, subclasses of which can be used
to calculate these properties.
"""
def __init__(self, examples: "Examples"):
self.example_stack: "List[int]" = []
self.examples = examples
self.bytes_read = 0
self.example_count = 0
self.block_count = 0
self.ir_node_count = 0
def run(self) -> Any:
"""Rerun the test case with this visitor and return the
results of ``self.finish()``."""
self.begin()
blocks = self.examples.blocks
for record in self.examples.trail:
if record == DRAW_BITS_RECORD:
self.__push(0)
self.bytes_read = blocks.endpoints[self.block_count]
self.block(self.block_count)
self.block_count += 1
self.__pop(discarded=False)
elif record == IR_NODE_RECORD:
data = self.examples.ir_nodes[self.ir_node_count]
self.ir_node(data)
self.ir_node_count += 1
elif record >= START_EXAMPLE_RECORD:
self.__push(record - START_EXAMPLE_RECORD)
else:
assert record in (
STOP_EXAMPLE_DISCARD_RECORD,
STOP_EXAMPLE_NO_DISCARD_RECORD,
)
self.__pop(discarded=record == STOP_EXAMPLE_DISCARD_RECORD)
return self.finish()
def __push(self, label_index: int) -> None:
i = self.example_count
assert i < len(self.examples)
self.start_example(i, label_index=label_index)
self.example_count += 1
self.example_stack.append(i)
def __pop(self, *, discarded: bool) -> None:
i = self.example_stack.pop()
self.stop_example(i, discarded=discarded)
def begin(self) -> None:
"""Called at the beginning of the run to initialise any
relevant state."""
self.result = IntList.of_length(len(self.examples))
def start_example(self, i: int, label_index: int) -> None:
"""Called at the start of each example, with ``i`` the
index of the example and ``label_index`` the index of
its label in ``self.examples.labels``."""
def block(self, i: int) -> None:
"""Called with each ``draw_bits`` call, with ``i`` the index of the
corresponding block in ``self.examples.blocks``"""
def stop_example(self, i: int, *, discarded: bool) -> None:
"""Called at the end of each example, with ``i`` the
index of the example and ``discarded`` being ``True`` if ``stop_example``
was called with ``discard=True``."""
def ir_node(self, node: "IRNode") -> None:
"""Called when an ir node is drawn."""
def finish(self) -> Any:
return self.result
def calculated_example_property(cls: Type[ExampleProperty]) -> Any:
"""Given an ``ExampleProperty`` as above we use this decorator
to transform it into a lazy property on the ``Examples`` class,
which has as its value the result of calling ``cls.run()``,
computed the first time the property is accessed.
This has the slightly weird result that we are defining nested
classes which get turned into properties."""
name = cls.__name__
cache_name = "__" + name
def lazy_calculate(self: "Examples") -> IntList:
result = getattr(self, cache_name, None)
if result is None:
result = cls(self).run()
setattr(self, cache_name, result)
return result
lazy_calculate.__name__ = cls.__name__
lazy_calculate.__qualname__ = cls.__qualname__
return property(lazy_calculate)
DRAW_BITS_RECORD = 0
STOP_EXAMPLE_DISCARD_RECORD = 1
STOP_EXAMPLE_NO_DISCARD_RECORD = 2
START_EXAMPLE_RECORD = 3
IR_NODE_RECORD = calc_label_from_name("ir draw record")
class ExampleRecord:
"""Records the series of ``start_example``, ``stop_example``, and
``draw_bits`` calls so that these may be stored in ``Examples`` and
replayed when we need to know about the structure of individual
``Example`` objects.
Note that there is significant similarity between this class and
``DataObserver``, and the plan is to eventually unify them, but
they currently have slightly different functions and implementations.
"""
def __init__(self) -> None:
self.labels = [DRAW_BYTES_LABEL]
self.__index_of_labels: "Optional[Dict[int, int]]" = {DRAW_BYTES_LABEL: 0}
self.trail = IntList()
self.ir_nodes: List[IRNode] = []
def freeze(self) -> None:
self.__index_of_labels = None
def record_ir_draw(self, ir_type, value, *, kwargs, was_forced):
self.trail.append(IR_NODE_RECORD)
node = IRNode(
ir_type=ir_type,
value=value,
kwargs=kwargs,
was_forced=was_forced,
index=len(self.ir_nodes),
)
self.ir_nodes.append(node)
def start_example(self, label: int) -> None:
assert self.__index_of_labels is not None
try:
i = self.__index_of_labels[label]
except KeyError:
i = self.__index_of_labels.setdefault(label, len(self.labels))
self.labels.append(label)
self.trail.append(START_EXAMPLE_RECORD + i)
def stop_example(self, *, discard: bool) -> None:
if discard:
self.trail.append(STOP_EXAMPLE_DISCARD_RECORD)
else:
self.trail.append(STOP_EXAMPLE_NO_DISCARD_RECORD)
def draw_bits(self) -> None:
self.trail.append(DRAW_BITS_RECORD)
class Examples:
"""A lazy collection of ``Example`` objects, derived from
the record of recorded behaviour in ``ExampleRecord``.
Behaves logically as if it were a list of ``Example`` objects,
but actually mostly exists as a compact store of information
for them to reference into. All properties on here are best
understood as the backing storage for ``Example`` and are
described there.
"""
def __init__(self, record: ExampleRecord, blocks: "Blocks") -> None:
self.trail = record.trail
self.ir_nodes = record.ir_nodes
self.labels = record.labels
self.__length = (
self.trail.count(STOP_EXAMPLE_DISCARD_RECORD)
+ record.trail.count(STOP_EXAMPLE_NO_DISCARD_RECORD)
+ record.trail.count(DRAW_BITS_RECORD)
)
self.blocks = blocks
self.__children: "Optional[List[Sequence[int]]]" = None
class _starts_and_ends(ExampleProperty):
def begin(self):
self.starts = IntList.of_length(len(self.examples))
self.ends = IntList.of_length(len(self.examples))
def start_example(self, i: int, label_index: int) -> None:
self.starts[i] = self.bytes_read
def stop_example(self, i: int, *, discarded: bool) -> None:
self.ends[i] = self.bytes_read
def finish(self) -> Tuple[IntList, IntList]:
return (self.starts, self.ends)
starts_and_ends: "Tuple[IntList, IntList]" = calculated_example_property(
_starts_and_ends
)
@property
def starts(self) -> IntList:
return self.starts_and_ends[0]
@property
def ends(self) -> IntList:
return self.starts_and_ends[1]
class _ir_starts_and_ends(ExampleProperty):
def begin(self):
self.starts = IntList.of_length(len(self.examples))
self.ends = IntList.of_length(len(self.examples))
def start_example(self, i: int, label_index: int) -> None:
self.starts[i] = self.ir_node_count
def stop_example(self, i: int, *, discarded: bool) -> None:
self.ends[i] = self.ir_node_count
def finish(self) -> Tuple[IntList, IntList]:
return (self.starts, self.ends)
ir_starts_and_ends: "Tuple[IntList, IntList]" = calculated_example_property(
_ir_starts_and_ends
)
@property
def ir_starts(self) -> IntList:
return self.ir_starts_and_ends[0]
@property
def ir_ends(self) -> IntList:
return self.ir_starts_and_ends[1]
class _discarded(ExampleProperty):
def begin(self) -> None:
self.result: "Set[int]" = set() # type: ignore # IntList in parent class
def finish(self) -> FrozenSet[int]:
return frozenset(self.result)
def stop_example(self, i: int, *, discarded: bool) -> None:
if discarded:
self.result.add(i)
discarded: FrozenSet[int] = calculated_example_property(_discarded)
class _trivial(ExampleProperty):
def begin(self) -> None:
self.nontrivial = IntList.of_length(len(self.examples))
self.result: "Set[int]" = set() # type: ignore # IntList in parent class
def block(self, i: int) -> None:
if not self.examples.blocks.trivial(i):
self.nontrivial[self.example_stack[-1]] = 1
def stop_example(self, i: int, *, discarded: bool) -> None:
if self.nontrivial[i]:
if self.example_stack:
self.nontrivial[self.example_stack[-1]] = 1
else:
self.result.add(i)
def finish(self) -> FrozenSet[int]:
return frozenset(self.result)
trivial: FrozenSet[int] = calculated_example_property(_trivial)
class _parentage(ExampleProperty):
def stop_example(self, i: int, *, discarded: bool) -> None:
if i > 0:
self.result[i] = self.example_stack[-1]
parentage: IntList = calculated_example_property(_parentage)
class _depths(ExampleProperty):
def begin(self):
self.result = IntList.of_length(len(self.examples))
def start_example(self, i: int, label_index: int) -> None:
self.result[i] = len(self.example_stack)
depths: IntList = calculated_example_property(_depths)
class _ir_tree_nodes(ExampleProperty):
def begin(self):
self.result = []
def ir_node(self, ir_node):
self.result.append(ir_node)
ir_tree_nodes: "List[IRNode]" = calculated_example_property(_ir_tree_nodes)
class _label_indices(ExampleProperty):
def start_example(self, i: int, label_index: int) -> None:
self.result[i] = label_index
label_indices: IntList = calculated_example_property(_label_indices)
class _mutator_groups(ExampleProperty):
def begin(self) -> None:
self.groups: "Dict[Tuple[int, int], List[int]]" = defaultdict(list)
def start_example(self, i: int, label_index: int) -> None:
depth = len(self.example_stack)
self.groups[label_index, depth].append(i)
def finish(self) -> Iterable[Iterable[int]]:
# Discard groups with only one example, since the mutator can't
# do anything useful with them.
return [g for g in self.groups.values() if len(g) >= 2]
mutator_groups: List[List[int]] = calculated_example_property(_mutator_groups)
@property
def children(self) -> List[Sequence[int]]:
if self.__children is None:
children = [IntList() for _ in range(len(self))]
for i, p in enumerate(self.parentage):
if i > 0:
children[p].append(i)
# Replace empty children lists with a tuple to reduce
# memory usage.
for i, c in enumerate(children):
if not c:
children[i] = () # type: ignore
self.__children = children # type: ignore
return self.__children # type: ignore
def __len__(self) -> int:
return self.__length
def __getitem__(self, i: int) -> Example:
assert isinstance(i, int)
n = len(self)
if i < -n or i >= n:
raise IndexError(f"Index {i} out of range [-{n}, {n})")
if i < 0:
i += n
return Example(self, i)
@dataclass_transform()
@attr.s(slots=True, frozen=True)
class Block:
"""Blocks track the flat list of lowest-level draws from the byte stream,
within a single test run.
Block-tracking allows the shrinker to try "low-level"
transformations, such as minimizing the numeric value of an
individual call to ``draw_bits``.
"""
start: int = attr.ib()
end: int = attr.ib()
# Index of this block inside the overall list of blocks.
index: int = attr.ib()
# True if this block's byte values were forced by a write operation.
# As long as the bytes before this block remain the same, modifying this
# block's bytes will have no effect.
forced: bool = attr.ib(repr=False)
# True if this block's byte values are all 0. Reading this flag can be
# more convenient than explicitly checking a slice for non-zero bytes.
all_zero: bool = attr.ib(repr=False)
@property
def bounds(self) -> Tuple[int, int]:
return (self.start, self.end)
@property
def length(self) -> int:
return self.end - self.start
@property
def trivial(self) -> bool:
return self.forced or self.all_zero
class Blocks:
"""A lazily calculated list of blocks for a particular ``ConjectureResult``
or ``ConjectureData`` object.
Pretends to be a list containing ``Block`` objects but actually only
contains their endpoints right up until the point where you want to
access the actual block, at which point it is constructed.
This is designed to be as space efficient as possible, so will at
various points silently transform its representation into one
that is better suited for the current access pattern.
In addition, it has a number of convenience methods for accessing
properties of the block object at index ``i`` that should generally
be preferred to using the Block objects directly, as it will not
have to allocate the actual object."""
__slots__ = ("endpoints", "owner", "__blocks", "__count", "__sparse")
owner: "Union[ConjectureData, ConjectureResult, None]"
__blocks: Union[Dict[int, Block], List[Optional[Block]]]
def __init__(self, owner: "ConjectureData") -> None:
self.owner = owner
self.endpoints = IntList()
self.__blocks = {}
self.__count = 0
self.__sparse = True
def add_endpoint(self, n: int) -> None:
"""Add n to the list of endpoints."""
assert isinstance(self.owner, ConjectureData)
self.endpoints.append(n)
def transfer_ownership(self, new_owner: "ConjectureResult") -> None:
"""Used to move ``Blocks`` over to a ``ConjectureResult`` object
when that is read to be used and we no longer want to keep the
whole ``ConjectureData`` around."""
assert isinstance(new_owner, ConjectureResult)
self.owner = new_owner
self.__check_completion()
def start(self, i: int) -> int:
"""Equivalent to self[i].start."""
i = self._check_index(i)
if i == 0:
return 0
else:
return self.end(i - 1)
def end(self, i: int) -> int:
"""Equivalent to self[i].end."""
return self.endpoints[i]
def bounds(self, i: int) -> Tuple[int, int]:
"""Equivalent to self[i].bounds."""
return (self.start(i), self.end(i))
def all_bounds(self) -> Iterable[Tuple[int, int]]:
"""Equivalent to [(b.start, b.end) for b in self]."""
prev = 0
for e in self.endpoints:
yield (prev, e)
prev = e
@property
def last_block_length(self):
return self.end(-1) - self.start(-1)
def __len__(self) -> int:
return len(self.endpoints)
def __known_block(self, i: int) -> Optional[Block]:
try:
return self.__blocks[i]
except (KeyError, IndexError):
return None
def trivial(self, i: int) -> Any:
"""Equivalent to self.blocks[i].trivial."""
if self.owner is not None:
return self.start(i) in self.owner.forced_indices or not any(
self.owner.buffer[self.start(i) : self.end(i)]
)
else:
return self[i].trivial
def _check_index(self, i: int) -> int:
n = len(self)
if i < -n or i >= n:
raise IndexError(f"Index {i} out of range [-{n}, {n})")
if i < 0:
i += n
return i
def __getitem__(self, i: int) -> Block:
i = self._check_index(i)
assert i >= 0
result = self.__known_block(i)
if result is not None:
return result
# We store the blocks as a sparse dict mapping indices to the
# actual result, but this isn't the best representation once we
# stop being sparse and want to use most of the blocks. Switch
# over to a list at that point.
if self.__sparse and len(self.__blocks) * 2 >= len(self):
new_blocks: "List[Optional[Block]]" = [None] * len(self)
assert isinstance(self.__blocks, dict)
for k, v in self.__blocks.items():
new_blocks[k] = v
self.__sparse = False
self.__blocks = new_blocks
assert self.__blocks[i] is None
start = self.start(i)
end = self.end(i)
# We keep track of the number of blocks that have actually been
# instantiated so that when every block that could be instantiated
# has been we know that the list is complete and can throw away
# some data that we no longer need.
self.__count += 1
# Integrity check: We can't have allocated more blocks than we have
# positions for blocks.
assert self.__count <= len(self)
assert self.owner is not None
result = Block(
start=start,
end=end,
index=i,
forced=start in self.owner.forced_indices,
all_zero=not any(self.owner.buffer[start:end]),
)
try:
self.__blocks[i] = result
except IndexError:
assert isinstance(self.__blocks, list)
assert len(self.__blocks) < len(self)
self.__blocks.extend([None] * (len(self) - len(self.__blocks)))
self.__blocks[i] = result
self.__check_completion()
return result
def __check_completion(self):
"""The list of blocks is complete if we have created every ``Block``
object that we currently good and know that no more will be created.
If this happens then we don't need to keep the reference to the
owner around, and delete it so that there is no circular reference.
The main benefit of this is that the gc doesn't need to run to collect
this because normal reference counting is enough.
"""
if self.__count == len(self) and isinstance(self.owner, ConjectureResult):
self.owner = None
def __iter__(self) -> Iterator[Block]:
for i in range(len(self)):
yield self[i]
def __repr__(self) -> str:
parts: "List[str]" = []
for i in range(len(self)):
b = self.__known_block(i)
if b is None:
parts.append("...")
else:
parts.append(repr(b))
return "Block([{}])".format(", ".join(parts))
class _Overrun:
status = Status.OVERRUN
def __repr__(self):
return "Overrun"
Overrun = _Overrun()
global_test_counter = 0
MAX_DEPTH = 100
class DataObserver:
"""Observer class for recording the behaviour of a
ConjectureData object, primarily used for tracking
the behaviour in the tree cache."""
def conclude_test(
self,
status: Status,
interesting_origin: Optional[InterestingOrigin],
) -> None:
"""Called when ``conclude_test`` is called on the
observed ``ConjectureData``, with the same arguments.
Note that this is called after ``freeze`` has completed.
"""
def kill_branch(self) -> None:
"""Mark this part of the tree as not worth re-exploring."""
def draw_integer(
self, value: int, *, kwargs: IntegerKWargs, was_forced: bool
) -> None:
pass
def draw_float(
self, value: float, *, kwargs: FloatKWargs, was_forced: bool
) -> None:
pass
def draw_string(
self, value: str, *, kwargs: StringKWargs, was_forced: bool
) -> None:
pass
def draw_bytes(
self, value: bytes, *, kwargs: BytesKWargs, was_forced: bool
) -> None:
pass
def draw_boolean(
self, value: bool, *, kwargs: BooleanKWargs, was_forced: bool
) -> None:
pass
@attr.s(slots=True, repr=False, eq=False)
class IRNode:
ir_type: IRTypeName = attr.ib()
value: IRType = attr.ib()
kwargs: IRKWargsType = attr.ib()
was_forced: bool = attr.ib()
index: Optional[int] = attr.ib(default=None)
def copy(self, *, with_value: IRType) -> "IRNode":
# we may want to allow this combination in the future, but for now it's
# a footgun.
assert not self.was_forced, "modifying a forced node doesn't make sense"
# explicitly not copying index. node indices are only assigned via
# ExampleRecord. This prevents footguns with relying on stale indices
# after copying.
return IRNode(
ir_type=self.ir_type,
value=with_value,
kwargs=self.kwargs,
was_forced=self.was_forced,
)
@property
def trivial(self):
"""
A node is trivial if it cannot be simplified any further. This does not
mean that modifying a trivial node can't produce simpler test cases when
viewing the tree as a whole. Just that when viewing this node in
isolation, this is the simplest the node can get.
"""
if self.was_forced:
return True
if self.ir_type == "integer":
shrink_towards = self.kwargs["shrink_towards"]
min_value = self.kwargs["min_value"]
max_value = self.kwargs["max_value"]
if min_value is not None:
shrink_towards = max(min_value, shrink_towards)
if max_value is not None: