def evaluate_test_data(data):
if perform_health_check and not performed_random_check[0]:
initial_state = getglobalrandomstate()
performed_random_check[0] = True
else:
initial_state = None
try:
result = test_runner(
data, reify_and_execute(
search_strategy,
test,
))
if result is not None and settings.perform_health_check:
fail_health_check(
('Tests run under @given should return None, but '
'%s returned %r instead.') %
(test.__name__, result), HealthCheck.return_value)
return False
except UnsatisfiedAssumption:
data.mark_invalid()
except (
HypothesisDeprecationWarning,
FailedHealthCheck,
StopTest,
):
raise
except Exception:
last_exception[0] = traceback.format_exc()
verbose_report(last_exception[0])
data.mark_interesting()
finally:
if (initial_state is not None
and getglobalrandomstate() != initial_state):
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.',
HealthCheck.random_module,
)
def evaluate_test_data(data):
if perform_health_check and not performed_random_check[0]:
initial_state = getglobalrandomstate()
performed_random_check[0] = True
else:
initial_state = None
try:
result = test_runner(data, reify_and_execute(
search_strategy, test,
))
if result is not None and settings.perform_health_check:
fail_health_check((
'Tests run under @given should return None, but '
'%s returned %r instead.'
) % (test.__name__, result), HealthCheck.return_value)
return False
except UnsatisfiedAssumption:
data.mark_invalid()
except (
HypothesisDeprecationWarning, FailedHealthCheck,
StopTest,
):
raise
except Exception:
last_exception[0] = traceback.format_exc()
verbose_report(last_exception[0])
data.mark_interesting()
finally:
if (
initial_state is not None and
getglobalrandomstate() != initial_state
):
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.',
HealthCheck.random_module,
)
def is_template_example(xs):
if perform_health_check and not warned_random[0]:
initial_state = getglobalrandomstate()
record_repr = [None]
try:
result = test_runner(reify_and_execute(
search_strategy, xs, test,
record_repr=record_repr,
))
if result is not None:
note_deprecation((
'Tests run under @given should return None, but '
'%s returned %r instead.'
'In Hypothesis 2.0 this will become an error.'
) % (test.__name__, result), settings)
return False
except HypothesisDeprecationWarning:
raise
except UnsatisfiedAssumption as e:
raise e
except Exception as e:
last_exception[0] = traceback.format_exc()
repr_for_last_exception[0] = record_repr[0]
verbose_report(last_exception[0])
return True
finally:
if (
not warned_random[0] and
perform_health_check and
getglobalrandomstate() != initial_state
):
warned_random[0] = True
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.'
)
def is_template_example(xs):
if perform_health_check and not warned_random[0]:
initial_state = getglobalrandomstate()
record_repr = [None]
try:
result = test_runner(
reify_and_execute(
search_strategy,
xs,
test,
record_repr=record_repr,
))
if result is not None:
note_deprecation(
('Tests run under @given should return None, but '
'%s returned %r instead.'
'In Hypothesis 2.0 this will become an error.') %
(test.__name__, result), settings)
return False
except HypothesisDeprecationWarning:
raise
except UnsatisfiedAssumption as e:
raise e
except Exception as e:
last_exception[0] = traceback.format_exc()
repr_for_last_exception[0] = record_repr[0]
verbose_report(last_exception[0])
return True
finally:
if (not warned_random[0] and perform_health_check
and getglobalrandomstate() != initial_state):
warned_random[0] = True
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.')
def wrapped_test(*arguments, **kwargs):
settings = wrapped_test._hypothesis_internal_use_settings
if wrapped_test._hypothesis_internal_use_seed is not None:
random = Random(wrapped_test._hypothesis_internal_use_seed)
elif settings.derandomize:
random = Random(function_digest(test))
else:
random = new_random()
import hypothesis.strategies as sd
selfy = None
arguments, kwargs = convert_positional_arguments(
wrapped_test, arguments, kwargs)
# If the test function is a method of some kind, the bound object
# will be the first named argument if there are any, otherwise the
# first vararg (if any).
if argspec.args:
selfy = kwargs.get(argspec.args[0])
elif arguments:
selfy = arguments[0]
test_runner = executor(selfy)
for example in reversed(
getattr(wrapped_test, 'hypothesis_explicit_examples', ())):
if example.args:
example_kwargs = dict(
zip(original_argspec.args[-len(example.args):],
example.args))
else:
example_kwargs = example.kwargs
example_kwargs.update(kwargs)
# Note: Test may mutate arguments and we can't rerun explicit
# examples, so we have to calculate the failure message at this
# point rather than than later.
message_on_failure = 'Falsifying example: %s(%s)' % (
test.__name__, arg_string(test, arguments, example_kwargs))
try:
with BuildContext() as b:
test_runner(lambda: test(*arguments, **example_kwargs))
except BaseException:
report(message_on_failure)
for n in b.notes:
report(n)
raise
arguments = tuple(arguments)
given_specifier = sd.tuples(
sd.just(arguments),
sd.fixed_dictionaries(generator_kwargs).map(
lambda args: dict(args, **kwargs)))
def fail_health_check(message):
message += (
'\nSee http://hypothesis.readthedocs.org/en/latest/health'
'checks.html for more information about this.')
if settings.strict:
raise FailedHealthCheck(message)
else:
warnings.warn(FailedHealthCheck(message))
search_strategy = given_specifier
search_strategy.validate()
if settings.database:
storage = settings.database.storage(fully_qualified_name(test))
else:
storage = None
start = time.time()
warned_random = [False]
perform_health_check = settings.perform_health_check
if Settings.default is not None:
perform_health_check &= Settings.default.perform_health_check
if perform_health_check:
initial_state = getglobalrandomstate()
health_check_random = Random(random.getrandbits(128))
count = 0
bad_draws = 0
filtered_draws = 0
errors = 0
while (count < 10 and time.time() < start + 1
and filtered_draws < 50 and bad_draws < 50):
try:
with Settings(settings, verbosity=Verbosity.quiet):
test_runner(
reify_and_execute(
search_strategy,
search_strategy.draw_template(
health_check_random,
search_strategy.draw_parameter(
health_check_random, )),
lambda *args, **kwargs: None,
))
count += 1
except BadTemplateDraw:
bad_draws += 1
except UnsatisfiedAssumption:
filtered_draws += 1
except Exception:
if errors == 0:
report(traceback.format_exc())
errors += 1
if test_runner is default_executor:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
"an error in a function yo've passed to "
'it to construct your data.')
else:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
'an error in a function you\'ve passed to '
'it to construct your data. Additionally, '
'you have a custom executor, which means '
'that this could be your executor failing '
'to handle a function which returns None. ')
if filtered_draws >= 50:
fail_health_check((
'It looks like your strategy is filtering out a lot '
'of data. Health check found %d filtered examples but '
'only %d good ones. This will make your tests much '
'slower, and also will probably distort the data '
'generation quite a lot. You should adapt your '
'strategy to filter less.') % (filtered_draws, count))
if bad_draws >= 50:
fail_health_check(
'Hypothesis is struggling to generate examples. '
'This is often a sign of a recursive strategy which '
'fans out too broadly. If you\'re using recursive, '
'try to reduce the size of the recursive step or '
'increase the maximum permitted number of leaves.')
runtime = time.time() - start
if runtime > 1.0 or count < 10:
fail_health_check(
('Data generation is extremely slow: Only produced '
'%d valid examples in %.2f seconds. Try decreasing '
"size of the data yo're generating (with e.g."
'average_size or max_leaves parameters).') %
(count, runtime))
if getglobalrandomstate() != initial_state:
warned_random[0] = True
fail_health_check(
'Data generation depends on global random module. '
'This makes results impossible to replay, which '
'prevents Hypothesis from working correctly. '
'If you want to use methods from random, use '
'randoms() from hypothesis.strategies to get an '
'instance of Random you can use. Alternatively, you '
'can use the random_module() strategy to explicitly '
'seed the random module.')
last_exception = [None]
repr_for_last_exception = [None]
def is_template_example(xs):
if perform_health_check and not warned_random[0]:
initial_state = getglobalrandomstate()
record_repr = [None]
try:
result = test_runner(
reify_and_execute(
search_strategy,
xs,
test,
record_repr=record_repr,
))
if result is not None and settings.perform_health_check:
raise FailedHealthCheck(
('Tests run under @given should return None, but '
'%s returned %r instead.') %
(test.__name__, result), settings)
return False
except (HypothesisDeprecationWarning, FailedHealthCheck,
UnsatisfiedAssumption):
raise
except Exception:
last_exception[0] = traceback.format_exc()
repr_for_last_exception[0] = record_repr[0]
verbose_report(last_exception[0])
return True
finally:
if (not warned_random[0] and perform_health_check
and getglobalrandomstate() != initial_state):
warned_random[0] = True
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.')
is_template_example.__name__ = test.__name__
is_template_example.__qualname__ = qualname(test)
with settings:
falsifying_template = None
try:
falsifying_template = best_satisfying_template(
search_strategy,
random,
is_template_example,
settings,
storage,
start_time=start,
)
except NoSuchExample:
return
assert last_exception[0] is not None
try:
test_runner(
reify_and_execute(search_strategy,
falsifying_template,
test,
print_example=True,
is_final=True))
except UnsatisfiedAssumption:
report(traceback.format_exc())
raise Flaky(
'Unreliable assumption: An example which satisfied '
'assumptions on the first run now fails it.')
report(
'Failed to reproduce exception. Expected: \n' +
last_exception[0], )
try:
test_runner(
reify_and_execute(search_strategy,
falsifying_template,
test_is_flaky(
test,
repr_for_last_exception[0]),
print_example=True,
is_final=True))
except UnsatisfiedAssumption:
raise Flaky(
'Unreliable test data: Failed to reproduce a failure '
'and then when it came to recreating the example in '
'order to print the test data with a flaky result '
'the example was filtered out (by e.g. a '
'call to filter in your strategy) when we didn\'t '
'expect it to be.')
def wrapped_test(*arguments, **kwargs):
if settings.derandomize:
random = Random(function_digest(test))
else:
random = provided_random or new_random()
import hypothesis.strategies as sd
from hypothesis.internal.strategymethod import strategy
selfy = None
arguments, kwargs = convert_positional_arguments(
wrapped_test, arguments, kwargs)
for arg in hypothesis_owned_arguments:
try:
value = kwargs[arg]
except KeyError:
continue
if not isinstance(value, HypothesisProvided):
note_deprecation(
'Passing in explicit values to override Hypothesis '
'provided values is deprecated and will no longer '
'work in Hypothesis 2.0. If you need to do this, '
'extract a common function and call that from a '
'Hypothesis based test.', settings
)
# Anything in unused_kwargs hasn't been injected through
# argspec.defaults, so we need to add them.
for k in unused_kwargs:
if k not in kwargs:
kwargs[k] = unused_kwargs[k]
# If the test function is a method of some kind, the bound object
# will be the first named argument if there are any, otherwise the
# first vararg (if any).
if argspec.args:
selfy = kwargs.get(argspec.args[0])
elif arguments:
selfy = arguments[0]
if isinstance(selfy, HypothesisProvided):
selfy = None
test_runner = executor(selfy)
for example in reversed(getattr(
wrapped_test, u'hypothesis_explicit_examples', ()
)):
if example.args:
example_kwargs = dict(zip(
argspec.args[-len(example.args):], example.args
))
else:
example_kwargs = dict(example.kwargs)
for k, v in kwargs.items():
if not isinstance(v, HypothesisProvided):
example_kwargs[k] = v
# Note: Test may mutate arguments and we can't rerun explicit
# examples, so we have to calculate the failure message at this
# point rather than than later.
message_on_failure = u'Falsifying example: %s(%s)' % (
test.__name__, arg_string(test, arguments, example_kwargs)
)
try:
with BuildContext() as b:
test_runner(
lambda: test(*arguments, **example_kwargs)
)
except BaseException:
report(message_on_failure)
for n in b.notes:
report(n)
raise
if not any(
isinstance(x, HypothesisProvided)
for xs in (arguments, kwargs.values())
for x in xs
):
# All arguments have been satisfied without needing to invoke
# hypothesis
test_runner(lambda: test(*arguments, **kwargs))
return
def convert_to_specifier(v):
if isinstance(v, HypothesisProvided):
return strategy(v.value, settings)
else:
return sd.just(v)
given_specifier = sd.tuples(
sd.tuples(*map(convert_to_specifier, arguments)),
sd.fixed_dictionaries(dict(
(k, convert_to_specifier(v)) for (k, v) in kwargs.items()))
)
def fail_health_check(message):
message += (
'\nSee http://hypothesis.readthedocs.org/en/latest/health'
'checks.html for more information about this.'
)
if settings.strict:
raise FailedHealthCheck(message)
else:
warnings.warn(FailedHealthCheck(message))
search_strategy = strategy(given_specifier, settings)
search_strategy.validate()
if settings.database:
storage = settings.database.storage(
fully_qualified_name(test))
else:
storage = None
start = time.time()
warned_random = [False]
perform_health_check = settings.perform_health_check
if Settings.default is not None:
perform_health_check &= Settings.default.perform_health_check
if perform_health_check:
initial_state = getglobalrandomstate()
health_check_random = Random(random.getrandbits(128))
count = 0
bad_draws = 0
filtered_draws = 0
errors = 0
while (
count < 10 and time.time() < start + 1 and
filtered_draws < 50 and bad_draws < 50
):
try:
with Settings(settings, verbosity=Verbosity.quiet):
test_runner(reify_and_execute(
search_strategy,
search_strategy.draw_template(
health_check_random,
search_strategy.draw_parameter(
health_check_random,
)),
lambda *args, **kwargs: None,
))
count += 1
except BadTemplateDraw:
bad_draws += 1
except UnsatisfiedAssumption:
filtered_draws += 1
except Exception:
if errors == 0:
report(traceback.format_exc())
errors += 1
if test_runner is default_executor:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
"an error in a function you've passed to "
'it to construct your data.'
)
else:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
'an error in a function you\'ve passed to '
'it to construct your data. Additionally, '
'you have a custom executor, which means '
'that this could be your executor failing '
'to handle a function which returns None. '
)
if filtered_draws >= 50:
fail_health_check((
'It looks like your strategy is filtering out a lot '
'of data. Health check found %d filtered examples but '
'only %d good ones. This will make your tests much '
'slower, and also will probably distort the data '
'generation quite a lot. You should adapt your '
'strategy to filter less.') % (
filtered_draws, count
))
if bad_draws >= 50:
fail_health_check(
'Hypothesis is struggling to generate examples. '
'This is often a sign of a recursive strategy which '
'fans out too broadly. If you\'re using recursive, '
'try to reduce the size of the recursive step or '
'increase the maximum permitted number of leaves.'
)
runtime = time.time() - start
if runtime > 1.0 or count < 10:
fail_health_check((
'Data generation is extremely slow: Only produced '
'%d valid examples in %.2f seconds. Try decreasing '
"size of the data you're generating (with e.g."
'average_size or max_leaves parameters).'
) % (count, runtime))
if getglobalrandomstate() != initial_state:
warned_random[0] = True
fail_health_check(
'Data generation depends on global random module. '
'This makes results impossible to replay, which '
'prevents Hypothesis from working correctly. '
'If you want to use methods from random, use '
'randoms() from hypothesis.strategies to get an '
'instance of Random you can use. Alternatively, you '
'can use the random_module() strategy to explicitly '
'seed the random module.'
)
last_exception = [None]
repr_for_last_exception = [None]
def is_template_example(xs):
if perform_health_check and not warned_random[0]:
initial_state = getglobalrandomstate()
record_repr = [None]
try:
result = test_runner(reify_and_execute(
search_strategy, xs, test,
record_repr=record_repr,
))
if result is not None:
note_deprecation((
'Tests run under @given should return None, but '
'%s returned %r instead.'
'In Hypothesis 2.0 this will become an error.'
) % (test.__name__, result), settings)
return False
except HypothesisDeprecationWarning:
raise
except UnsatisfiedAssumption as e:
raise e
except Exception as e:
last_exception[0] = traceback.format_exc()
repr_for_last_exception[0] = record_repr[0]
verbose_report(last_exception[0])
return True
finally:
if (
not warned_random[0] and
perform_health_check and
getglobalrandomstate() != initial_state
):
warned_random[0] = True
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.'
)
is_template_example.__name__ = test.__name__
is_template_example.__qualname__ = qualname(test)
falsifying_template = None
try:
falsifying_template = best_satisfying_template(
search_strategy, random, is_template_example,
settings, storage, start_time=start,
)
except NoSuchExample:
return
assert last_exception[0] is not None
with settings:
test_runner(reify_and_execute(
search_strategy, falsifying_template, test,
print_example=True, is_final=True
))
report(
u'Failed to reproduce exception. Expected: \n' +
last_exception[0],
)
test_runner(reify_and_execute(
search_strategy, falsifying_template,
test_is_flaky(test, repr_for_last_exception[0]),
print_example=True, is_final=True
))
def wrapped_test(*arguments, **kwargs):
settings = wrapped_test._hypothesis_internal_use_settings
if wrapped_test._hypothesis_internal_use_seed is not None:
random = Random(
wrapped_test._hypothesis_internal_use_seed)
elif settings.derandomize:
random = Random(function_digest(test))
else:
random = new_random()
import hypothesis.strategies as sd
selfy = None
arguments, kwargs = convert_positional_arguments(
wrapped_test, arguments, kwargs)
# If the test function is a method of some kind, the bound object
# will be the first named argument if there are any, otherwise the
# first vararg (if any).
if argspec.args:
selfy = kwargs.get(argspec.args[0])
elif arguments:
selfy = arguments[0]
test_runner = new_style_executor(selfy)
for example in reversed(getattr(
wrapped_test, 'hypothesis_explicit_examples', ()
)):
if example.args:
if len(example.args) > len(original_argspec.args):
raise InvalidArgument(
'example has too many arguments for test. '
'Expected at most %d but got %d' % (
len(original_argspec.args), len(example.args)))
example_kwargs = dict(zip(
original_argspec.args[-len(example.args):],
example.args
))
else:
example_kwargs = example.kwargs
example_kwargs.update(kwargs)
# Note: Test may mutate arguments and we can't rerun explicit
# examples, so we have to calculate the failure message at this
# point rather than than later.
message_on_failure = 'Falsifying example: %s(%s)' % (
test.__name__, arg_string(test, arguments, example_kwargs)
)
try:
with BuildContext() as b:
test_runner(
None,
lambda data: test(*arguments, **example_kwargs)
)
except BaseException:
report(message_on_failure)
for n in b.notes:
report(n)
raise
if settings.max_examples <= 0:
return
arguments = tuple(arguments)
given_specifier = sd.tuples(
sd.just(arguments),
sd.fixed_dictionaries(generator_kwargs).map(
lambda args: dict(args, **kwargs)
)
)
def fail_health_check(message):
message += (
'\nSee http://hypothesis.readthedocs.org/en/latest/health'
'checks.html for more information about this.'
)
raise FailedHealthCheck(message)
search_strategy = given_specifier
search_strategy.validate()
perform_health_check = settings.perform_health_check
perform_health_check &= Settings.default.perform_health_check
from hypothesis.internal.conjecture.data import TestData, Status, \
StopTest
if perform_health_check:
initial_state = getglobalrandomstate()
health_check_random = Random(random.getrandbits(128))
# We "pre warm" the health check with one draw to give it some
# time to calculate any cached data. This prevents the case
# where the first draw of the health check takes ages because
# of loading unicode data the first time.
data = TestData(
max_length=settings.buffer_size,
draw_bytes=lambda data, n, distribution:
distribution(health_check_random, n)
)
with Settings(settings, verbosity=Verbosity.quiet):
try:
test_runner(data, reify_and_execute(
search_strategy,
lambda *args, **kwargs: None,
))
except BaseException:
pass
count = 0
overruns = 0
filtered_draws = 0
start = time.time()
while (
count < 10 and time.time() < start + 1 and
filtered_draws < 50 and overruns < 20
):
try:
data = TestData(
max_length=settings.buffer_size,
draw_bytes=lambda data, n, distribution:
distribution(health_check_random, n)
)
with Settings(settings, verbosity=Verbosity.quiet):
test_runner(data, reify_and_execute(
search_strategy,
lambda *args, **kwargs: None,
))
count += 1
except UnsatisfiedAssumption:
filtered_draws += 1
except StopTest:
if data.status == Status.INVALID:
filtered_draws += 1
else:
assert data.status == Status.OVERRUN
overruns += 1
except Exception:
report(traceback.format_exc())
if test_runner is default_new_style_executor:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
"an error in a function yo've passed to "
'it to construct your data.'
)
else:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
'an error in a function you\'ve passed to '
'it to construct your data. Additionally, '
'you have a custom executor, which means '
'that this could be your executor failing '
'to handle a function which returns None. '
)
if overruns >= 20 or (
not count and overruns > 0
):
fail_health_check((
'Examples routinely exceeded the max allowable size. '
'(%d examples overran while generating %d valid ones)'
'. Generating examples this large will usually lead to'
' bad results. You should try setting average_size or '
'max_size parameters on your collections and turning '
'max_leaves down on recursive() calls.') % (
overruns, count
))
if filtered_draws >= 50 or (
not count and filtered_draws > 0
):
fail_health_check((
'It looks like your strategy is filtering out a lot '
'of data. Health check found %d filtered examples but '
'only %d good ones. This will make your tests much '
'slower, and also will probably distort the data '
'generation quite a lot. You should adapt your '
'strategy to filter less. This can also be caused by '
'a low max_leaves parameter in recursive() calls') % (
filtered_draws, count
))
runtime = time.time() - start
if runtime > 1.0 or count < 10:
fail_health_check((
'Data generation is extremely slow: Only produced '
'%d valid examples in %.2f seconds (%d invalid ones '
'and %d exceeded maximum size). Try decreasing '
"size of the data you're generating (with e.g."
'average_size or max_leaves parameters).'
) % (count, runtime, filtered_draws, overruns))
if getglobalrandomstate() != initial_state:
fail_health_check(
'Data generation depends on global random module. '
'This makes results impossible to replay, which '
'prevents Hypothesis from working correctly. '
'If you want to use methods from random, use '
'randoms() from hypothesis.strategies to get an '
'instance of Random you can use. Alternatively, you '
'can use the random_module() strategy to explicitly '
'seed the random module.'
)
last_exception = [None]
repr_for_last_exception = [None]
performed_random_check = [False]
def evaluate_test_data(data):
if perform_health_check and not performed_random_check[0]:
initial_state = getglobalrandomstate()
performed_random_check[0] = True
else:
initial_state = None
try:
result = test_runner(data, reify_and_execute(
search_strategy, test,
))
if result is not None and settings.perform_health_check:
raise FailedHealthCheck((
'Tests run under @given should return None, but '
'%s returned %r instead.'
) % (test.__name__, result), settings)
return False
except UnsatisfiedAssumption:
data.mark_invalid()
except (
HypothesisDeprecationWarning, FailedHealthCheck,
StopTest,
):
raise
except Exception:
last_exception[0] = traceback.format_exc()
verbose_report(last_exception[0])
data.mark_interesting()
finally:
if (
initial_state is not None and
getglobalrandomstate() != initial_state
):
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.')
from hypothesis.internal.conjecture.engine import TestRunner
falsifying_example = None
database_key = str_to_bytes(fully_qualified_name(test))
start_time = time.time()
runner = TestRunner(
evaluate_test_data,
settings=settings, random=random,
database_key=database_key,
)
runner.run()
run_time = time.time() - start_time
timed_out = (
settings.timeout > 0 and
run_time >= settings.timeout
)
if runner.last_data.status == Status.INTERESTING:
falsifying_example = runner.last_data.buffer
if settings.database is not None:
settings.database.save(
database_key, falsifying_example
)
else:
if runner.valid_examples < min(
settings.min_satisfying_examples,
settings.max_examples,
):
if timed_out:
raise Timeout((
'Ran out of time before finding a satisfying '
'example for '
'%s. Only found %d examples in ' +
'%.2fs.'
) % (
get_pretty_function_description(test),
runner.valid_examples, run_time
))
else:
raise Unsatisfiable((
'Unable to satisfy assumptions of hypothesis '
'%s. Only %d examples considered '
'satisfied assumptions'
) % (
get_pretty_function_description(test),
runner.valid_examples,))
return
assert last_exception[0] is not None
try:
with settings:
test_runner(
TestData.for_buffer(falsifying_example),
reify_and_execute(
search_strategy, test,
print_example=True, is_final=True
))
except (UnsatisfiedAssumption, StopTest):
report(traceback.format_exc())
raise Flaky(
'Unreliable assumption: An example which satisfied '
'assumptions on the first run now fails it.'
)
report(
'Failed to reproduce exception. Expected: \n' +
last_exception[0],
)
filter_message = (
'Unreliable test data: Failed to reproduce a failure '
'and then when it came to recreating the example in '
'order to print the test data with a flaky result '
'the example was filtered out (by e.g. a '
'call to filter in your strategy) when we didn\'t '
'expect it to be.'
)
try:
test_runner(
TestData.for_buffer(falsifying_example),
reify_and_execute(
search_strategy,
test_is_flaky(test, repr_for_last_exception[0]),
print_example=True, is_final=True
))
except (UnsatisfiedAssumption, StopTest):
raise Flaky(filter_message)
def wrapped_test(*arguments, **kwargs):
settings = wrapped_test._hypothesis_internal_use_settings
if wrapped_test._hypothesis_internal_use_seed is not None:
random = Random(wrapped_test._hypothesis_internal_use_seed)
elif settings.derandomize:
random = Random(function_digest(test))
else:
random = new_random()
import hypothesis.strategies as sd
selfy = None
arguments, kwargs = convert_positional_arguments(
wrapped_test, arguments, kwargs)
# If the test function is a method of some kind, the bound object
# will be the first named argument if there are any, otherwise the
# first vararg (if any).
if argspec.args:
selfy = kwargs.get(argspec.args[0])
elif arguments:
selfy = arguments[0]
test_runner = new_style_executor(selfy)
for example in reversed(
getattr(wrapped_test, 'hypothesis_explicit_examples', ())):
if example.args:
if len(example.args) > len(original_argspec.args):
raise InvalidArgument(
'example has too many arguments for test. '
'Expected at most %d but got %d' %
(len(original_argspec.args), len(example.args)))
example_kwargs = dict(
zip(original_argspec.args[-len(example.args):],
example.args))
else:
example_kwargs = example.kwargs
if Phase.explicit not in settings.phases:
continue
example_kwargs.update(kwargs)
# Note: Test may mutate arguments and we can't rerun explicit
# examples, so we have to calculate the failure message at this
# point rather than than later.
message_on_failure = 'Falsifying example: %s(%s)' % (
test.__name__, arg_string(test, arguments, example_kwargs))
try:
with BuildContext() as b:
test_runner(
None,
lambda data: test(*arguments, **example_kwargs))
except BaseException:
report(message_on_failure)
for n in b.notes:
report(n)
raise
if settings.max_examples <= 0:
return
arguments = tuple(arguments)
given_specifier = sd.tuples(
sd.just(arguments),
sd.fixed_dictionaries(generator_kwargs).map(
lambda args: dict(args, **kwargs)))
def fail_health_check(message, label):
if label in settings.suppress_health_check:
return
message += (
'\nSee http://hypothesis.readthedocs.org/en/latest/health'
'checks.html for more information about this. ')
message += (
'If you want to disable just this health check, add %s '
'to the suppress_health_check settings for this test.') % (
label, )
raise FailedHealthCheck(message)
search_strategy = given_specifier
if selfy is not None:
search_strategy = WithRunner(search_strategy, selfy)
search_strategy.validate()
perform_health_check = settings.perform_health_check
perform_health_check &= Settings.default.perform_health_check
from hypothesis.internal.conjecture.data import TestData, Status, \
StopTest
if not (Phase.reuse in settings.phases
or Phase.generate in settings.phases):
return
if perform_health_check:
initial_state = getglobalrandomstate()
health_check_random = Random(random.getrandbits(128))
# We "pre warm" the health check with one draw to give it some
# time to calculate any cached data. This prevents the case
# where the first draw of the health check takes ages because
# of loading unicode data the first time.
data = TestData(max_length=settings.buffer_size,
draw_bytes=lambda data, n, distribution:
distribution(health_check_random, n))
with Settings(settings, verbosity=Verbosity.quiet):
try:
test_runner(
data,
reify_and_execute(
search_strategy,
lambda *args, **kwargs: None,
))
except BaseException:
pass
count = 0
overruns = 0
filtered_draws = 0
start = time.time()
while (count < 10 and time.time() < start + 1
and filtered_draws < 50 and overruns < 20):
try:
data = TestData(
max_length=settings.buffer_size,
draw_bytes=lambda data, n, distribution:
distribution(health_check_random, n))
with Settings(settings, verbosity=Verbosity.quiet):
test_runner(
data,
reify_and_execute(
search_strategy,
lambda *args, **kwargs: None,
))
count += 1
except UnsatisfiedAssumption:
filtered_draws += 1
except StopTest:
if data.status == Status.INVALID:
filtered_draws += 1
else:
assert data.status == Status.OVERRUN
overruns += 1
except InvalidArgument:
raise
except Exception:
report(traceback.format_exc())
if test_runner is default_new_style_executor:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
"an error in a function you've passed to "
'it to construct your data.',
HealthCheck.exception_in_generation,
)
else:
fail_health_check(
'An exception occurred during data '
'generation in initial health check. '
'This indicates a bug in the strategy. '
'This could either be a Hypothesis bug or '
'an error in a function you\'ve passed to '
'it to construct your data. Additionally, '
'you have a custom executor, which means '
'that this could be your executor failing '
'to handle a function which returns None. ',
HealthCheck.exception_in_generation,
)
if overruns >= 20 or (not count and overruns > 0):
fail_health_check((
'Examples routinely exceeded the max allowable size. '
'(%d examples overran while generating %d valid ones)'
'. Generating examples this large will usually lead to'
' bad results. You should try setting average_size or '
'max_size parameters on your collections and turning '
'max_leaves down on recursive() calls.') %
(overruns, count),
HealthCheck.data_too_large)
if filtered_draws >= 50 or (not count and filtered_draws > 0):
fail_health_check((
'It looks like your strategy is filtering out a lot '
'of data. Health check found %d filtered examples but '
'only %d good ones. This will make your tests much '
'slower, and also will probably distort the data '
'generation quite a lot. You should adapt your '
'strategy to filter less. This can also be caused by '
'a low max_leaves parameter in recursive() calls') %
(filtered_draws, count),
HealthCheck.filter_too_much)
runtime = time.time() - start
if runtime > 1.0 or count < 10:
fail_health_check(
('Data generation is extremely slow: Only produced '
'%d valid examples in %.2f seconds (%d invalid ones '
'and %d exceeded maximum size). Try decreasing '
"size of the data you're generating (with e.g."
'average_size or max_leaves parameters).') %
(count, runtime, filtered_draws, overruns),
HealthCheck.too_slow,
)
if getglobalrandomstate() != initial_state:
fail_health_check(
'Data generation depends on global random module. '
'This makes results impossible to replay, which '
'prevents Hypothesis from working correctly. '
'If you want to use methods from random, use '
'randoms() from hypothesis.strategies to get an '
'instance of Random you can use. Alternatively, you '
'can use the random_module() strategy to explicitly '
'seed the random module.',
HealthCheck.random_module,
)
last_exception = [None]
repr_for_last_exception = [None]
performed_random_check = [False]
def evaluate_test_data(data):
if perform_health_check and not performed_random_check[0]:
initial_state = getglobalrandomstate()
performed_random_check[0] = True
else:
initial_state = None
try:
result = test_runner(
data, reify_and_execute(
search_strategy,
test,
))
if result is not None and settings.perform_health_check:
fail_health_check(
('Tests run under @given should return None, but '
'%s returned %r instead.') %
(test.__name__, result), HealthCheck.return_value)
return False
except UnsatisfiedAssumption:
data.mark_invalid()
except (
HypothesisDeprecationWarning,
FailedHealthCheck,
StopTest,
):
raise
except Exception:
last_exception[0] = traceback.format_exc()
verbose_report(last_exception[0])
data.mark_interesting()
finally:
if (initial_state is not None
and getglobalrandomstate() != initial_state):
fail_health_check(
'Your test used the global random module. '
'This is unlikely to work correctly. You should '
'consider using the randoms() strategy from '
'hypothesis.strategies instead. Alternatively, '
'you can use the random_module() strategy to '
'explicitly seed the random module.',
HealthCheck.random_module,
)
from hypothesis.internal.conjecture.engine import TestRunner
falsifying_example = None
database_key = str_to_bytes(fully_qualified_name(test))
start_time = time.time()
runner = TestRunner(
evaluate_test_data,
settings=settings,
random=random,
database_key=database_key,
)
runner.run()
run_time = time.time() - start_time
timed_out = (settings.timeout > 0 and run_time >= settings.timeout)
if runner.last_data is None:
return
if runner.last_data.status == Status.INTERESTING:
falsifying_example = runner.last_data.buffer
if settings.database is not None:
settings.database.save(database_key, falsifying_example)
else:
if runner.valid_examples < min(
settings.min_satisfying_examples,
settings.max_examples,
):
if timed_out:
raise Timeout(
('Ran out of time before finding a satisfying '
'example for '
'%s. Only found %d examples in ' + '%.2fs.') %
(get_pretty_function_description(test),
runner.valid_examples, run_time))
else:
raise Unsatisfiable(
('Unable to satisfy assumptions of hypothesis '
'%s. Only %d examples considered '
'satisfied assumptions') % (
get_pretty_function_description(test),
runner.valid_examples,
))
return
assert last_exception[0] is not None
try:
with settings:
test_runner(
TestData.for_buffer(falsifying_example),
reify_and_execute(search_strategy,
test,
print_example=True,
is_final=True))
except (UnsatisfiedAssumption, StopTest):
report(traceback.format_exc())
raise Flaky(
'Unreliable assumption: An example which satisfied '
'assumptions on the first run now fails it.')
report(
'Failed to reproduce exception. Expected: \n' +
last_exception[0], )
filter_message = (
'Unreliable test data: Failed to reproduce a failure '
'and then when it came to recreating the example in '
'order to print the test data with a flaky result '
'the example was filtered out (by e.g. a '
'call to filter in your strategy) when we didn\'t '
'expect it to be.')
try:
test_runner(
TestData.for_buffer(falsifying_example),
reify_and_execute(search_strategy,
test_is_flaky(
test, repr_for_last_exception[0]),
print_example=True,
is_final=True))
except (UnsatisfiedAssumption, StopTest):
raise Flaky(filter_message)
