Viewing file:  check_raises.py (4.27 KB) -rw-r--r--
Select action/file-type:
 (+) |  (+) |  (+) | Code (+) | Session (+) |  (+) | SDB (+) |  (+) |  (+) |  (+) |  (+) |  (+) |
from __future__ import annotations
from typing import Iterable, Any
from .check_log import log_failure
_stop_on_fail = False
# TODO: Returning Any isn't ideal, but returning CheckRaisesContext | None
# would require callers to type ignore or declare the type when using `with`.
# Or, it could always return CheckRaisesContext, just an empty one after
# calling the passed function.
def raises(
expected_exception: type | Iterable[type], *args: Any, **kwargs: object
) -> Any:
"""
Check that a given callable or context raises an error of a given type.
Can be used as either a context manager:
>>> with raises(AssertionError):
>>> raise AssertionError
or as a function:
>>> def raises_assert():
>>> raise AssertionError
>>> raises(AssertionError, raises_assert)
`expected_exception` follows the same format rules as the second argument
to `issubclass`, so multiple possible exception types can be used.
When args[0] is callable, the remainder of args and all of kwargs except
for any called `msg` are passed to args[0] as arguments.
Note that because `raises` is implemented using a context manager, the
usual control flow warnings apply: within the context, execution stops on
the first error encountered *and does not resume after this error has been
logged*. Therefore, the line you expect to raise an error must be the last
line of the context: any subsequent lines won't be executed. Pull such
lines out of the context if they don't raise errors, or use more calls to
`raises` if they do.
This function is modeled loosely after Pytest's own `raises`, except for
the latter's `match`-ing logic. We should strive to keep the call
signature of this `raises` as close as possible to the other `raises`.
"""
__tracebackhide__ = True
if isinstance(expected_exception, type):
expected_exceptions: Iterable[type] = (expected_exception,)
else:
expected_exceptions = expected_exception
assert all(
isinstance(exc, type) or issubclass(exc, BaseException)
for exc in expected_exceptions
)
msg = kwargs.pop("msg", None)
if not args:
assert not kwargs, f"Unexpected kwargs for pytest_check.raises: {kwargs}"
return CheckRaisesContext(*expected_exceptions, msg=msg)
else:
func = args[0]
assert callable(func)
with CheckRaisesContext(*expected_exceptions, msg=msg):
func(*args[1:], **kwargs)
class CheckRaisesContext:
"""
Helper context for `raises` that can be parameterized by error type.
Note that CheckRaisesContext is instantiated whenever needed; it is not a
global variable like `check`. Therefore, we don't need to curate
`self.msg` in `__exit__` for this class like we do with
CheckContextManager.
"""
def __init__(self, *expected_excs: type, msg: object = None) -> None:
self.expected_excs = expected_excs
self.msg = msg
def __enter__(self) -> "CheckRaisesContext":
return self
def __exit__(self, exc_type: type, exc_val: object, exc_tb: object) -> bool:
__tracebackhide__ = True
if exc_type is not None and issubclass(exc_type, self.expected_excs):
# This is the case where an error has occured within the context,
# but it is the type we're expecting. Therefore, we return True
# to silence this error and proceed with execution outside the
# context.
return True
if not _stop_on_fail:
# Returning something falsey here will cause the context
# manager to *not* suppress an exception not in
# `expected_excs`, thus allowing the higher-level Pytest
# context to handle it like any other unhandle exception during
# test execution, including display and tracebacks. That is the
# behavior we want when `_stop_on_fail` is True, so we let that
# case fall through. If *not* `_stop_on_fail`, then we want to
# log the error as a failed check but then continue execution
# without raising an error, hence `return True`.
log_failure(self.msg if self.msg else exc_val)
return True
# Stop on fail, so return False
return False
|