You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is conceptually independent of the immediately preceding
super() call refactoring, but serves the same goal of simplifying
and clarifying super() calls.
In test.performance.lib, the TestBigRepoR and TestBigRepoRW
classes' setUp methods had calls to setUp through super proxies,
which were wrapped in try-blocks to swallow AttributeError
exceptions. This removes that, relying on the presence of setUp
methods in some parent or sibling class in the MRO.
The intent appeared to be solely to account for the possibility
that no class in the MRO would define a setUp method. However,
the unittest.TestCase base class defines noop setUp and tearDown
methods to ensure this does not have to be done.
This may also make the code more robust, because the form in which
AttributeError was being swallowed was:
try:
super().setUp()
except AttributeError:
pass
But that has the disadvantage of also catching AttributeError due
to a bug or other problem in code that runs *in* an ancestor or
sibling class's existing setUp method. This could alternatively
be addressed by using:
try:
other_setUp = super().setUp
except AttributeError:
pass
else:
other_setUp()
However, because unittest.TestCase provides empty setUp and
tearDown methods to allow such special-casing to be avoided (both
in cases like this and for the test runner), this isn't needed.
0 commit comments