Sorry, something went wrong.
|
Marking as "DO-NOT-MERGE" for now, as this appears more controversial than I realised. |
Sorry, something went wrong.
|
I've removed the tests that assert undesirable behaviour, as per the consensus in python/typing#1363. |
Sorry, something went wrong.
| class C: | ||
| @property | ||
| def attr(self): | ||
| return 42 |
There was a problem hiding this comment.
What about side-effects in @property like raise ValueError? Should we test this case?
Sorry, something went wrong.
There was a problem hiding this comment.
We should test this case, definitely! But the behaviour for this case may be about to change if any of the patches discussed in python/typing#1363 is implemented (and the consensus seems to be that we should implement one of those patches). If so, we should add the tests in the same PR as we change the behaviour.
Whether or not we decide to change the behaviour around properties that have side effects, I'd prefer to add those tests in a separate PR, so that this PR is solely focussed on adding tests for uncontroversial behaviour.
Sorry, something went wrong.
There was a problem hiding this comment.
A couple nits, but these tests look good to me! Thank you!
Sorry, something went wrong.
| class BadPG1(Protocol[T]): | ||
| attr: T | ||
|
|
||
| for obj in PG[T], PG[C], PG1[T], PG1[C], BadP, BadP1, BadPG, BadPG1: |
There was a problem hiding this comment.
nit: these are all protocol classes still; the name 'protocol_class used in the above "good" loop seems much clearer than the name obj
Sorry, something went wrong.
There was a problem hiding this comment.
No strong opinion here, but the reason I avoided "protocol_class" for these ones is that the parameterised ones (PG[T], etc) aren't strictly speaking classes anymore — they're generic aliases to protocol classes.
Sorry, something went wrong.
|
Thanks @AlexWaygood for the PR 🌮🎉.. I'm working now to backport this PR to: 3.10, 3.11. |
Sorry, something went wrong.
|
GH-102592 is a backport of this pull request to the 3.11 branch. |
Sorry, something went wrong.
|
GH-102593 is a backport of this pull request to the 3.10 branch. |
Sorry, something went wrong.
|
Thanks, all! |
Sorry, something went wrong.
It appears we currently have no tests for how classes with properties interact with isinstance checks for protocols decorated with @runtime_checkable.
I'm trying to only add tests here for uncontroversial behaviour that we won't want changed, whether or not any of the patches discussed in python/typing#1363 is implemented.