TYP: Use Self instead of class-bound TypeVar (generic.py/frame.py/series.py) #51493
Conversation
topper-123
changed the title
topper-123
force-pushed
the
use_typing.Self
branch
from
5089ea8 to
089920e
Compare
topper-123
force-pushed
the
use_typing.Self
branch
from
089920e to
86b5fad
Compare
pandas/core/indexes/base.py
Outdated
|
|
||
| @final | ||
| def _left_indexer_unique(self: _IndexT, other: _IndexT) -> npt.NDArray[np.intp]: | ||
| def _left_indexer_unique(self, other: Index) -> npt.NDArray[np.intp]: |
There was a problem hiding this comment.
Is this change intended? Is _left_indexer_unique(self, other: Self) to strict?
There was a problem hiding this comment.
(same for the following methods)
There was a problem hiding this comment.
Yeah, I was in doubt about this too. Fir the built-in index classes I think it will always require same class (so other: Self), but custom user-created classes:
>>> MyIndex(Index):
... passIn this case, MyIndex has same dtype as the base index, so should be comparable to Index in this method? (I'm leaning to yes; that this should be possible, but I'm not 100% sure).
There was a problem hiding this comment.
For the built-in index classes I think it will always require same class (so
other: Self)
I would then stick with Self it is also stricter which could help catch bugs (or causes a false mypy error which can then be fixed by someone who knows for sure).
|
Thanks @topper-123! |
3 participants
Progress towards #51233.
There were more updates to do than I expected, so I split it up a bit. In this PR I do
pandas/core/generic.py,pandas/core/frame.py,pandas/core/series.py&pandas/indexes/.