ENH: Add annotations for np.lib.ufunclike
#18377
Conversation
| @@ -189,7 +189,8 @@ def isposinf(x, out=None): | |||
| try: | |||
| signbit = ~nx.signbit(x) | |||
| except TypeError as e: | |||
| raise TypeError('This operation is not supported for complex values ' | |||
| dtype = nx.asanyarray(x).dtype | |||
| raise TypeError(f'This operation is not supported for {dtype} values ' | |||
There was a problem hiding this comment.
The exception message would previously assume that that the dtype of the (invalid) array was complex. This is not necessarily true though, as the very same exception will also be triggered by timedelta64 and datetime64 arrays.
| x: _ArrayLikeFloat_co, | ||
| out: None = ..., | ||
| ) -> _ArrayOrScalar[floating[Any]]: ... | ||
| @overload |
There was a problem hiding this comment.
Be good to have 2 blank lines between functions (PEP8).
There was a problem hiding this comment.
I know you are trying to group the functions, but I find it more difficult to read without the blank lines.
There was a problem hiding this comment.
For overloads it's generally common practice to exclude the blank lines (example from typeshed).
Note that this area is not quite covered by PEP 8 though: the description of a single function with multiply signatures.
|
Just some style nits, otherwise LGTM. |
|
Thanks Bas. |
Per the title: this PR adds annotations for
np.lib.ufunclike.DType support also turned out to be fairly straightforward to implement here, so that was definitely a plus.