Bug summary

In testing type-checking integration between matplotlib 3.8.0rc1 andcmyt, I found mypy flagged a couple errors that I think may be spurious ?

Here's a minimal reproducer of the issue I'm talking about.

Code for reproduction

# t.py , linted with `mypy t.py`frommatplotlib.colorsimportColormapimportnumpyasnpdefrgb_from_cmap(cmap:Colormap)->np.ndarray:RGBA=cmap(np.linspace(0,1,155))RGB=RGBA[:, :3]returnRGB

Actual outcome

t.py: note: In function "rgb_from_cmap":t.py:5: error: No overload variant of "__getitem__" of "tuple" matches argument type "Tuple[slice, slice]"  [call-overload]t.py:5: note: Possible overload variants:t.py:5: note:     def __getitem__(self, SupportsIndex, /) -> floatt.py:5: note:     def __getitem__(self, slice, /) -> Tuple[float, ...]t.py:6: error: Incompatible return value type (got "Union[float, ndarray[Any, Any]]", expected "ndarray[Any, Any]")  [return-value]Found 2 errors in 1 file (checked 1 source file)

Expected outcome

no error. The second one is easy enough to handle in user code by returningnp.asarray(RGB) (though I am not 100% sure that this should be necessary), but the first error really feels undesired to me;

Quoting fromColormap.__call__'s docstring:

   Returns   -------   Tuple of RGBA values if X is scalar, otherwise an array of   RGBA values with a shape of ``X.shape + (4, )``.

it should be clear that the return value is necessarily an array withndim >= 2 in my case ?
I think the problem is that the return type of this function is annotated as a union type

However it's not clear to me how to fix it.

Additional information

previous versions of matplotlib "type-checked" correctly (because everything was assumed to beAny !)

Operating system

ubuntu

Matplotlib Version

3.8.0rc1

Matplotlib Backend

N/A

Python version

3.11.4

Jupyter version

N/A

Installation

pip