Cast numpy integer into Mojo integer

Mojo
,
1

I’m updating this little library I have going to 25.1, and am having a bit of trouble with casting from numpy types to Mojo types.

Essentially I used to be able to cast numpy.uint8 to a Mojo Int as:
py_array[rows][columns][channels].__int__(), where the elements of py_array are numpy.uint8.

But now I get:
Unhandled exception caught during execution: an integer is required

.__as_int__() has the same behaviour, as well as Int(py_array[rows][columns][channels]). This worked at least up to 24.5, so I am trying to see in the Changelog if anything about typing changed in the meantime, but nothing stands out to me.

In the function I am trying to fix, I call I am comparing the values in a numpy array to the values in a Mojo tensor.

The numpy array is initialized as var py_array = np.array(PillowImage.open(image_path))

So is there a new/better way of casting numpy types to Mojo?

2

Opened a bug report: [BUG] Numpy ints no longer cast to Mojo · Issue #4119 · modular/max · GitHub

3

Thanks @ferdinand have a fix landed that passes your tests PR: [stdlib] Add tests for python and numpy casting by fnands · Pull Request #4120 · modular/max · GitHub

Will land those tests when the fix lands in the max repo, thanks for adding tests much appreciated.

2 Likes