Conversation
|
Adding Win/Py3.13 to CI matrix to see if we pass there now. |
|
This seems to work, but it is likely that numpy 2 will break outer user workflows. It would be prudent to test well before merging. |
|
Thanks for working on this! I just tried it out, and it works great on my end. I’m happy to provide an approval, but just wanted to check first if it would be good for me to test anything specific? I tested this on a Windows computer using Python 3.13.5. I’ve attached a notebook with the workflow I used. The workflow is pretty basic, so I’d be happy to test additional workflow items as well. |
|
The major thing is that I believe downstream packages/usage might break, but we can't insulate against that. Particularly users that frequently cast to np.array() will find that there are problems with the new array primitives, but that will hopefully help reduce that - it's bad practice anyway. |
|
The CI testing covers the usage, the main question is "does the change make sense", at least that's how I look at it. |
|
Ah, okay. The tracked changes seem fine to me, but I don't think I'm very knowledgeable on what all the implications might be. |
2 participants
#208
This branch is an attempt to understand what the pain point(s) are with numpy2.