✨ HasDType, Array

[nstarman]

Relies on #47

A new PR to precede #32.

[jorenham]

sorry if this is premature, but maybe these comments can help 🤷🏻

[jorenham]

is this needed for the redefinition of _?

[nstarman]

Yes.

[nstarman]

I didn't want to name all the variables.

[jorenham]

There are only two hard things in Computer Science: cache invalidation and naming things.

-- Phil Karlton

[jorenham]

Or even better:

There are 2 hard problems in computer science: cache invalidation, naming things, and off-by-1 errors.

-- Leon Bambrick

[nstarman]

numpy 2.0's array_api stuff is infinitely better than 1.0's. It's a real struggle to write tests that both pass and test what we want. 😮‍💨

[nstarman]

@jorenham. Ok. this PR was much easier than #32 and hopefully will help with that one, particularly in writing more comprehensive tests. Might go one Can* binop at a time given some of the complexity with bool-dtyped arrays.
But we should probably have more array libraries being tested... and other type checkers doing the testing...

[nstarman]

Pushed to add docs to HasArrayNamespace

[nstarman]

Forgot to have Array inherit from HasArrayNamespace. That's kind of the most important one 😆

[nstarman]

Ok. @jorenham now it should really be ready for a review.

[jorenham]

numpy 2.0's array_api stuff is infinitely better than 1.0's. It's a real struggle to write tests that both pass and test what we want. 😮‍💨

Maybe we could use array-api-compat for numpy<2.0?

[nstarman]

That's a good idea. After this PR because things are passing?

[jorenham]

But we should probably have more array libraries being tested... and other type checkers doing the testing...

Yea, I'll get on that soon (after I fix optype and scipy-stubs apparently being broken on numpy<2.3 😅).

[jorenham]

should be good to merge