Side note: I could not find good documentation on default integer/float/complex type besides the somewhat confusing https://numpy.org/devdocs/numpy_2_0_migration_guide.html#windows-default-integer; perhaps we could add a subsection somewhere in this file?

Sorry, I was out while this happened. But would also value someone else's input here.

To me, this is a detail about result_type. And I feel that it should be mentioned there mainly?
(It is also something that we probably should consider changing there!)

The default types are documented in the list of types, I think. But, I suppose we could move the footnote about it up and just state the fact that the default is int64 on 64bit systems (which are most systems, so we use that).

Thanks, I added a reference to the promotion list in 904a6f5. I also reworded the footnote slightly.

To me, this is a detail about result_type. And I feel that it should be mentioned there mainly?
(It is also something that we probably should consider changing there!)

Do you mean, should this whole subsection be moved to the result_type doc, or just the default integer discussion?

I was wondering about this whole addition. The reason is that I don't think we mention the syntax of result_type to allow dtypes at all (which is the crux, int here would ideally not be a dtype instance, but an abstract DType class/type. But result_tyee uses np.dtype(int) which does a silly guess at the default).

So I am wondering if adding it here just adds another layer of complexity that is mainly needed with result_type. OTOH, explaining result_type in more detail here could make sense...

I see. Just as an additional wrinkle, promote_types also accepts int...