Sebastian Berg

Right, the thing that I misunderstood, is that I had the impression that new behavior was chosen because the old behavior was very strange. But that seems wrong, I am...

Closing, this is reverted. Note that the shape *did* change in 2.0 for the first case you mentioned and will stay changed. The first example actually *didn't* work previously.

> So my question is: should numpy.isdtype support checking custom numpy dtypes? Yes this should definitely work (or at least not fail the way it does). There is the caveat...

FWIW, you can use `fromiter` if you use a structured dtype, but that only gives you a ~30% speedup or so (might be slower also due to the structured part...

Note that to do this, you will need to refine the `__array__` protocol to takek precedence in subclass cases: ``` if type(subclass).__array__ != ndarray.__array__: arr = subclass.__array__() ``` (strictly speaking...

> Is it really worth it to force all those users to change their code? In contrast, __array_function__ would solve the problem in np.ma itself. Right, this is what I...

Just to note that CuPy is compiling to C++, so it now switched to C++ mutexes to avoid the dependency and unlock 3.14. Which doesn't mean that 3.14 support isn't...

Well, we never really adopted NEP 29 either I believe, which is why it is in the "informational" category :). Just like SPEC 0 it was always a recommendation and...

Hmmm, I feel there was a time where I had occasional errors similar to this locally. But I don't think there were in `matmul`. (I feel I might have tried...

I was faintly aware, but not that it was an issue here. A solution may be to make sure that the default fill value is typed to e.g. int64. That...