QuadDtype Wishlist

Question

QuadDtype Wishlist

Opened this issue 4 months ago · 14 comments

Answer 1 · 2025-07-21T16:12:54.000Z

It'd also be nice to have proper API docs for the Python API generated by e.g. Sphinx. We should also make sure there are docstrings (preferably doctested in CI) on the full public Python API.

Answer 2 · 2025-07-21T16:22:19.000Z

Once we get to writing documentation, we should also document the guaranteed rounding error in ULPs for each operation - thankfully SLEEF documents it fully - since people who need float128, myself included, then likely also care about rounding errors

Answer 3 · 2025-09-08T10:18:45.000Z

@ngoldbaum @SwayamInSync Would it be possible to implement np.issubdtype(dtype, np.floating) support by making QuadPrecision a subclass of np.floating? While something like finfo would still be blocked on other issues, it would still be an improvement

Answer 4 · 2025-09-08T11:04:35.000Z

It is possible and we did that back when it is in early development, and it leads to some issues like np.finfo got populated with the float64 properties, which later leads to some precision issues in tests

Answer 5 · 2025-09-08T13:47:32.000Z

I encourage you both to not think of NumPy as a static thing that can't be changed or fixed. We should definitely make finfo play nicer with user dtypes, it'll just take some work inside NumPy to get that going. Happy to mentor on contributing to NumPy.

Answer 6 · 2025-09-08T14:07:26.000Z

it'll just take some work inside NumPy to get that going. Happy to mentor on contributing to NumPy.

Yes after fixing the packaging and setting things up on conda-forge, I'll be studying the DType API more on extending to these ideas. AFAIK JAX team also got these issues and they are using monkey patching for this. We can do that too but it'll be ugly

Answer 7 · 2025-10-22T07:42:28.000Z

@SwayamInSync Could we add a test for np.nan_to_num now working (since you added finfo support)?

Answer 8 · 2025-10-22T07:42:59.000Z

@SwayamInSync Is it still true that the dtype character will be '\0'?

Answer 9 · 2025-10-22T08:36:59.000Z

@SwayamInSync Could we add a test for np.nan_to_num now working (since you added finfo support)?

It works but array printing is having some issues, the operation itself works. I'll keep this in mind and debug

Answer 10 · 2025-10-22T08:37:57.000Z

@SwayamInSync Is it still true that the dtype character will be '\0'?

the kind is '\x00'

Answer 11 · 2025-10-22T11:01:02.000Z

@SwayamInSync Could we add a test for np.nan_to_num now working (since you added finfo support)?

It works but array printing is having some issues, the operation itself works. I'll keep this in mind and debug

It'll work but in the environment following this has to be set

np.set_printoptions(formatter={'float_kind': lambda x: str(x)})

Otherwise the operation will properly work, its just you can't be able to print it

Answer 12 · 2025-10-24T21:48:55.000Z

having a numpy_quaddtype.__version__ would be nice

Answer 13 · 2025-10-27T23:22:22.000Z

Are there any plans for implementing all and any yet? Because I noticed that calling QuadPrecision(1).{all,any}() raises a UFuncTypeError at the moment.

Answer 14 · 2025-10-27T23:41:08.000Z

Are there any plans for implementing all and any yet? Because I noticed that calling QuadPrecision(1).{all,any}() raises a UFuncTypeError at the moment.

Because they underline call the logical_and and logical_or ufunc with <class 'numpy.dtypes.BoolDType'>, <class 'numpy_quaddtype.QuadPrecDType'>) -> <class 'numpy.dtypes.BoolDType'>
BUT currently those ufunc are only set to accept the arguments of type {&QuadPrecDType, &QuadPrecDType, &PyArray_BoolDType}

Thanks a lot for noticing this case, I'll fix it

QuadDtype Wishlist

Constants

Ufuncs

Dtype

Binary