array-api
array-api copied to clipboard
data-apis
RFC document, tooling and other content related to the array API standard
The `clip` function is quite handy for constraining data to a certain range. Would be nice to include this in the spec
https://github.com/data-apis/array-api/blob/main/spec/API_specification/array_object.rst#17 says: > Furthermore, a conforming implementation of the array API standard must support array objects of arbitrary rank `N` (i.e., number of dimensions), where `N` is greater than or...
In the 2021.12 standard it is here: https://data-apis.org/array-api/2021.12/API_specification/data_types.html#float64, which says it "will be added to Type Promotion Rules". But it disappears in both Data Types and Type Promotion Rules sections...
`xp.__array_api_version__` has been specified in our [Versioning](https://data-apis.org/array-api/latest/future_API_evolution.html#versioning) section for a while now (~2yrs), but no such mention of it is found in the API specification section, which is probably why...
This RFC proposes that n-dimensional FFT APIs should support specifying a placeholder value when specifying the shape of the transformed output. ## Prior Art In [PyTorch](https://pytorch.org/docs/stable/generated/torch.fft.fftn.html#torch.fft.fftn), when specifying the shape...
[This table](https://data-apis.org/array-api/draft/API_specification/type_promotion.html#floating-point-type-promotion-table) needs to be updated to include mixed real and complex dtypes (making it a 4x4 table). Currently the complex type promotion is completely unspecified, and the [lattice diagram](https://github.com/data-apis/array-api/blob/main/spec/_static/images/dtype_promotion_complex.png)...
Functions like `full` and `ones` and maybe others don't have positional only arguments in their definition. I wanted to know the reason behind this. https://data-apis.org/array-api/latest/API_specification/creation_functions.html?highlight=positional
We've had discussions several times about whether some function that is a composite of other functions already present in the standard can be added. The most recent example is gh-460,...
Ok, the idea of this PR is to maybe resolve #411. A few things have come up that might make folk think ultimately a first-party package is a bad idea....
I was not around when this may have been discussed in https://github.com/numpy/numpy/pull/21912. In-place matmul is a weird beast, but for NumPy it would be strange to make `@=` and out-of-place...
