🌐 Array API support tracker

[jorenham]

upstream tracker: scipy/scipy#18867

Typing the array API

... that'll have to wait until someone manages to find the time to work on array-api-typing 😅

Relevant scipy bits

The Array API is partially supported scipy. The following packages support it fully:

• cluster
  • vq
    • whiten
    • [py_]vq
    • kmeans[2]
  • hierarchy
    • int_floor
    • linkage (used by single, complete, average, weighted, centroid, median, ward)
    • {cut,to}_tree
    • optimal_leaf_ordering
    • cophenet
    • inconsistent
    • {from,to}_mlab_linkage
    • is_monotonic
    • is_valid_{im,linkage}
    • num_obs_linkage
    • correspond
    • fcluster[data]
    • leaves_list
    • dendrogram
    • max{dists,inconsts,Rstat}
    • leaders
• constants
  • convert_temperature
• datasets (trivial)
• differentiate [1.15+]
  • derivative
  • jacobian
  • hessian
• fft
  • [i][r|h]fft[2|n]
  • [i]dct[n]
  • [i]dst[n]
  • [r]fftfreq
  • [i]fftshift
• integrate [1.15+]
  • cubature
  • tanhsinh
  • trapezoid
  • nsum
• io (trivial)
• linalg (not yet, see scipy/scipy#19068)
• ndimage (not yet, see scipy/scipy#21280)
• optimize
  • _differentiable_functions
  • elementwise.* [1.15+]
  • rosen[_{der,hess}] [1.15+]
• signal (not yet, see scipy/scipy#20678)
• sparse (not yet, see scipy/scipy#18867)
• special
  • log_ndtr
  • ndtr[i]
  • chdtr [1.15+]
  • chdtrc
  • erf[c]
  • i0[e]
  • i1[e]
  • betainc[c] [1.15+]
  • gammaln
  • gammainc[c]
  • {log,exp}it
  • [rel_]entr
  • xlogy
• stats (see scipy/scipy#20544)
  • describe
  • moment
  • variation
  • sem
  • skew
  • kurtosis
  • pearsonr
  • kstat[var]
  • z{map,score} [1.15+]
  • gzscore [1.15+]
  • t{mean,var,str,sem,min,max} [1.15+]
  • {g,h,p}mean [1.15+]
  • entropy
  • differential_entropy [1.15+]
  • power_divergence
  • circ{mean,var,std}
  • directional_stats [1.15+]
  • ttest_1samp
  • ttest_{ind,rel} [1.15+]
  • {skew,kurtosis,normal,monte_carlo}test
  • chisquare
  • jarque_bera
  • bartlett
  • boxcox_llf [1.15+]
  • combine_pvalues [1.15+]

How Scipy implements it

Internally, scipy implements the support in scipy._lib._array_api. From the looks of it, this module will come with a bunch of extra utility functions in scipy 1.15 (37174ad), as well as an additional scipy._lib._array_api_no_0d module in 281a8c3

Libraries that support it

Some libraries require array-api-compat for them to support the array-api (2022):

• numpy < 2
• cupy
• pytorch
• dask > 2023.12.0
• ndonnx

the following libraries have native support:

• numpy >= 2
• cupy (but not really)
• jax
• sparse
• ndonnx

Other relevant links:

• https://docs.scipy.org/doc/scipy/dev/api-dev/array_api.html
• https://data-apis.org/array-api/latest/index.html
• https://github.com/data-apis/array-api