[Python][Doc] Add from_numpy examples for sparse tensor constructors

[chilin0525]

Describe the enhancement requested

Follow-up from #49147 (comment: #49147 (comment)).

PR #49147 added class-level docstring examples for SparseCOOTensor, SparseCSRMatrix, SparseCSCMatrix, and SparseCSFTensor. Currently, all examples only demonstrate construction via from_dense_numpy(). As @rok suggested, it would be useful to also add examples for the non-dense constructors that build sparse tensors directly from sparse data structures.

The following factory methods in python/pyarrow/tensor.pxi are missing docstring examples:

SparseCOOTensor

• from_numpy(data, coords, shape, dim_names=None)

SparseCSRMatrix

• from_numpy(data, indptr, indices, shape, dim_names=None)

SparseCSCMatrix

• from_numpy(data, indptr, indices, shape, dim_names=None)

SparseCSFTensor

• from_numpy(data, indptr, indices, shape, axis_order=None, dim_names=None)

Component(s)

Python

[rok]

Thanks for working on this @chilin0525 !
I would suggest to only add from_numpy examples as they demonstrate the optimal use of the underlying data structures. And covering others would just be unnecessarily verbose.

[chilin0525]

@rok Thanks for the suggestion! I’ll update the issue title and adjust the content accordingly.

[chilin0525]

take