Change sparse matrix to array

[Lakshmi-bashyam]

Replace spmatrix with sparray for SciPy compatibility

This pull request updates all references to scipy.sparse.spmatrix to use the new scipy.sparse.sparray class, in line with SciPy's ongoing deprecation of spmatrix. This change ensures compatibility with recent and future versions of SciPy.

Changes Made

• Replaced all instances of spmatrix type checks and imports with sparray.
• Modified test cases to make them compatible with the new change.

Related Issue

Fixes stwfsapy#89

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (cf652c4) to head (04e04fc).
⚠️ Report is 5 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##            master       #96   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           16        16           
  Lines          943       943           
=========================================
  Hits           943       943           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[gmmajal]

Can you make the changes specified for the doctrings and further clarify if text_features needs to be modified?

[gmmajal]

I think in the docstrings for the predict() method we can replace "A sparse matrix of shape ..." with a sparse array for the sake of consistency.

[gmmajal]

Can you explain why over here(line 433) and in line 458, when the transform method is applied we access index 0. I see that it was modified for the text_vectorizer attribute i.e. index 0 is not accessed. Should it be also modified for the text_features attribute or does it have a different data structure?

[gmmajal]

I looked at what the transform methods are doing for text_vectorizer and text_features, respectively. The one for text_vectorizer returns a sparse matrix whereas for text_features a numpy array is returned. The data structure does indeed seem to be different.

[Lakshmi-bashyam]

The text vectorizer produces a csr_matrix from scikit-learn, so we can’t switch it to a sparray at this point.

[Lakshmi-bashyam]

We can’t safely transition from spmatrix to the sparray hierarchy just yet. Our dependency on scikit-learn still poses compatibility risks. While scikit-learn has begun its migration toward supporting sparray, the internal transition is still in progress.

Specifically, scikit-learn’s PR #31072 (“First steps toward sparray migration pass 2”) is still open, indicating that full adoption isn’t complete yet.

[gmmajal]

We can’t safely transition from spmatrix to the sparray hierarchy just yet. Our dependency on scikit-learn still poses compatibility risks. While scikit-learn has begun its migration toward supporting sparray, the internal transition is still in progress.

Specifically, scikit-learn’s PR #31072 (“First steps toward sparray migration pass 2”) is still open, indicating that full adoption isn’t complete yet.

Good catch! I had a look at the pull request you referenced. The team at scikit-learn are in the process of migrating as you mentioned. As part of their release v1.8, they have this particular pull request: scikit-learn/scikit-learn#31177. The release is expected to be available by mid Nov 2025, see here: https://github.com/scikit-learn/scikit-learn/milestone/66. We can wait till scikit-learn has a transition mechanism in place before migrating to sparray ourselves.