Fix edge cases of csr_array

[aarmey]

A csr_array can have non-sparse zeros potentially.

[Copilot]

Pull Request Overview

This PR improves get_deviance by ensuring sparse data has no redundant or zero entries and by simplifying how expected counts (mu_ij_nn) are computed.

• Adds in-place deduplication and zero-elimination on the input CSR matrix.
• Updates mu_ij_nn calculation to use 1D indexing on flattened row/column sums.

Comments suppressed due to low confidence (2)

parafac2/normalize.py:83

• sum_duplicates() and eliminate_zeros() modify the input in place. If callers of get_deviance rely on the original matrix, consider working on a copy or clearly document this side effect.

    data.sum_duplicates()

parafac2/normalize.py:110

• If n_i or pi_j are 2D arrays (e.g., numpy matrices), indexing with [row]/[col] could yield unexpected shapes. Ensure these sums are explicitly flattened (e.g., via ravel() or squeeze()) so that elementwise multiplication produces a 1D array of scalars.

    mu_ij_nn = n_i[row] * pi_j[col]

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.

Project coverage is 91.45%. Comparing base (ccf708f) to head (8a3de9c).

Files with missing lines	Patch %	Lines
parafac2/normalize.py	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #39      +/-   ##
==========================================
- Coverage   91.87%   91.45%   -0.43%     
==========================================
  Files          10       10              
  Lines         431      433       +2     
==========================================
  Hits          396      396              
- Misses         35       37       +2     

Flag	Coverage Δ
unittests	91.45% <0.00%> (-0.43%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.