Honor masks in PyTorch nonzero

[FlorianPfaff]

Bug

The PyTorch backend wrapper converted NumPy MaskedArray inputs with torch.as_tensor. That conversion discards the mask and exposes the hidden payload, so nonzero could report masked entries as present.

For example, a masked 2×2 input whose only unmasked nonzero value is at (1, 0) returned (0, 1), (1, 0), and (1, 1) on the PyTorch backend.

Fix

• fill masked entries with zero before converting to a PyTorch tensor;
• preserve the existing rejection of zero-dimensional inputs;
• preserve ordinary list, NumPy-array, and tensor handling;
• add regression coverage for both the public and raw PyTorch backend helpers.

Validation

• reproduced the pre-fix mask loss with NumPy 2.3.5 and PyTorch 2.10.0;
• verified the patched helper matches numpy.nonzero for the masked input;
• verified existing scalar rejection and ordinary matrix behavior in the isolated harness;
• syntax-compiled both modified files;
• branch comparison: 2 commits ahead, 0 behind main, with only 2 files changed.

The full supported backend and Python-version matrix is delegated to GitHub Actions.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	20.05s
✅ JSON	prettier	7	0	0	0	0.79s
✅ JSON	v8r	7		0	0	5.0s
✅ MARKDOWN	markdownlint	68	0	0	0	1.24s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.58s
✅ PYTHON	black	1809	58	0	0	64.98s
✅ PYTHON	isort	1809	90	0	0	2.45s
✅ REPOSITORY	betterleaks	yes		no	no	1.95s
✅ REPOSITORY	checkov	yes		no	no	38.81s
✅ REPOSITORY	gitleaks	yes		no	no	11.35s
✅ REPOSITORY	git_diff	yes		no	no	0.11s
✅ REPOSITORY	secretlint	yes		no	no	59.25s
✅ REPOSITORY	syft	yes		no	no	4.48s
✅ REPOSITORY	trivy-sbom	yes		no	no	4.47s
✅ REPOSITORY	trufflehog	yes		no	no	25.18s
✅ YAML	prettier	11	0	0	0	0.63s
✅ YAML	v8r	11		0	0	9.2s
✅ YAML	yamllint	11		0	0	0.42s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.6.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository

[FlorianPfaff]

Closing as low value. This introduces NumPy masked-array semantics into the PyTorch nonzero compatibility layer, while masked arrays are not a native PyTorch input contract and are not supported consistently across PyRecEst. Isolated mask handling here would make backend behavior less coherent.