Reject masked JAX norm axes

[FlorianPfaff]

What changed

• reject genuinely masked NumPy axis values before JAX/static-axis coercion
• cover direct masked axes and masked scalar entries nested in lists or tuples
• preserve valid ordinary static-axis inputs and fully unmasked nested scalar axes
• add a focused backend-portable regression

Root cause

The JAX linalg.norm compatibility layer normalizes list and tuple axes entry by entry. For a masked NumPy scalar such as np.ma.array(1, mask=True), operator.index() exposes the hidden payload. Consequently, axis=[np.ma.array(1, mask=True)] was silently normalized to (1,) and the norm was evaluated along axis 1 instead of rejecting a missing axis value.

Direct masked arrays could also leak backend-specific ValueErrors during JAX conversion rather than producing the documented axis validation error.

Impact

Missing or masked configuration can no longer become a real norm axis through its hidden payload. Callers now receive a deterministic TypeError for all genuinely masked axis forms.

Validation

• reproduced the pre-fix nested-axis behavior: a masked scalar payload of 1 produced [5.0, 13.0]
• exercised the patched normalizer against scalar, vector, nested-list, and nested-tuple masked axes
• verified all masked cases raise the documented TypeError
• verified existing NumPy/JAX scalar, array, list, and tuple axis forms still produce [5.0, 13.0]
• final comparison is 2 commits ahead and 0 behind main, limited to 4 source lines and one focused regression file

GitHub Actions will run the repository's full supported backend and Python-version matrix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	19.06s
✅ JSON	prettier	7	0	0	0	1.03s
✅ JSON	v8r	7		0	0	3.3s
✅ MARKDOWN	markdownlint	68	0	0	0	1.59s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.69s
✅ PYTHON	black	1810	56	0	0	70.7s
✅ PYTHON	isort	1810	89	0	0	2.19s
✅ REPOSITORY	betterleaks	yes		no	no	1.65s
✅ REPOSITORY	checkov	yes		no	no	43.83s
✅ REPOSITORY	gitleaks	yes		no	no	13.12s
✅ REPOSITORY	git_diff	yes		no	no	0.31s
✅ REPOSITORY	secretlint	yes		no	no	65.53s
✅ REPOSITORY	syft	yes		no	no	4.24s
✅ REPOSITORY	trivy-sbom	yes		no	no	3.43s
✅ REPOSITORY	trufflehog	yes		no	no	26.32s
✅ YAML	prettier	11	0	0	0	0.57s
✅ YAML	v8r	11		0	0	9.42s
✅ YAML	yamllint	11		0	0	0.4s

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 adds masked-array semantics to a JAX batched-solve path, but masked arrays are not part of the normal JAX/PyRecEst array contract. Supporting them here in isolation would create an inconsistent API without broader end-to-end masked-array support.