Reject masked shared random parameters

[FlorianPfaff]

Summary

• reject genuinely masked distribution parameters before shared NumPy/Autograd random wrappers coerce them with asarray
• cover uniform bounds, normal location/scale, multivariate-normal mean/covariance, and choice probabilities
• preserve ordinary NumPy inputs and fully unmasked MaskedArray values

Root cause

The shared validators converted inputs with asarray before checking NumPy mask semantics. For a masked scalar or partially masked array, that conversion exposes the hidden data payload and discards the mask. Missing configuration values could therefore become real sampling parameters.

Examples include a masked normal scale being silently used as its hidden standard deviation and a masked choice probability being normalized and sampled from.

Impact

NumPy and Autograd backend samplers now fail early instead of drawing reproducible-looking samples from missing data. The earlier masked-input fix for NumPy randint and multinomial did not cover these shared distribution samplers.

Validation

• reproduced mask loss for scalar and partially masked arrays through NumPy asarray
• isolated exact-module regression: eight masked cases rejected and the fully unmasked compatibility path passed
• Python syntax compilation passed for the modified source and regression module
• branch comparison: 2 commits ahead, 0 behind main
• final diff is limited to 8 validator lines and one focused test file

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	28.77s
✅ JSON	prettier	7	0	0	0	1.27s
✅ JSON	v8r	7		0	0	4.52s
✅ MARKDOWN	markdownlint	68	0	0	0	1.97s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.69s
✅ PYTHON	black	1809	55	0	0	99.08s
✅ PYTHON	isort	1809	87	0	0	2.83s
✅ REPOSITORY	betterleaks	yes		no	no	2.03s
✅ REPOSITORY	checkov	yes		no	no	59.07s
✅ REPOSITORY	gitleaks	yes		no	no	17.1s
✅ REPOSITORY	git_diff	yes		no	no	0.3s
✅ REPOSITORY	secretlint	yes		no	no	89.56s
✅ REPOSITORY	syft	yes		no	no	6.2s
✅ REPOSITORY	trivy-sbom	yes		no	no	7.03s
✅ REPOSITORY	trufflehog	yes		no	no	32.86s
✅ YAML	prettier	11	0	0	0	0.76s
✅ YAML	v8r	11		0	0	13.45s
✅ YAML	yamllint	11		0	0	0.72s

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 expands masked-array rejection across shared random-distribution parameters, but masked arrays are not a supported cross-backend sampling contract in PyRecEst. Adding piecemeal validation here would increase inconsistency rather than establish coherent missing-data behavior.