test(validation): one bad cell must trip each rule, and nothing may mutate

[kevincostner17]

Why

Three properties the suite did not assert directly.

1. One-value violations

A rule tested only against a wholly-broken frame can pass while detecting something else entirely; a rule never tested against a clean frame can pass by always failing. Each rule is now asserted both ways.

Rule	one bad cell detected	clean frame stays silent
unique	✅	✅
allowed_values	✅	✅
min_value/max_value	✅	✅
nullable=False	✅	✅
max_missing_ratio	✅	✅
regex	✅	✅
compound_unique	✅	✅
min_rows / max_rows	✅	n/a
missing required column	✅	n/a
strict_columns extras	✅	n/a

All 16 already behave correctly.

2. Non-mutation, by digest

The existing checks use assert_frame_equal, which compares values and dtypes but ignores .attrs and the index name — an in-place change to either would go unnoticed. These hash content + labels + dtypes + attrs + index name.

run_suite, validate_fields, profile, suggest_plan, infer_roles, explain_clean and clean all leave the input byte-identical.

3. Cross-field, including min <= max

The inventory recorded "no explicit min<=max test". The capability does exist, through caller-supplied cross_rules callables — so this was a coverage gap, not a missing feature.

On a frame with one date inversion and one price inversion on the same row: both rules fire, only that row is reported, the action is manual_review rather than an automatic repair, and a row with a missing half of either pair is correctly not reported — absence is not inversion.

4. Remediation integrity

• accepted + quarantined + rejected + needs_review == len(df) — no row silently dropped.
• The original "apple" stays recoverable from the quarantine frame.
• The audit entry carries original, action, applied, classification and the reason "expected numeric in 'amount' but got text value 'apple'; not silently converted".
• The report's claim and the frames agree — row 1 absent from accepted, present in quarantined.

Verification

• 31 tests. Every property already holds on main, so these are regression guards rather than bug reports.
• Full suite py3.12: 6674 passed, 22 skipped, 0 failed, coverage 93.91%.
• ruff check . clean repo-wide.
• No library code changed.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: db96216d-882e-45e3-9b03-16a1acbeeca7

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}