fix: report unmeasured quality-debt dimensions as not assessed; list columns aren't mixed types

[kevincostner17]

Summary

Two public scores reported results for things they never measured, or misread the data:

• fd.evaluate_quality_debt gave a dimension a clean 0.0 when it couldn't be computed, so the gate could say pass without having checked it.
• compute_trust_score flagged object columns of lists or dicts as "mixed types", which lowered consistency, while the equivalent nested Arrow column scored 100.

Root cause

• quality.py: several fallbacks returned (0.0, detail) instead of an unmeasured score:

  • type_instability when profiling raised
  • pii_risk when the PII scan raised or wasn't available
  • schema_drift and category_churn when no baseline was given
  • any dimension missing from the scores

  Each of these became an assessed item that counted toward the total and was written to the ledger.

• enterprise/metrics.py: infer_dtype returns "mixed" for any object column of containers, even when every value is a list. A nested Arrow list infers as "unknown-array", so the same data scored differently depending on how it was stored.

Behaviour change

• Unmeasured quality-debt dimensions now use the DebtItem.assessed mechanism from fix: report undetectable duplicates and uniqueness as unknown instead of clean #414. This covers type_instability on profiling failure, pii_risk when the scan fails or is unavailable, schema_drift and category_churn without baseline=, and any dimension that wasn't scored. For each such item:
  • score and over_threshold serialise as None, and the detail says why (exception type only, no message).
  • It's listed in gate.unassessed, printed as ? <dimension>: not assessed — <why>, and shown as n/a in HTML.
  • It counts toward neither the total, the gate status nor the ledger, so escalation keeps comparing against the last run that measured it.
• Runs with a baseline and a working profile and PII scan are byte-identical to before.
• Runs without a baseline now report schema_drift and category_churn as not assessed instead of 0.0. baseline= is documented as what enables those dimensions. A run without a baseline also no longer writes a 0.0 that reset drift escalation history.
• Object columns whose non-null values are all lists, all dicts or all tuples are no longer "mixed types", and score like the nested Arrow equivalent (consistency 66.7 → 100 for the repro frame). Genuinely mixed columns are still flagged: strings with numbers, lists with scalars, lists with dicts or tuples. Plain-frame trust scores are unchanged.

Tests

• Failures: build_profile and detect_pii patched to raise, with several exception types including ImportError. The affected dimension is unassessed, and the other eight items match an unpatched run.
• Baseline: without one, drift and churn are unassessed; with one, they're measured. A dimension missing from the scores is unassessed.
• Ledger: a run without a baseline between two drifting runs records no drift or churn rows and keeps previous, and the third run escalates to fail.
• Parity: exact to_dict() items, total, status and summary() for a fully measured dirty frame with a baseline.
• Trust score: uniform lists, empty lists, dicts and tuples are consistent. An object list column matches the Arrow list<int64> column. Five genuinely mixed layouts are still flagged.
• Existing tests: the fix: report undetectable duplicates and uniqueness as unknown instead of clean #414 tests now pass baseline=, so they still assert that duplicates is the only unassessed dimension.

Verification

• ruff and mypy clean.
• pytest -m "not online and not large": py3.12 / pandas 2.3.3, 6274 passed; py3.9 / pandas 1.5.3, 6245 passed.
• tests/truthbench: 290 passed.
• Normal-frame outputs (to_dict, summary, to_frame, ledger rows, trust scores) diffed against main on both pandas versions: identical.

[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: 27606f3b-658e-46ac-9912-00a138105c15

---

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.}

[github-actions]

FreshData benchmark report — performance

• freshdata: ?
• python: ?
• platform: ?

fixture	n_rows	n_cols	p50 s	p95 s	peak MB	repair %	false-repair %	preserve %	trust	monotonic	export %

Authored-code reduction (Metric 6)