fix(impute): reject unknown impute_strategy columns; single fallback record for predictor-less MissForest columns

[kevincostner17]

Summary

• Pipeline.impute(columns=) and impute_strategy silently ignore unknown column names #310 — impute_strategy keys (including those produced by Pipeline.impute(columns=)) that match no column were silently ignored: a pre-rename name like "Age" after normalize_columns(), or a typo like "agee", imputed nothing with no error, warning or report action. run_pipeline now validates the keys right after column renaming and raises a ValueError in the same style as duplicate_subset: it lists the unknown keys and available columns and, when a key is a pre-normalization name, suggests the normalized one ('Age' -> 'age'). Validation runs before any drop step, so keys for columns a later step removes are still accepted, and before any mutation of the input.
• MissForest on a single-column frame logs the column as imputed twice and claims a model #324 — MissForest on a frame with no predictor columns (a one-column frame) filled the column via the fallback, then also logged it as a random-forest regressor/classifier imputation and appended it to columns_imputed twice. Such columns are now filled once by the simple fallback before scikit-learn is loaded and removed from the model loop, so the report shows a single missforest_fallback action and scikit-learn is not required when every column falls back.

Behaviour change

Configs with impute_strategy keys that name no column (after renaming) now raise instead of being silently ignored. This includes a reused config across frames where the column is missing from some of them.

Tests

New tests/test_impute_column_validation.py:

• Pipeline.impute(columns=) and impute_strategy silently ignore unknown column names #310: pipeline pre-rename name (error + rename hint), fd.clean typo, default renaming with the original name, unknown key without renaming (no rename note), normalized name imputes, original name imputes when renaming is off, key for a column later dropped by drop_empty_columns is accepted, non-string labels matched by string key, input untouched on error.
• MissForest on a single-column frame logs the column as imputed twice and claims a model #324: single-column MissForest records exactly one fallback action and one columns_imputed entry; same with scikit-learn imports blocked, via both impute="missforest" and impute_strategy={"x": "missforest"}; frames with predictors still fit a regressor exactly once.

Verification

• ruff check . — clean
• mypy src/freshdata — no issues (202 files)
• pytest -m "not online and not large" — py3.12: 4258 passed, 6 skipped; py3.9: 4254 passed, 10 skipped

Closes #310
Closes #324

[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: e3b4ef50-4f6c-409c-a193-b7e6b7ccc09d

---

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)