fix(missforest): keep integer dtypes by rounding imputed values

[kevincostner17]

Summary

impute="missforest" turned nullable integer columns (Int64, Int32, ...) into object dtype holding a mix of Python ints and fractional floats, with no note in the report. MissForest fills from a float64 working frame. Series.where refuses float predictions on a masked integer array, so _assign_success fell back to astype(object). _fallback_fill did the same when the median was fractional.

For numpy and nullable integer columns, both paths now:

• round imputed values half-to-even to Python ints and keep the column's own dtype;
• use exact_int_stat for the fallback median when the column holds values beyond 2**53, so no digits are lost;
• note the rounding in the action rationale and set a new rounded_to_integer metadata flag. The flag is True only when regressor predictions or a median were actually rounded.

Numeric columns that still cannot hold the fill are cast to float64 with a rationale note, never to object. Float, bool, categorical and string targets are unchanged.

Tests

New tests/test_missforest_nullable_int.py (skipped without scikit-learn):

• Int64/Int32/Int16 columns keep their dtype, imputed values are integers, observed values are untouched, and the report note and flag are present
• integer predictions equal the rounded float predictions for the same frame
• float columns keep fractional predictions and have no rounding note
• numpy int64 columns come back unchanged
• the small-frame fallback keeps Int64, rounds a 3.5 median to 4, and records the note
• values beyond 2**53 stay exact in both the fallback and regressor paths

Verification

• ruff check . clean; mypy src/freshdata clean
• pytest -m "not online and not large": Python 3.12 4446 passed / 6 skipped; Python 3.9 4442 passed / 10 skipped (before rebasing onto current main; MissForest suites re-run after the rebase on both)
• Merges cleanly with fix(impute): reject unknown impute_strategy columns; single fallback record for predictor-less MissForest columns #373, which edits other hunks of missforest.py

Closes #263

[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: 5b4cefb6-4fdf-48f4-8ea7-8e4387f59fed

---

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)