docs: changelog for the #199–#215 fixes

[kevincostner17]

Summary

CHANGELOG.md → [Unreleased] entries for the fixes that closed #199–#215 (#217–#221, #223, #224, #226).

• Changed, for four behaviour changes users will notice:
  • polars>=1.0 floor (Declared polars>=0.20 floor is wrong: the Polars engine needs polars 1.0+ (LazyFrame.collect_schema) #214)
  • Polars reads native float NaN as null (Polars engine treats float NaN as a value: empty rows/columns kept, missing counts wrong, inf not excluded from outlier fences #200)
  • ValueError when a native output format is requested from an engine that can't produce it (Mismatched engine/output_format silently returns the wrong type, even with fallback_policy='error' #205)
  • a zero-column result keeps its row count (DuckDB engine reports dropping all-empty columns but returns them unchanged #201)
• Fixed: one entry per issue, DuckDB engine crashes with 'STDDEV_SAMP is out of range' on float columns containing NaN or inf #199–Missing-pyarrow error says 'Reading Parquet metadata requires pyarrow' for output_format='arrow' #215.

Rebased onto #225, whose numpy-floor entry is kept.

[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: 58c9afbb-09de-494f-9314-4e9bc9a30ce6

---

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.}