Roadmap

Arrow export and bounded schema validation

Latest public release. It adds to_arrow, bool dtype detection, and max_errors for schema validation.

Large production feature wave

• CSV hardening: on_bad_lines, dtype overrides, skip rows, decimal separators, encoding errors, Unicode paths, duplicate headers, and clearer row diagnostics.
• Frame ergonomics: from_records, schema_summary, describe, to_dict, and stronger column guards.
• Interop: DuckDB registration, Parquet writing, Arrow-facing work, pandas accessor improvements, and scikit-learn safety checks.
• Quality and schema: near-constant warnings, high-cardinality warnings, schema YAML export, custom validators, and validation Markdown/Pandas output.

Polish before the next release

• Frame selection and dropping methods are now more ergonomic.
• Quality reports can exclude sensitive columns and export JSON.
• Benchmark regression checks and dry-run tests improve maintenance confidence.
• Whitespace-only duplicate headers and tuple mapping keys are handled more safely.

Performance parity and streaming depth

• Continue optimizing parser and cleaning paths using benchmark-driven changes.
• Expand chunked processing beyond reading into more streaming-friendly transformation and validation workflows.
• Add clearer benchmark baselines for release comparisons across supported platforms.

Docs, release, and contributor experience

• Keep website docs aligned with public releases and current main.
• Preserve release automation through Release Please, GitHub Actions, wheels, and PyPI Trusted Publishing.
• Maintain label quality, issue scope, GSSoC review flow, and test expectations for merged contributions.