Changelog

Added

• DataQualityReport.to_json() for JSON report export.
• profile(exclude_columns=...) and DataQualityReport.to_dict(exclude_columns=...) for privacy-aware quality exports.
• ArFrame.__getitem__ column selection and ArFrame.drop_columns.
• Lightweight benchmark regression checks and dry-run coverage for the pandas accessor.

Fixed

• Reject headers that differ only by whitespace.
• Safely handle tuple mapping keys in replace_values.

Features

• Added Arrow export API with ar.to_arrow(frame) and bool dtype detection.
• Added max_errors support to Schema.validate() and ar.validate().

Bug fixes

• Improved cleaning mapping validation errors by including the received type.

Documentation

• Clarified that chunked schema validation is not a separate streaming schema-validation contract.

Features

• Added URL allowed_schemes, ArFrame.from_records, ArFrame.schema_summary, schema YAML export, and ArFrame.describe.
• Added configurable bad-line handling, CSV dtype support, skip rows, encoding error handling, decimal separators, and DuckDB registration.
• Added write_parquet, to_dict, drop_empty_columns, winsorize_outliers, select_columns, near-constant warnings, and high-cardinality quality warnings.
• Added pipeline context and verbose diagnostics, plus Python interface and extension stub updates.

Bug fixes

• Hardened CSV parsing, malformed rows, permission messages, Unicode paths, duplicate headers, extra fields, and unterminated quoted-field diagnostics.
• Fixed zero-column row-count preservation, duplicate add-column guards, drop-duplicates row-key collisions, JSONL nrows, Windows clean target, and safer cleaning validation paths.

Performance and docs

• Optimized CSV parser allocation paths, integer parsing, string mutation paths, and unmodified-column moves.
• Added sparse-null benchmarks, optimized theme logos, schema validation tutorial docs, and Windows build troubleshooting.