Support optional Parquet page index writes #3847

Feature Request / Improvement

PyIceberg doesn't expose Parquet page index writing, even though PyArrow supports it with ParquetWriter(write_page_index=True).

Use case / motivation

Page indexes let readers such as ClickHouse skip non-matching pages during predicate evaluation, avoiding decoding work for selective queries. Tables written by PyIceberg currently lack these indexes, so page-level pruning is not possible regardless of reader support.

Proposed change

Add an opt-in write.parquet.page-index-enabled table property, default false, to preserve current output and file sizes. When enabled, PyIceberg passes write_page_index=True to PyArrow's ParquetWriter.

Implementation

PR #3829 adds the table property, threads it through _get_parquet_writer_kwargs, documents it in mkdocs/docs/configuration.md, and includes unit and integration test coverage.

Tooling note

I developed this with assistance from DS v4 Pro and reviewed the changes myself.

References

• Add optional Parquet page index writes #3829