Parquet column chunk
A Parquet column chunk is a contiguous sequence of encoded pages for one column within one row group.
Layout: a chunk belongs to a row group, not to all rows of the file. File metadata locates the chunks a reader needs; a column's chunks across multiple row groups are distinct. [1]
Pages: a chunk may contain multiple data pages and an optional dictionary page. Encoding, compression and optional page checksums operate below the chunk level; a chunk is not necessarily one page or one checksum. [2] [3]
Coverage: reading one column does not establish integrity of an excluded column. Hesela's experiment verifies selected values but does not measure physical reads or prefetch. [4]
Relations
- related
- parquet, parquet-row-group, parquet-page-checksum
Citations
- Apache Parquet: file layout and metadata
- Apache Parquet: column chunks and page organization
- Apache Parquet: page checksumming
- Hesela: Parquet checksum and read-coverage experiment
Published by Hesela team · Updated
Report a correction · Editorial standards · Reuse guidance
Cite this page
Hesela team. Parquet column chunk. Hesela. 2026-10-17. https://hesela.dev/terms/parquet-column-chunk/