zone-map · shipping · also min-max index

Zone map (min-max index)

A zone map is a per-region summary of column-value bounds that lets a reader skip regions that cannot satisfy a query predicate.

Related Observatory article: Zone map (min-max index) on hesela.com

A zone map supports exclusion, not membership proof. An overlapping interval remains a candidate even when it contains no matching value. Clustering can make bounds more selective. [1]

Parquet has column-chunk statistics and a separate optional page index; their granularity and reader support must not be conflated. Conservative bounds can include values absent from the data. They must still enclose the relevant values under the specified ordering. [2] [3]

Correctness depends on valid metadata. Hesela's deliberately narrowed footer minimum caused PyArrow 21.0.0 to prune a matching row although all four page-body CRCs remained valid. Wider valid bounds and absent-statistics controls restored it. This is a controlled fixture, not an observed field failure or a cross-reader result. [4]

Relations

related

parquet, parquet-page-checksum, bloom-filter

Citations

1. Zeng et al. An Empirical Evaluation of Columnar Storage Formats. PVLDB 17(2), 148-161, 2023. DOI: 10.14778/3626292.3626298 (accessed 2026-10-17)
2. Apache Parquet format 2.10.0: Statistics bounds and ColumnOrder (accessed 2026-10-17)
3. Apache Parquet: optional page index (accessed 2026-10-17)
4. Hesela: controlled Parquet footer-statistics observations (accessed 2026-10-17)

JSON: /data/terms/zone-map.json