Parquet, byte by byte · Build a Parquet reader while you read about one

Part V: Beyond one file

What changes when files are protected, or many?

ch13 reads encrypted files and asks what stays visible. ch14 puts many files behind one table, and shows file pruning working above the row-group pruning of Part III. ch15 changes that table, and measures where Parquet works badly: finding one row, deleting one, and the small files and delete files that changes leave behind until compaction. ch16 maps what Apache Iceberg adds above Parquet files to the chapters that show it, and compares it with Delta Lake and Apache Hudi.

By Chris Snow, in collaboration with Claude (Anthropic) · Prose and figures: CC BY-NC 4.0 · Code: Apache 2.0

Built from commit c32f48e.