BeakGraph HDF5 Files

Architecture of the on-disk format

Dictionary-encoded, columnar RDF quad stores — SPARQL-queryable at rest.

Start: What a BeakGraph file is →

These pages are the original architecture slides as web pages, one page per slide (the deck, BeakGraph-HDF5-Architecture.pptx, remains in the repository’s git history). They describe format version 3 and the five writers of that time. Since then, format v4 added the langDirs dataset for RDF 1.2 base-direction literals, v5 the tripleTerms component store for RDF 1.2 triple terms, and -method 5 (plaid) joined the writers. The file format specification is normative; the changelog has the version history.

---

Table of contents

• What a BeakGraph file is
• File layout
• Dictionary design
• Inside a dictionary section
• GSPO / GPOS indexes
• Rank / select directories
• Bit-packed buffers
• Query path
• Five writers, one format