Apache Iceberg table specification
The canonical description of the metadata graph, sequence numbers, snapshots, optimistic commits, partition specs, schemas, and row-level deletes.
Apache Iceberg · Learning path
How does an Apache Iceberg table work?
A path through Iceberg’s metadata tree, snapshots, manifests, atomic commits, partition evolution, schema evolution, and scan planning.
Engineers who need to explain why Iceberg is a table format, how it avoids directory listing, and what makes concurrent changes safe.
- 2
- stages
- 2
- selected readings
- 01
Stage 1
Read the table format from the metadata down
Trace table metadata to snapshots, manifest lists, manifests, data files, and delete files, then explain atomic replacement of table metadata.
- 02
Stage 2
Understand evolution without rewriting data
Explain how stable field IDs and partition transforms allow schema and partition evolution without rewriting existing files.
Evolution in Apache Iceberg
Concentrates the schema, partition, sort-order, and property changes that Iceberg applies as metadata operations.