pg-dump Reference

Schema extraction with pg_dump reads PostgreSQL system catalogs to generate DDL statements. pg_dump queries pg_class, pg_attribute, pg_type, and pg_constraint tables to identify objects, their definitions, and dependencies, then outputs CREATE statements in dependency order.

pg_dump queries pg_class for tables and indexes, pg_attribute for columns, pg_type for data types, pg_constraint for constraints, pg_proc for functions, pg_trigger for triggers, pg_depend for dependencies, and pg_description for comments to reconstruct the complete schema.

Different PostgreSQL versions (14–17) have variations in system catalog structure and object support. Reference pg_dump's version-aware logic in pg_dump.c and pg_dump_sort.c to conditionally query catalogs and format DDL based on the target PostgreSQL version.

Yes. pg_dump's main implementation files (pg_dump.c, pg_dump.h, pg_backup_archiver.c, common.c) demonstrate correct system catalog queries and DDL formatting. Study these patterns to validate and fix schema extraction logic in tools like pgschema's ir/inspector.go.

pg_dump queries pg_depend to map object relationships and uses pg_dump_sort.c logic to order DDL statements so dependencies are created before dependents, ensuring clean schema recreation without constraint or reference violations.

Schema-only extraction uses pg_dump's schema query patterns to generate DDL for tables, functions, and constraints without data. Full backup includes schema plus row data. Both use the same system catalog queries; schema mode filters out data export steps.