External Table
An external table exposes data through an engine without placing it under that engine’s ordinary managed-table lifecycle. Its metadata identifies a schema, location or table-metadata reference, and the format the reader must understand.
Registration usually avoids copying source rows. The engine still needs permissions for metadata, objects, encryption keys, and any result storage. A table grant may not cover every access path, while direct object access can bypass an engine-only restriction.
In a plain partition-directory table, registered locations determine discovery. Newly written files inside an existing location may become visible independently of catalog refresh. Registration does not establish that a multi-file write is complete.
External tables can also use a transactional table format. State which catalog and commit protocol all readers honor, how snapshots remain available, and who owns cleanup. “External” alone says little about freshness, update support, or query performance.
See External Tables and the Lake/Warehouse Boundary for worked examples.
Reference: Official documentation.
Discover more from Insightful Data Lab
Subscribe to get the latest posts sent to your email.
