Skip to content

Declare the dependency between element type and corpus IDs in PostgreSQL stats

Erwan Rouchet requested to merge optimize-type-filter into master

Closes #1309 (closed) again

See https://gitlab.com/teklia/arkindex/backend/-/issues/1309#note_1401556847 for the whole context.

TL;DR: This tells the query planner that if you filter by type, then filtering by corpus will not remove any rows because all elements of the same type are in the same corpus. This makes the query planner take much better decisions and always favor the very fast GIN index when it needs to join on ElementPath.

Edited by Erwan Rouchet

Merge request reports