paulononato/backstage
1
0
Fork 0
Code Issues Pull Requests Actions 44 Packages Projects Releases Wiki Activity

Add extended statistics on search(key, value) for planner accuracy

Browse Source 
The search table is an EAV schema where key and value are highly
correlated, but standard single-column statistics can't model this.
This caused the planner to underestimate compound filter queries by
up to 1300x, choosing materialize-then-sort plans instead of LIMIT
short-circuit index scans.

CREATE STATISTICS captures the (key, value) correlation via
dependencies, ndistinct, and most-common-values metadata. The planner
then correctly estimates row counts and chooses the index-driven plan.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
This commit is contained in:
Fredrik Adelöw
2026-05-19 23:11:19 +02:00
parent cfb9f07c28
commit 39c5fbb633
2 changed files with 92 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.changeset/search-extended-statistics.md
+5
View File
@@ -0,0 +1,5 @@
---
'@backstage/plugin-catalog-backend': patch
---
Added extended multi-column statistics on `(key, value)` in the `search` table (PostgreSQL only). This tells the query planner about the correlation between the `key` and `value` columns, fixing severe row count estimation errors on compound filter queries. Without this, the planner could choose to materialize and sort thousands of rows instead of using the LIMIT short-circuit index scan — causing 10-40x slower catalog list views when multiple filters are active.