Remove NOT IN unstitched guard from facets no-filter path

The NOT IN guard against null final_entity rows was preventing the
query planner from using parallel workers and (with a future covering
index) index-only scans. Without the covering index it caused a 2.6x
regression on the no-filter path (7s -> 18.4s).

The filtered path already excludes unstitched entities via the
whereNotNull('final_entities.final_entity') in the inner
entityIdSubquery, so no guard is needed there.

The no-filter path now matches 1.49.x behavior. A followup migration
adding a covering index can re-introduce the guard efficiently.

Signed-off-by: Fredrik Adelöw <freben@spotify.com>
Made-with: Cursor

plugins/catalog-backend/src/service/DefaultEntitiesCatalog.test.ts

@@ -2452,7 +2452,7 @@ describe('DefaultEntitiesCatalog', () => {
     }
 
     it.each(databases.eachSupportedId())(
-      'excludes not-yet-stitched entities, %p',
+      'excludes not-yet-stitched entities from filtered facets, %p',
       async databaseId => {
         await createDatabase(databaseId);
 
@@ -2500,19 +2500,8 @@ describe('DefaultEntitiesCatalog', () => {
           stitcher,
         });
 
-        // Without filters: unstitched entity should be excluded
-        await expect(
-          catalog.facets({
-            facets: ['metadata.name'],
-            credentials: mockCredentials.none(),
-          }),
-        ).resolves.toEqual({
-          facets: {
-            'metadata.name': [{ value: 'stitched', count: 1 }],
-          },
-        });
-
-        // With filter: unstitched entity should also be excluded
+        // With filter: unstitched entity should be excluded because the
+        // inner entityIdSubquery requires final_entity IS NOT NULL
         await expect(
           catalog.facets({
             facets: ['metadata.name'],

plugins/catalog-backend/src/service/DefaultEntitiesCatalog.ts

@@ -676,20 +676,12 @@ export class DefaultEntitiesCatalog implements EntitiesCatalog {
   }
 
   async facets(request: EntityFacetsRequest): Promise<EntityFacetsResponse> {
-    // Exclude not-yet-stitched (or future tombstoned) entities by
-    // anti-joining the tiny set of null final_entity rows. This is
-    // nearly free (~3k rows) compared to confirming ~520k non-null rows.
-    const unstitchedSubquery = this.database('final_entities')
-      .select('final_entities.entity_id')
-      .whereNull('final_entities.final_entity');
-
     const query = this.database<DbSearchRow>('search')
       .whereIn(
         'search.key',
         request.facets.map(f => f.toLocaleLowerCase('en-US')),
       )
       .whereNotNull('search.original_value')
-      .whereNotIn('search.entity_id', unstitchedSubquery)
       .select({
         facet: 'search.key',
         value: 'search.original_value',
@@ -701,8 +693,8 @@ export class DefaultEntitiesCatalog implements EntitiesCatalog {
       // Build a subquery that finds matching entity IDs via
       // final_entities, so that the EXISTS-based filters correlate
       // against one-row-per-entity rather than the much larger search
-      // table. This keeps the facets aggregation fast even with many
-      // filter clauses or permission conditions.
+      // table. The whereNotNull guard on final_entity excludes
+      // not-yet-stitched (or future tombstoned) entities.
       const entityIdSubquery = this.database('final_entities')
         .select('final_entities.entity_id')
         .whereNotNull('final_entities.final_entity');