Improve account media query

[abcang]

By narrowing down by statuses.account_id, index_statuses_20180106 can be used and it becomes faster.

before:

SELECT  DISTINCT "statuses"."id" FROM "statuses" INNER JOIN "media_attachments" ON "media_attachments"."status_id" = "statuses"."id" WHERE "media_attachments"."account_id" = xxx AND ("statuses"."visibility" IN (0, 1) OR "statuses"."id" IN (SELECT "mentions"."status_id" FROM "mentions" WHERE "mentions"."account_id" = yyy)) AND (statuses.created_at <= '2019-02-26 09:21:23.846620') ORDER BY "statuses"."id" DESC LIMIT 20

Limit  (cost=3609.37..29375.14 rows=20 width=8)
  ->  Unique  (cost=3609.37..75266716.19 rows=58421 width=8)
        ->  Merge Join  (cost=3609.37..75266570.13 rows=58421 width=8)
              Merge Cond: (statuses.id = media_attachments.status_id)
              ->  Index Scan Backward using statuses_pkey on statuses  (cost=3374.35..74945413.91 rows=121241046 width=8)
                    Filter: ((created_at <= '2019-02-26 09:21:23.84662'::timestamp without time zone) AND ((visibility = ANY ('{0,1}'::integer[])) OR (hashed SubPlan 1)))
                    SubPlan 1
                      ->  Index Only Scan using index_mentions_on_account_id_and_status_id on mentions  (cost=0.56..3370.75 rows=1211 width=8)
                            Index Cond: (account_id = yyy)
              ->  Index Only Scan Backward using index_media_attachments_on_account_id_and_status_id on media_attachments  (cost=0.56..17317.88 rows=60723 width=8)
                    Index Cond: (account_id = xxx)

after:

SELECT  DISTINCT "statuses"."id" FROM "statuses" INNER JOIN "media_attachments" ON "media_attachments"."status_id" = "statuses"."id" WHERE "statuses"."account_id" = xxx AND "media_attachments"."account_id" = xxx AND ("statuses"."visibility" IN (0, 1) OR "statuses"."id" IN (SELECT "mentions"."status_id" FROM "mentions" WHERE "mentions"."account_id" = yyy)) AND (statuses.created_at <= '2019-02-26 09:21:23.846620') ORDER BY "statuses"."id" DESC LIMIT 20

Limit  (cost=3609.37..42304.73 rows=4 width=8)
  ->  Unique  (cost=3609.37..42304.73 rows=4 width=8)
        ->  Merge Join  (cost=3609.37..42304.72 rows=4 width=8)
              Merge Cond: (statuses.id = media_attachments.status_id)
              ->  Index Scan using index_statuses_20180106 on statuses  (cost=3374.35..24816.73 rows=7419 width=8)
                    Index Cond: (account_id = xxx)
                    Filter: ((created_at <= '2019-02-26 09:21:23.84662'::timestamp without time zone) AND ((visibility = ANY ('{0,1}'::integer[])) OR (hashed SubPlan 1)))
                    SubPlan 1
                      ->  Index Only Scan using index_mentions_on_account_id_and_status_id on mentions  (cost=0.56..3370.75 rows=1211 width=8)
                            Index Cond: (account_id = yyy)
              ->  Index Only Scan Backward using index_media_attachments_on_account_id_and_status_id on media_attachments  (cost=0.56..17317.88 rows=60723 width=8)
                    Index Cond: (account_id = xxx)

[ClearlyClaire]

LGTM!

Hm, looking at the query, I'm wondering where does that created_at query come from? I don't know how expensive it is, but it doesn't seem relevent.

[abcang]

Oh, I'm sorry. Please ignore created_at because it is due to extension of Pawoo. It does not affect query improvement.