Freegle / Iznik / 24124

72%

Build Type

push

circleci

Committed by invalid-email-address

Commit Message

perf(rippling-stats): restore index use on the reply/mean/taken KPIs

The first-ripple anchor put COALESCE(rip.first_ripple, mg.arrival) in the WHERE,
which made the messages_groups arrival filter non-sargable — a ~4M-row scan (~20s).
Filter on mg.arrival directly (indexed) instead; the day/window/cohort still key on
COALESCE(first_ripple, appearance), so dating/classification are unchanged. Verified
0 rippled posts are dropped by the appearance-window filter (a post's appearance is
always within the window that contains its first ripple).

Also bound the first-ripple derived table by arrival (indexed) — messages_groups has
no index on rippled_in, so an unbounded MIN-per-msgid scanned ~8M rows (~15s); all
rippled rows are recent and any <=90-day window is covered, so this is ~0.7s. Net:
~20s -> ~5s.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014aXkxBPEiPSJBaRMjz7opk

Coverage Stats

11883 of 15790 branches covered (75.26%)

Branch coverage included in aggregate %.

6 of 6 new or added lines in 1 file covered. (100.0%)

18 existing lines in 3 files now uncovered.

128102 of 179745 relevant lines covered (71.27%)

36.98 hits per line