fix: batch enqueue_push_intel from O(N) to O(1) DB round-trips
Phase 10 was adding ~1.4s to every push (18 sequential SELECT+INSERT pairs at ~80ms Docker postgres latency). Replace the per-job loop in enqueue_push_intel with a single SELECT … IN to find already-pending types, then session.add_all for the missing ones — two round-trips total regardless of job count.
Also remove autouse=True from _stub_push_background_tasks so tests declare their dependencies explicitly. test_push_enqueues_intel_jobs now opts in via an explicit fixture parameter; enqueue unit tests call the real implementation and verify DB state directly.
sha256:34844ccdc278ce0371354d4b298302848ce48f8399946af20d41dd7a1ef5d8cd
sha
sha256:c10f13af59dea93995b3c8d4410d84b1f42080ffecc98cd6719f07dba3d5dc2b
snapshot
Older
perf(timing): absurd instrumentation across wire_push_presign, w…
sha256:3c7d4bccfd75c58133785eedaf80376da6dc02149518a1c799b2ce21d0dd18af
All commits
Newer
fix: eliminate 850ms cold-import penalty on first push
sha256:02c83bb8bd449e68b9c4cc8d34134a22191f3b32ea21202c500ab908becf5c4b
0 comments
To add a comment, use the Muse CLI:
muse hub commit comment sha256:34844ccdc278ce0371354d4b298302848ce48f8399946af20d41dd7a1ef5d8cd --body "your comment"
No comments yet. Be the first to start the discussion.