solidjs / solid / 34540201364

72%

main: 89%

push

github

perf(signals): mark heap insertions in place instead of invalidating the markHeap memo (#3356)

* perf(signals): mark heap insertions in place instead of invalidating the markHeap memo

Mounting N rows in one flush was O(N²) when each row created a user effect
whose source was written during row creation (the ref-effect pattern). An
unmarked node entering an already-marked pure heap set heap._marked = false,
so every later mid-tick memo pull re-walked the whole heap: N full walks,
N²/2 nodes visited. The insertion now runs markNode(n) in place — the same
flags the deferred walk would have produced (#2922 semantics unchanged) —
and the mount is linear (8000 rows, one ref effect each, prod: 231 → 12 ms).

Adds a mount-scaling bench and a tripwire + mid-flush freshness test.

Fixes #3350

Co-authored-by: Claude via Cursor <noreply@cursor.com>

* test(signals): make the #3350 mount tripwire relative

The absolute bound (250 ms at 8000 rows) measured 403 ms on the
coverage-instrumented CI job (12× slower than local). Compare 1000 vs 8000
rows within one process instead: ~8-10× fixed, ~50× on next; threshold 24.

Co-authored-by: Claude via Cursor <noreply@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: Claude via Cursor <noreply@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

519 of 790 branches covered (65.7%)

Branch coverage included in aggregate %.

772 of 1007 relevant lines covered (76.66%)

15.02 hits per line