Kikobeats / superlock / 39996650921

100%

push

github

perf(createLock): reuse one resolved promise on a free acquire (#37)

* test: fail when src coverage drops below 100%

c8 already reported full coverage for src. npm test did not fail when a line went untested. An unused function now makes the suite exit 1.

Co-authored-by: Jose Francisco 'Kiko' Verdú Gambín <josefrancisco.verdu@gmail.com>

* perf(createLock): reuse one resolved promise when a slot is free

A free slot allocated a promise on every acquire. It now returns one promise created with the lock.

Median of 7 runs and 1e6 acquires, 0 errors: 135.09 ms before, 96.16 ms after. The 1000-waiter drain stayed within noise, 56.37 ms before and 56.73 ms after.

Co-authored-by: Jose Francisco 'Kiko' Verdú Gambín <josefrancisco.verdu@gmail.com>

* bench: time acquire and release, and count errors

The old mutex script timed fibonacci in worker threads, each with its own lock. The fair script printed one ratio and a budget it never checked.

node benchmark/index.js now times a free acquire, an async withLock, and a queue drain. It prints the median, the range, the work count, and the error count.

Co-authored-by: Jose Francisco 'Kiko' Verdú Gambín <josefrancisco.verdu@gmail.com>

* docs: record the clean timing and the open double-release

The earlier log row timed a loop that caught errors on every acquire. The comparison to keep is the loop without that catch.

Co-authored-by: Jose Francisco 'Kiko' Verdú Gambín <josefrancisco.verdu@gmail.com>

---------

Co-authored-by: Cursor Agent <cursoragent@cursor.com>

117 of 117 branches covered (100.0%)

Branch coverage included in aggregate %.

12 of 12 new or added lines in 1 file covered. (100.0%)

310 of 310 relevant lines covered (100.0%)

3437.71 hits per line