microlinkhq / browserless / 36837637077

81%

push

github

feat(function): name the spans spent reaching the page (#953)

* feat(function): name the spans spent reaching the page

Reaching the page is the runner's work, not the snippet's, so it lands in
`run` with nothing to tell the parts apart. Measured in production on
`({ page }) => page.title()`, timing the snippet from inside itself:

  fn.run=3536   inside=10   (browser=0  pages=2  title=8)
  fn.run=6466   inside=5    (browser=0  pages=0  title=5)
  fn.run=117    inside=5    (browser=0  pages=0  title=5)

The snippet's own work is 5-10ms whether the request is fast or slow, so
the time is spent before its body runs, and `run` being one number is why
the last investigation narrowed by elimination and reached the wrong
step.

`connect`, `pages` and `resolve` are now named spans. `connect` is the
remaining suspect for that tail and this is what turns it into a
measurement rather than an inference.

The spans need `isolated-function` 0.2.11 for `globalThis.__isolated_time`.
An older one is not an error: the helper falls back to calling the thunk,
so nothing is recorded and behaviour is unchanged. The stub path never
reaches a browser and so names nothing, which is asserted.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* chore(deps): isolated-function to 0.2.12 for the span accounting

0.2.11 reports a named span alongside `run` while it is inside it, so the
row double-counts. 0.2.12 takes a timed span back out, which is what
makes these three readable next to the phases they sit between.

Verified against a real browser rather than the test doubles:

  install=0  build=46  spawn=142  connect=10  pages=5  resolve=NaN  run=9  total=200

`resolve` is absent because that path has one page and no `strictTarget`,
so the scan is skipped; a non-finite phase is already dropped downstream.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* feat(function): time the module load and the disconnect too

`connect`, `pages` and ... (continued)

811 of 936 branches covered (86.65%)

Branch coverage included in aggregate %.

16 of 16 new or added lines in 1 file covered. (100.0%)

3072 of 3891 relevant lines covered (78.95%)

92.09 hits per line