Document marker filter syntax #5493
Conversation
5ccfdca to
315d0a0
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #5493 +/- ##
=======================================
Coverage 86.15% 86.15%
=======================================
Files 307 307
Lines 29637 29637
Branches 7999 7999
=======================================
Hits 25533 25533
Misses 3516 3516
Partials 588 588 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
canova
left a comment
There was a problem hiding this comment.
LGTM, thanks!
One thing I discovered while trying to document this is that a negative match without a key will be treated as a positive match against a substring with a
-in it: the search stringa,-DOMwill match anything with "a" in it or with "-DOM" in it, whereasa,-name:DOMwill match anything with "a" in it except markers with "DOM" in their names. That's fine functionality to have (I see markers with-in their fields, for example), but unexpected to me. I would probably rather have them error out until quoted strings are supported (to allowa,"-DOM"). I'm not sure it's worth opening an issue for, though.
Yeah, this was done this way because of a real use case that we had back then. SP3 profiles had UserTiming markers with names like <test-name>-start and <test-name>-end, and performance people who looked at the sp3 profiles wanted to be able to filter all those by adding -start,-end in the marker filter.
I think this use case has changed now, with the proper duration markers that we have in sp3 profiles. But there could be other use cases like this.
Also if we change this behavior now, the old published urls that include -something won't work. We can handle this with an upgrade, but still it's something to consider.
|
Ah yes, fair, since I ran into exactly these markers in the investigation that spawned this PR. Only in my case, I want to exclude the -start and -end markers, since they're just noise now that the proper interval markers are there (and I was trying out Fouineur and it seemed easiest to filter things down with the profiler). I ended up tacking |
Changes: [Nazım Can Altınova] Fix issues related to the track borders (#5484) [Nazım Can Altınova] Remove the active tab and origins views (#5483) [Julien Wajsberg] Automatically request reviews for dependency updates (#5490) [Julien Wajsberg] Update fetch-mock-jest to @fetch-mock/jest (#5488) [Steve Fink] Document marker filter syntax (#5493) [Nazım Can Altınova] Update all of jest 29.7.0 → 30.0.0 (major) (#5495) [Nazım Can Altınova] Order global tracks by activity and select the most active non-parent process by default (#5491) [Paul Adenot] Allow searching for HTTP response status in marker views (#5504) [Nazım Can Altınova] Expose a `totalMarkerDuration` function in console (#5507) [Nazım Can Altınova] 🔃 Sync: l10n -> main (July 11, 2025) (#5510) And thanks to our localizers: el: Jim Spentzos kab: ZiriSut tr: Grk tr: Selim Şumlu uk: Artem Polivanchuk
One thing I discovered while trying to document this is that a negative match without a key will be treated as a positive match against a substring with a
-in it: the search stringa,-DOMwill match anything with "a" in it or with "-DOM" in it, whereasa,-name:DOMwill match anything with "a" in it except markers with "DOM" in their names. That's fine functionality to have (I see markers with-in their fields, for example), but unexpected to me. I would probably rather have them error out until quoted strings are supported (to allowa,"-DOM"). I'm not sure it's worth opening an issue for, though.