fixes #504, avoids global list of all MutationObservers #505
Conversation
|
@arv mind taking a look? |
| @@ -180,13 +183,11 @@ | |||
| record.oldValue = associatedStrings[uid]; | |||
|
|
|||
| // 8. | |||
| if (!observer.records_.length) { | |||
| scheduleCallback(observer); | |||
| } | |||
| observer.records_.push(record); | |||
There was a problem hiding this comment.
Is this reordering intentional?
|
LGTM Nice work around |
|
Actually. Could we move the call to scheduleCallback out of the loop. No need to do that work more than once. |
Done. Yeah was thinking it probably didn't matter much perf wise (because of the length check, and the isScheduled check inside scheduleCallback), and I liked having one less state variable and slightly better "encapsulation" of the global list (only schedule/notify touch it). But those are very minor things in the grand scheme of things. :) |
|
by the way, @arv do you usually prefer to squash into a single commit before submitting a pull req? |
|
LGTM Squashing is preferred. Less churn. |
|
thanks! will squash the follow up & land :) |
jmesserly
force-pushed
the
observer_leak
branch
from
d8fcfe2
to
7c8b908
Compare
fixes #504, avoids global list of all MutationObservers
They're now added to the global list on-demand, and it is sorted by ID before delivering. Sounds like this is similar to the Blink implementation (https://github.com/Polymer/ShadowDOM/issues/504#issuecomment-55013158)