Sizzle Performance Regression

Reported by:	rwaldron	Owned by:
Priority:	blocker	Milestone:	1.5
Component:	selector	Version:	1.4.4
Keywords:		Cc:
Blocked by:		Blocking:

Description

http://www.readwriteweb.com/hack/2011/01/jquery-upgrade-twitter.php

http://www.dustindiaz.com/about-that-slowness-on-twitter/

I just wanted to document these here, close if we have tickets

Attachments (0)

Change History (4)

Changed January 20, 2011 02:34PM UTC by jitter comment:1

_comment0:	That sounds strange to me as 1.4.3 specifically included an update to Sizzle to '''improve''' the performance of element rooted queries. Also the [http://blog.jquery.com/2010/10/16/jquery-143-released/ blog post] on the 1.4.3 release documents [http://www.flickr.com/photos/jeresig/5078513793/ improvements on the performance of .find] → 1295534217423541

That sounds strange to me as 1.4.3 specifically included an update to Sizzle to improve the performance of element rooted queries. Also the blog post on the 1.4.3 release documents improvements on the performance of .find. Maybe we need a jsperf test on this

Changed January 20, 2011 03:25PM UTC by petersendidit comment:2

http://jsperf.com/jquery-context-find-class

Changed January 20, 2011 08:12PM UTC by danheberden comment:3

resolution:	→ fixed
status:	new → closed

Changed January 20, 2011 08:15PM UTC by john comment:4

component:	unfiled → selector
milestone:	1.next → 1.5
priority:	undecided → blocker

So yeah - this was from moving to using qSA in jQuery 1.4.3. qSA(".class") is slower than doing getElementsByClassName("class") (which is what Sizzle would use internally). Amusingly the case of jQuery(".class") would/has been always using qSA if possible so that would've been "slow" for them as well.

I've moved some perf improvements from jQuery into Sizzle and added a ".class" check and quickly use getElementsByClassName if it exists. This will help to make the code faster in all browsers (and for other projects that use Sizzle).

Side navigation