carefully allow more strings to use innerHTML

[cmcnulty]

.html() uses rnocache to determine both if a snippet should be cached by buildFragment() and also whether or not html() should use append or innerHTMLto insert the snippet. The problem is that buildFragment requires more stringent rules than innerHTML does, and therefore fewer strings use innerHTML than otherwise could, and that hurts performance, particularly by stripping all snippets with an <option anywhere in the string.

I therefore suggest a new regex rule explicitly to determine if innerHTML or append method should be used. Not only will the improve performance, but it also makes the code cleaner, more readable and make more sense.

The new regex only excludes <style and <script for the following documented reasons:

style:  http://dev.jquery.com/ticket/5977

script:  http://poeticcode.wordpress.com/2007/10/03/innerhtml-and-script-tags/

<object and <embed

Have tests added to them, and cause no new failures in IE7, FF3.7 or Chrome 5.

I discussed these changes a while back in this topic:

 http://forum.jquery.com/topic/html-performance-vs-cruft#14737000000784382

[snover]

Resetting milestone to future.

[cmcnulty]

Created a jsperf test to show the performance increase:

 http://jsperf.com/tweaks-to-rnocache

I extend jquery to create a .html2() which is identical to .html() except that rather than using nocache it uses a regex test that it optimized for .html()

[cmcnulty]

 Pull requested

[john]

#7156 is a duplicate of this ticket.

[john]

#7338 is a duplicate of this ticket.

[wookiehangover]

I'm gonna take crack at this.

[cmcnulty]

Glad to hear this has been assigned. I've been pushing for this change for a long time and am very familiar with at least some of the issues involved. Feel free to let me know if you have any questions! I'm not convinced that this little change is going to be a cure-all for the #7341 problems, but it should help a bit in IE 7,8,9.

[john]

Nominating ticket for 1.7 discussion.

[jaubourg]

+0, What about IE6? Need proof.

[ajpiano]

+1, If the perf here is demonstrable in IE, then let's go for it.

[timmywil]

+1, I'm all for perf improvements.

[cmcnulty]

Replying to jaubourg:

+0, What about IE6? Need proof.

IE6 sees near 100% performance increase, passes all unit tests.

[rwaldron]

@cmcnulty can you provide a test case that we can view in IE6?

[cmcnulty]

Replying to rwaldron:

@cmcnulty can you provide a test case that we can view in IE6?

I'd be happy to, but I'm not sure exactly what you're looking for. I've tested the change in  jsperf with IE6, and the  pull request passes all of the unit tests. The nice thing about this change is that existing unit tests already cover pretty much this entire change except for <EMBED which is the only thing the pull request adds a unit test for. (None currently exist for EMBED anywhere in the unit tests) Other than that, there are already tests in place for, for instance, making sure that an option stays selected, making sure the <OBJECT is inserted properly, etc.

[dmethvin]

+1

[john]

+1, If we think that this might work as we expect it to, then let's go for it.

[scott.gonzalez]

+1

[addyosmani]

+1

[rwaldron]

+1

[jzaefferer]

+0

[dmethvin]

Landed for 1.7:  https://github.com/jquery/jquery/commit/d638aa9c6d6394db0c06431560b3d8b83cacc61c