listview.filter / data-filter performanceoptimization

[nsaleh]

As I experienced huge performanceproblems with data-filter on long lists, I optimized it by:

• not allways showing all items on change (and hiding them right after that), but only show those really matching the filter
• if only chars are added, hidden listelements are not touched at all
• introducing optional attribute (data-)filtertext Several patches - Faster data filter / control what is filtered + Load dialog content from other base URL #1418

[toddparker]

Thanks, we'll take a look!

[toddparker]

Thanks again, we're a little behind on reviews but we'll take a look as soon as we can.

[eddiemonge]

Can you format your changes according to http://docs.jquery.com/JQuery_Core_Style_Guidelines please?

[nsaleh]

done, hope i didn't miss anything.
edif: ahhh okay i see now, my xcode doesn't show me the truth ... i'll fix it

[ianeiko]

thanks nsaleh, its a very significant performance boost!

the official version just hangs when testing this on iOS 4 : http://jquerymobile.com/demos/1.0a4.1/#docs/api/../forms/../api/../../docs/lists/lists-performance.html

[hakanson]

I had similar data-filtertext thoughts in pull request 1221 (#1221), but being new to github, my pull request was overload with changes and I only got around to resubmitting one of them.

Also, I got faster show/hide performance based on http://www.learningjquery.com/2010/05/now-you-see-me-showhide-performance.

[nsaleh]

@hakanson great link! I'll include the class-based hiding soon.

[nsaleh]

I added class-based hiding and finally replaced all spaces with tabs

[jblas]

We're going to take the performance part of this patch, but perhaps not this part. I think this is no longer needed.

[jblas]

Ok, I landed the patch minus the jquery.mobile.navigation.js changes:

1387fd4

I'd still like to know if there's a specific reason the patch used toggleClass() with true/false state specified versus just calling the appropriate addClass/removeClass. The latter just seems easier to read/understand.

[jblas]

One other possible issue would be the initial check for jqmData("lastVal") ... it may return undefined in which case the initial comparison may be against a string of "undefined".

• Kin

[toddparker]

Great, thanks Kin!

[toddparker]

Seems to take less than a second to process a keystroke on the 500 items list performance test on an iPhone 4. Pretty snappy.

[eddiemonge]

The toggleClass part looks like it is supposed to be based on the childItems part but if true or false are being passed in directly then there is no point in using toggleClass. Its supposed to work with passing in a boolean variable that determines whether to add or remove. With true or false being passed in, I think performance is actually worse.

[jblas]

I probably should've asked this before I landed this code, but ... why can't we simply compare lastval against val?

if ( val !== lastval) {

Why do we have to do all this replace and and length checking business?

[jblas]

Disregard my previous note (duh) ... I forgot you were trying to distinguish between appended characters which is part of the optimization.