jQuery Forum

Debugged it a bit. IE seemed to hit duplicate check clause

if ( ret[r] === ret[n] {}

all the time, while Firefox did not (at all).

So in, jquery-1.4.1.js Line 3621, changed outer loop  start from "length" to 0

1.         for ( var i = 0, l = this.length; i < l; i++ ) {
               length = ret.length;
               jQuery.find( selector, this[i], ret );
   
               if ( i > 0 ) {
                   // Make sure that the results are unique
                   for ( var n = 0; n < ret.length; n++ ) {
                       for ( var r = 0; r < length; r++ ) {
                           if ( ret[r] === ret[n] ) {
                               ret.splice(n--, 1);
                               break;
                           }
                       }
                   }
               }
           }

that seems to fix my IE8 slow script problem. I don't claim to understand anything about jQuery internals so adverse effects are unknown. All seems to work well, and without any speed penalty for having a wider loop.

Of course it still seems like UI incompatibility, and real problem solution is elsewhere.

Thanks John,

I tried to create a demo snippet, but no luck. The app I work on is under NDA so can't share that. But, after some hours of debugging I found a simple clause that made a difference.

I have a function which resizes images (on 400ms timeout), and each time adds an absolutely positioned div (.fullsize) with a href and icon, .before the IMG. When page is resized, that icon div is erased and then redrawn. When I have

jQuery(window).resize(function() {
        jQuery("div").remove(".fullsize");
        // resize function follows...
});

IE8 goes very unresponsive on window resize, and starts to give "script is slow, abort?" messages. There is something in my DOM that hates that row. The "fix" in my first message fixes this.

When I write removal as

jQuery(window).resize(function() {
        jQuery("div.fullsize").remove();
        // resize function follows...
});

I8 is flawless with stock 1.4.1 and UI 1.8rc1.

The oddest thing is that when I changed above line, it made those "slow script" errors go away also for other functions outside window.resize() when window was NOT resized at all (only AJAX done and div filled with result, height checked and div height resized to conform UI). That one line change fixed IE8.

Nightly build did not help.

But, Scott's suggestion made the problem go away. I removed from UI lines

1. var _remove = $.fn.remove;
   
   $.fn.remove = function( selector, keepData ) {
       if ( !keepData ) {
           $( "*", this ).add( this ).each(function() {
               $( this ).triggerHandler( "remove" );
           });
       }
       return _remove.apply( this, arguments );
   };

And that fixed the IE8 slowness messages.

I added console.log to

1. if ( i > 0 ) {
                   // Make sure that the results are unique
                   console.log(i);
                   for ( var n = length; n < ret.length; n++ ) {
                       for ( var r = 0; r < length; r++ ) {
                           if ( ret[r] === ret[n] ) {
                               ret.splice(n--, 1);
                               break;
                           }
                       }
                   }
               }

With unmodied jQuery and JQuery UI and code
jQuery("div").remove(".fullsize");

Result was that when page was loaded with IE 8

LOG: 1
LOG: 2
LOG: 3
(and up to....)
LOG: 29

When page was resized

LOG: 1
LOG: 2
LOG: 3
LOG: 4
LOG: 5
LOG: 6
LOG: 7
LOG: 8
LOG: 9
LOG: 10
LOG: 11
...
LOG: 306
LOG: 307
LOG: 308
LOG: 309
  Operation aborted ( == I got "script slow message and aborted execution)
  jquery-nightly.js, line 3700 character 50
LOG: 1
LOG: 2
LOG: 3
LOG: 4
LOG: 5
LOG: 6
LOG: 7
LOG: 8
LOG: 9
LOG: 10
LOG: 11
...
LOG: 301
LOG: 302
LOG: 303
LOG: 304
LOG: 305
  Operation aborted  ( == I got "script slow message and aborted execution)
  jquery-nightly.js, line 3700 character 50



Firefox gets with on page load

LOG: 1
LOG: 2
LOG: 3
(and up to....)
LOG: 29

and when page is resized

LOG: 1
LOG: 2
LOG: 3
(and up to....)
LOG: 321


With modified JQuery UI (.remove removed) and code
jQuery("div").remove(".fullsize");

IE8:
No messages in console.log at all. It never gets i > 0

Firefox:
No messages in console.log at all. It never gets i > 0

-----

With unmodied jQuery and JQuery UI and code
jQuery("div.fullsize").remove();

Result was that when page was loaded with IE 8

LOG: 1
LOG: 2
LOG: 3
(and up to....)
LOG: 29


and when page is resized

-- no messages --


Firefox gets with on page load

LOG: 1
LOG: 2
LOG: 3
(and up to....)
LOG: 29

and when page is resized

-- no messages --

Hopefully this is helpful.

Scott,

Thanks and Kudos to you, it helps!

With your new remove function, the IE8 slowness error messages are gone, but there is a noticeable speed penalty with 
jQuery("div").remove(".fullsize");

It's something like 1/3rd speed of

jQuery('.fullsize').remove("div"); and
jQuery("div.fullsize").remove();

Nevertheless, most important is that there are no error message for end user.

Is this any more performant?

To me it seems both have exactly same performance. But of course there might be 50ms difference that plain eye can't see.