Sunday, April 8, 2012

Why is this Javascript much *slower* than its jQuery equivalent?



I have a HTML list of about 500 items and a "filter" box above it. I started by using jQuery to filter the list when I typed a letter (timing code added later):







$('#filter').keyup( function() {

var jqStart = (new Date).getTime();



var search = $(this).val().toLowerCase();

var $list = $('ul.ablist > li');



$list.each( function() {

if ( $(this).text().toLowerCase().indexOf(search) === -1 )

$(this).hide();

else

$(this).show();

} );



console.log('Time: ' + ((new Date).getTime() - jqStart));

} );







However, there was a couple of seconds delay after typing each letter (particularly the first letter). So I thought it may be slightly quicker if I used plain Javascript (I read recently that jQuery's each function is particularly slow). Here's my JS equivalent:







document.getElementById('filter').addEventListener( 'keyup', function () {

var jsStart = (new Date).getTime();



var search = this.value.toLowerCase();

var list = document.querySelectorAll('ul.ablist > li');

for ( var i = 0; i < list.length; i++ )

{

if ( list[i].innerText.toLowerCase().indexOf(search) === -1 )

list[i].style.display = 'none';

else

list[i].style.display = 'block';

}



console.log('Time: ' + ((new Date).getTime() - jsStart));

}, false );







To my surprise however, the plain Javascript is up to 10 times slower than the jQuery equivalent. The jQuery version takes around 2-3 seconds to filter on each letter, while the Javascript version takes 17+ seconds! I'm using Google Chrome on Ubuntu Linux.





This isn't for anything really important so it doesn't need to be super efficient. But am I doing something really dumb with my Javascript here?



Source: Tips4all

4 comments:

  1. You could try using textContent instead of innerText , I think it should be faster. Also timing the list-generation and loop separately would tell if there is problem in list-generation.

    ReplyDelete
  2. Another best practice for javascript speed is caching the list.length in a variable and calling the variable like:

    l = list.length;
    for (var i=0;i<l;i++):{ code here}


    And maybe timing with jsperf would be better.

    ReplyDelete
  3. Here, I've refactored your code a bit:

    var filter = document.getElementById( 'filter' ),
    ablist = document.querySelector( '.ablist' );

    filter.addEventListener( 'keyup', function () {
    var re, elems, i, len, elem;

    re = RegExp( this.value, 'i' );
    elems = ablist.children;

    for ( i = 0, len = elems.length; i < len; i += 1 ) {
    elem = elems[i];
    elem.style.display =
    elem.textContent.search( re ) > -1 ? 'list-item' : 'none';
    }
    }, false );


    Live demo: http://jsfiddle.net/MVFxn/

    Changes:


    with a regular expression and an i flag, there's no need for toLowerCase,
    if there is only one '.ablist' element on the page, querySelector should be the fastest way to grab it (since it aborts the query once it finds the first such element),
    there's no query for the LI elements since the children property already references them conveniently.


    I'd love to know how this code performs on your page...

    ReplyDelete
  4. I used while instead of for and did some minor improvements. Here is the final code.

    var list = list = document.querySelectorAll('ul.ablist > li');
    document.getElementById('javascriptFilter').addEventListener( 'keyup', function () {
    var jsStart = (new Date).getTime(),
    search = this.value.toLowerCase(),
    i = list.length - 1,
    listItem,
    result;
    while( i >= 0 )
    {
    listItem = list[i];
    if ( listItem.textContent.toLowerCase().indexOf(search) === -1 )
    listItem.style.display = 'none';
    else
    listItem.style.display = 'block';
    i--;
    }

    result = ((new Date).getTime() - jsStart);
    console.log(['Time: ', result, '<br />'].join(''));
    }, false );

    ReplyDelete