We have a site that is driven pretty much entirely out of knockout and we need to support all major browsers, including Internet Explorer back to IE7 (not IE6).
Chrome already supports almost all of the HTML 5 features that we actually care about, and Modernizr handles the CSS hacks like a champ. But there are times when we still have to resort to polyfills, two notable examples being the placeholder attribute and more recently the <details> element.
Most of the polyfills are or rely on jQuery plugins, which is great in theory. Unfortunately, they also tend to be ineffective at dealing with dynamically loaded content - which there's a ton of when you use knockout (or any templating engine, really). Further complicating matters is that we're using knockout as true MVVM, so there's no decent place to shoehorn in a bunch of JS hacks to reload the plugins (which is probably a good thing as far as our architecture is concerned, but frustrating on this front).
We were able to come up with a semi-reliable implementation using the DOMNodeInserted event (deprecated, I know) for Firefox and IE9. Unfortunately it didn't work in IE8, because legacy IE doesn't support it and it seems damn near impossible to replicate in those browsers. onreadystatechange seemed promising at first but the event tends to fire too early - even if readyState is explicitly checked - and the polyfills miss their targets, so to speak.
The only option we tried that actually worked reliably in IE7/IE8 was using a repeating timeout to rerun the polyfills every 50 ms. Unfortunately, that also consumed an entire CPU constantly, and bumping it up to even 100 ms caused a too-noticeable delay in the UI, so not really suitable for production use.
So: Is there any reliable way of combining traditional polyfill techniques with dynamic content and templating engines like knockoutjs, that works in every major browser down to at least IE7?
(FWIW, we did eventually manage a workaround using knockout's afterRender binding, but that kind of takes the "poly" out of "polyfill". I'm hoping for something that we can write once and then forget about.)
The way I solved the same problem was to wrap most of my jQuery plugins, and behaviors in general, in knockout bindings (http://knockoutjs.com/documentation/custom-bindings.html). So I had, for example, a placeholder "binding" that I used on each input like <input data-bind="placeholder:'Some Placeholder Text'"/> that either simply set the placeholder attribute or did some IE hack depending on the need.
A broader solution would be to augment knockouts binding provider (http://www.knockmeout.net/2011/09/ko-13-preview-part-2-custom-binding.html). The binding provider is the thing that traverses the DOM (both on load and when dynamically loaded) and identifies bindings. By default, that essentially means it's just looking for data-bind attributes and ko comments, but you can change this to also find attributes like placeholder, input types for date or number inputs, etc, and add your IE hacks.
I heard several times that jQuery's strongest asset is the way it queries and manipulates elements in the DOM: you can use CSS queries to create complex queries that would be very hard to do in regular javascript .
However , as far as I know, you can achieve the same result with document.querySelector or document.querySelectorAll, which are supported in Internet Explorer 8 and above.
So the question is this: why 'risk' jQuery's overhead if its strongest asset can be achieved with pure JavaScript?
I know jQuery has more than just CSS selectors, for example cross browser AJAX, nice event attaching etc. But its querying part is a very big part of the strength of jQuery!
Any thoughts?
document.querySelectorAll() has several inconsistencies across browsers and is not supported in older browsersThis probably won't cause any trouble anymore nowadays. It has a very unintuitive scoping mechanism and some other not so nice features. Also with javascript you have a harder time working with the result sets of these queries, which in many cases you might want to do. jQuery provides functions to work on them like: filter(), find(), children(), parent(), map(), not() and several more. Not to mention the jQuery ability to work with pseudo-class selectors.
However, I would not consider these things as jQuery's strongest features but other things like "working" on the dom (events, styling, animation & manipulation) in a crossbrowser compatible way or the ajax interface.
If you only want the selector engine from jQuery you can use the one jQuery itself is using: Sizzle That way you have the power of jQuerys Selector engine without the nasty overhead.
EDIT:
Just for the record, I'm a huge vanilla JavaScript fan. Nonetheless it's a fact that you sometimes need 10 lines of JavaScript where you would write 1 line jQuery.
Of course you have to be disciplined to not write jQuery like this:
$('ul.first').find('.foo').css('background-color', 'red').end().find('.bar').css('background-color', 'green').end();
This is extremely hard to read, while the latter is pretty clear:
$('ul.first')
.find('.foo')
.css('background-color', 'red')
.end()
.find('.bar')
.css('background-color', 'green')
.end();
The equivalent JavaScript would be far more complex illustrated by the pseudocode above:
1) Find the element, consider taking all element or only the first.
// $('ul.first')
// taking querySelectorAll has to be considered
var e = document.querySelector("ul.first");
2) Iterate over the array of child nodes via some (possibly nested or recursive) loops and check the class (classlist not available in all browsers!)
//.find('.foo')
for (var i = 0;i<e.length;i++){
// older browser don't have element.classList -> even more complex
e[i].children.classList.contains('foo');
// do some more magic stuff here
}
3) apply the css style
// .css('background-color', 'green')
// note different notation
element.style.backgroundColor = "green" // or
element.style["background-color"] = "green"
This code would be at least two times as much lines of code you write with jQuery. Also you would have to consider cross-browser issues which will compromise the severe speed advantage (besides from the reliability) of the native code.
If you are optimizing your page for IE8 or newer, you should really consider whether you need jquery or not. Modern browsers have many assets natively which jquery provides.
If you care for performance, you can have incredible performance benefits (2-10 faster) using native javascript:
http://jsperf.com/jquery-vs-native-selector-and-element-style/2
I transformed a div-tagcloud from jquery to native javascript (IE8+ compatible), the results are impressive. 4 times faster with just a little overhead.
Number of lines Execution Time
Jquery version : 340 155ms
Native version : 370 27ms
You Might Not Need Jquery provides a really nice overview, which native methods replace for which browser version.
http://youmightnotneedjquery.com/
Appendix: Further speed comparisons how native methods compete to jquery
Array: $.inArray vs Array.indexOf: Jquery 24% slower
DOM: $.empty vs Node.innerHtml: Jquery 97% slower
DOM: $.on vs Node.addEventListener: Jquery 90% slower
DOM: $.find vs Node.queryselectorall: Jquery 90% slower
Array: $.grep vs Array.filter: Native 70% slower
DOM: $.show/hide vs display none: Jquery 85% slower
AJAX: $.ajax vs XMLHttpRequest: Jquery 89% slower
Height: $.outerHeight vs offsetHeight: Jquery 87% slower
Attr: $.attr vs setAttribute: Jquery 86% slower
To understand why jQuery is so popular, it's important to understand where we're coming from!
About a decade ago, top browsers were IE6, Netscape 8 and Firefox 1.5. Back in those days, there were little cross-browser ways to select an element from the DOM besides Document.getElementById().
So, when jQuery was released back in 2006, it was pretty revolutionary. Back then, jQuery set the standard for how to easily select / change HTML elements and trigger events, because its flexibility and browser support were unprecedented.
Now, more than a decade later, a lot of features that made jQuery so popular have become included in the javaScript standard:
Instead of jQuery's $(), you can now now use Document.querySelectorAll()
Instead of jQuery's $el.on(), you can now use EventTarget.addEventListener()
Instead of jQuery's $el.toggleClass(), you can now use Element.classList.toggle()
...
These weren't generally available back in 2005. The fact that they are today obviously begs the question of why we should use jQuery at all. And indeed, people are increasingly wondering whether we should use jQuery at all.
So, if you think you understand JavaScript well enough to do without jQuery, please do! Don't feel forced to use jQuery, just because so many others are doing it!
That's because jQuery can do much more than querySelectorAll.
First of all, jQuery (and Sizzle, in particular), works for older browsers like IE7-8 that doesn't support CSS2.1-3 selectors.
Plus, Sizzle (which is the selector engine behind jQuery) offers you a lot of more advanced selector instruments, like the :selected pseudo-class, an advanced :not() selector, a more complex syntax like in $("> .children") and so on.
And it does it cross-browsers, flawlessly, offering all that jQuery can offer (plugins and APIs).
Yes, if you think you can rely on simple class and id selectors, jQuery is too much for you, and you'd be paying an exaggerated pay-off. But if you don't, and want to take advantage of all jQuery goodness, then use it.
jQuery's Sizzle selector engine can use querySelectorAll if it's available. It also smooths out inconsistencies between browsers to achieve uniform results. If you don't want to use all of jQuery, you could just use Sizzle separately. This is a pretty fundamental wheel to invent.
Here's some cherry-pickings from the source that show the kind of things jQuery(w/ Sizzle) sorts out for you:
Safari quirks mode:
if ( document.querySelectorAll ) {
(function(){
var oldSizzle = Sizzle,
div = document.createElement("div"),
id = "__sizzle__";
div.innerHTML = "<p class='TEST'></p>";
// Safari can't handle uppercase or unicode characters when
// in quirks mode.
if ( div.querySelectorAll && div.querySelectorAll(".TEST").length === 0 ) {
return;
}
If that guard fails it uses it's a version of Sizzle that isn't enhanced with querySelectorAll. Further down there are specific handles for inconsistencies in IE, Opera, and the Blackberry browser.
// Check parentNode to catch when Blackberry 4.6 returns
// nodes that are no longer in the document #6963
if ( elem && elem.parentNode ) {
// Handle the case where IE and Opera return items
// by name instead of ID
if ( elem.id === match[3] ) {
return makeArray( [ elem ], extra );
}
} else {
return makeArray( [], extra );
}
And if all else fails it will return the result of oldSizzle(query, context, extra, seed).
In terms of code maintainability, there are several reasons to stick with a widely used library.
One of the main ones is that they are well documented, and have communities such as ... say ... stackexchange, where help with the libraries can be found. With a custom coded library, you have the source code, and maybe a how-to document, unless the coder(s) spent more time documenting the code than writing it, which is vanishingly rare.
Writing your own library might work for you , but the intern sitting at the next desk may have an easier time getting up to speed with something like jQuery.
Call it network effect if you like. This isn't to say that the code will be superior in jQuery; just that the concise nature of the code makes it easier to grasp the overall structure for programmers of all skill levels, if only because there's more functional code visible at once in the file you are viewing. In this sense, 5 lines of code is better than 10.
To sum up, I see the main benefits of jQuery as being concise code, and ubiquity.
Here's a comparison if I want to apply the same attribute, e.g. hide all elements of class "my-class". This is one reason to use jQuery.
jQuery:
$('.my-class').hide();
JavaScript:
var cls = document.querySelectorAll('.my-class');
for (var i = 0; i < cls.length; i++) {
cls[i].style.display = 'none';
}
With jQuery already so popular, they should have made document.querySelector() behave just like $(). Instead, document.querySelector() only selects the first matching element which makes it only halfway useful.
Old question, but half a decade later, it’s worth revisiting. Here I am only discussing the selector aspect of jQuery.
document.querySelector[All] is supported by all current browsers, down to IE8, so compatibility is no longer an issue. I have also found no performance issues to speak of (it was supposed to be slower than document.getElementById, but my own testing suggests that it’s slightly faster).
Therefore when it comes to manipulating an element directly, it is to be preferred over jQuery.
For example:
var element=document.querySelector('h1');
element.innerHTML='Hello';
is vastly superior to:
var $element=$('h1');
$element.html('hello');
In order to do anything at all, jQuery has to run through a hundred lines of code (I once traced through code such as the above to see what jQuery was actually doing with it). This is clearly a waste of everyone’s time.
The other significant cost of jQuery is the fact that it wraps everything inside a new jQuery object. This overhead is particularly wasteful if you need to unwrap the object again or to use one of the object methods to deal with properties which are already exposed on the original element.
Where jQuery has an advantage, however, is in how it handles collections. If the requirement is to set properties of multiple elements, jQuery has a built-in each method which allows something like this:
var $elements=$('h2'); // multiple elements
$elements.html('hello');
To do so with Vanilla JavaScript would require something like this:
var elements=document.querySelectorAll('h2');
elements.forEach(function(e) {
e.innerHTML='Hello';
});
which some find daunting.
jQuery selectors are also slightly different, but modern browsers (excluding IE8) won’t get much benefit.
As a rule, I caution against using jQuery for new projects:
jQuery is an external library adding to the overhead of the project, and to your dependency on third parties.
jQuery function is very expensive, processing-wise.
jQuery imposes a methodology which needs to be learned and may compete with other aspects of your code.
jQuery is slow to expose new features in JavaScript.
If none of the above matters, then do what you will. However, jQuery is no longer as important to cross-platform development as it used to be, as modern JavaScript and CSS go a lot further than they used to.
This makes no mention of other features of jQuery. However, I think that they, too, need a closer look.
as the official site says:
"jQuery: The Write Less, Do More, JavaScript Library"
try to translate the following jQuery code without any library
$("p.neat").addClass("ohmy").show("slow");
I think the true answer is that jQuery was developed long before querySelector/querySelectorAll became available in all major browsers.
Initial release of jQuery was in 2006. In fact, even jQuery was not the first which implemented CSS selectors.
IE was the last browser to implement querySelector/querySelectorAll. Its 8th version was released in 2009.
So now, DOM elements selectors is not the strongest point of jQuery anymore. However, it still has a lot of goodies up its sleeve, like shortcuts to change element's css and html content, animations, events binding, ajax.
$("#id") vs document.querySelectorAll("#id")
The deal is with the $() function it makes an array and then breaks it up for you but with document.querySelectorAll() it makes an array and you have to break it up.
Just a comment on this, when using material design lite, jquery selector does not return the property for material design for some reason.
For:
<div class="logonfield mdl-textfield mdl-js-textfield mdl-textfield--floating-label">
<input class="mdl-textfield__input" type="text" id="myinputfield" required>
<label class="mdl-textfield__label" for="myinputfield">Enter something..</label>
</div>
This works:
document.querySelector('#myinputfield').parentNode.MaterialTextfield.change();
This does not:
$('#myinputfield').parentNode.MaterialTextfield.change();
Which is faster and why? Selecting div (for plugin needs) by $('div[data-something]') or $('div.something')? I lean towards the former since it's "cleaner".
Based on this SO question I know I shouldn't be using both. However I didn't find out whether there is a difference between these.
In Chrome 16 at least, there is no difference. However, if you make the class selector less specific ($(".test") for example), it does outperform the other methods:
That was somewhat unexpected, because as ShankarSangoli mentions, I thought the div.test class selector would be faster.
It will vary by browser. Nearly all browsers now support querySelectorAll, and jQuery will use it when it can. querySelectorAll can be used with attribute presence selectors, so if it's there jQuery doesn't have to do the work, it can offload it to the engine.
For older browsers without querySelectorAll, jQuery will obviously have to do more work, but even IE8 has it.
As with most of these things, your best bet is:
Don't worry about it until/unless you see a problem, and
If you see a problem, profile it on the browsers you intend to support and then make an informed decision.
Selecting by class is always faster than attribute selector because jQuery tries to use the native getElementByCalssName first if supported by browsers. If not it uses the querySelector which uses css selectors to find the elements within the page.
I am using IE9.js from this project.
For the most part it works well, enabling support for css such as opacity in Internet Explorer 7. However it doesn't support the :nth-child() selector. Is there a way around this purely using javascript? By that I mean my CSS would be read by the javascript and the css attributes applied to the matching elements.
I am aware that this problem can be solved by using classes on the elements or using jQuery (or other frameworks) to select the elements. However this is not ideal.
Ideally I want to have my presentation code in one place (css) not scattered around my project.
Edit Looks like this IE library might do it: link.
For non-IE6 browsers, you can use the adjacent sibling selector, if the type of child elements are always the same. For example, if you want select the 5th child li element in a list:
ul:first-child > li+li+li+li+li
{
...
}
There are a lot of ways to accomplish this but the easiest would be to adopt a helper library like jQuery. It supports those selectors across all browsers and also gives you a lot more tools.
If you can't add the dependency, just choose a similar library and extract that particular implementation (make sure the license allows it).
Sorry but I didn't see your reference to jQuery in the post... Please disregard the above. http://selectivizr.com/ seems a nice tool, though.
Back in 2005, Quirksmode.com released this article:
http://www.quirksmode.org/dom/classchange.html
that showed "proof" that changing the style of an element by changing its class (ie. "elem.className = x") was almost twice as fast as changing its style via its style property (ie. "elem.style.someStyle = x"), except in Opera. As a result of that article, we started using a className-based solution to do things like showing/hiding elements on our site.
The problem is, one of our developers would much rather use jQuery's equivalent methods for this kind of thing (ie. "$(something).hide()"), and I'm having a hard time convincing him that our className-based function is worth using, as I can only find a single article written four years ago.
Does anyone know of any more recent or more comprehensive investigations in to this issue?
Micro-optimization is evil. I think unless you are hiding a seriously large amount of elements at once or something, the difference in milliseconds is unimportant if by some chance that article is still relevant nowadays.
With that in mind, I would go with jQuery's methods as they are battle tested and more concise.
There is a flaw in the benchmark that article uses.
In my personal experience I've never seen a case where updating a className outperforms inline style setting. I have no concrete proof of this (I do vaguely remember an article I'm going to try to dig up), but I have noticed that large clientside apps (for example gmail, or google maps) prefer setting inline styles to classNames, and it was in the context of analysis of these apps that I first heard of the speed increase in doing so.
Note that I am not promoting one over the other: setting the className dynamically goes a long way in terms of maintainability/readability and separating concerns.
While I generally agree with the practice of using classes over style attributes for many reasons. Performance is one but consistency is another. I've often seen people do things like:
function toggle(element) {
element.style.display = element.style.display == 'none' ? 'block' : 'none';
}
(or the jQuery equivalent)
Seems reasonable right? It is until you apply it to table elements that have default display values of, for example, table-cell (not block). The class method:
.hidden { display: none; }
...
function toggle(element) {
$(element).toggleClass("hidden");
}
is much, much better for this and other reasons.
But the jQuery methods like hide() are an exception to this. They handle display setting correctly and give you the animations.
Test it yourself. There's a tester on the page you linked. Test performance on your target browsers to determine the best method to use, performance-wise.
Personally, I would choose readability over performance. Besides, the tides may turn later (if they haven't already**). If it makes sense to use a class (i.e. you use the style for many elements), you might as well use a class. If the CSS is for an animation on an element, use style. For either, prefer jQuery's functions as they are (1) more consistent and (2) more robust and tested.
**For Opera 10, at least, the speed has definitely improved. The tests are 5/12ms locally for Opera 10, 57/88ms for Firefox 3, 14/36ms for Google Chrome, and 125/118ms (!) for IE7. IE7 (perhaps your target browser) has about the same speed for either, with style changing slightly in the lead!
Premature optimization is the root of all evil.
Are you hiding hundreds of elements per second? If not, I'd say there are bigger fish to fry. Worse, a study done in 2005 says nothing about the performance of modern browsers. No IE7 or 8, No Firefox 3 (was 2 even out?), no Chrome.
But if you insist and want your fellow colleagues to follow suit, you should write a jQuery plugin that "hide()"'s using a class instead of a style.
No, but I can tell you that jQuery.hide() can hide an element with a smooth effect. This cannot be done by changing the className.
You are talking about client code, it runs in the user browser so it does not load your server. I ignore the javascript in your client side but I guess is not crunching much CPU.
Your coleague use of jQuery is probably not having a great impact while it results in a more readable code. Therefore I think he does not need to be convinced at all.