This is a brief revision to my previous snippet. It is still what it was: a small, simple system for using XMLHttpRequest without using globals.

I'll put the API spec in the code this time, so as not to clutter the front page further.

Changes:

Changed name to HTTPRequest (far more accurate)
Added HEAD requests
Shortened code by making a GenericRequest function
Used a better method for getting an XMLHttpRequest object (checks multiple MS versions and implements the latest one)

Flexible sorting algorithm based on Quicksort with extra functionality, such as:
- Direction (ie: ascending or descending)
- Sort-by-path (eg: item.name, item.name.firstName or item[5])
- Sorting function (returns true if two items are already sorted)
- Type checking
- All constants and support functions are members of the Sort() function
- Testsuite with hooks for cscript and in-browser javascript, so you can tweak and optimize, and make sure it still works
- Environment agnostic (can use with, say, SpiderMonkey or .Net's jsc)

libNog is a 9k library to provide somewhat easier browser-independence. It is not yet complete, and will be continuously revised. It grew from this snippet, and is still growing as I find things I use often.
Presently, libNog features:

A way to browser-independently get an XMLHttpRequest object.
A set of simple methods for GET, POST, HEAD, and to include other Scripts.
Simple methods for className control
A completely abstracted form of getElement*
Browser-independent methods for attaching and detaching event handlers
An event-normalization routine
A simple one-command 'Stop the Event' routine
And the big one: A simplified Class object by which classes can be built in JavaScript.

The above link will point you to the API documentation and history of libNog, as well as where to download the library.

This is a function I created to ease headaches caused by creating and inserting new stuff in an HTML document dynamically. Essentially, it allows you to specify an HTML tag, in its entirety (attributes, styles, event handling, etc), via a single command. You can nest it, so as to create an entire element tree programmatically in a lot less time as with document.createElement.

Don't let the goofy function acronym fool you; this is damned powerful.

First, the function spec:
HTMLElement acne(type[,attributes[,styles[,events[,children[,parent[,document]]]]]]);

Arguments:

String type: type of element to create; same information as the argument of document.createElement
mixed Object attributes: attributes you want the new element to have (ie: href, src)
mixed Object styles: styles for the new element (ie: width, borderLeft)
(function|function Array) Object events: event handlers for the new element
HTMLElement|HTMLElement Array children: elements you want to be in the new element (ie: nesting is possible)
HTMLElement parent: parent node of object
HTMLDocument document: document to create element in

Return value: The newly created/configured element

This is a quick set of overrides for Javascript so that any variable can be passed to PHP in a GET/POST activity. Just call myVar.toPHP();

I use this little set of functions extensively in a little Javascript/PHP RPC handler I wrote. I don't have the reverse function, as I pass JSON back to the browser for the return value.

Note: This lib is not safe for binaries or HTML Elements. The former will come out similar to FTP ASCII breaks, and the latter will cause infinite recursion. If you want to make a speical case for HTML Elements, do so; you could probably just test for parentNode and create a 'safe' object from that. I didn't need it, so I didn't code it.

As for binary safety, at some point between toPHP/escape/post/urldecode/unserialize, the object breaks. Rather than create a huge fix for something I didn't need to do, I put in a quick match/hack. Don't like it? Write the fix yourself.

Simple set of functions for capturing and manipulating the className member of an HTMLElement

This is the final version of my Javascript serializer targetted at PHP.

The point:

Objects are most easily passed over the network as serialized strings. Between serialization and unserialization, serialization is by far the easier of the two. Since object passing can sometimes be a process-hungry thing, we want to do things as quickly as possible.

My solution is to always do the hard part in compiled code, while doing the easy part in script. That is, whichever way you're passing an Object, you want to pass it in a natively decoded format for the target.

Since I work mostly in PHP, this meant writing a module that would be able to generate a string that can be decoded with PHP's unserialize() function into a PHP Associative Array (or other applicable type).

Notes:

This lib REQUIRES the Prototype lib. You can hack prototype out of it, of course (by replacing the references to Object.extend() with explicit assignments), but I can't imagine why you'd want to bother; it's used mostly with Ajax.Request anyway.

Previous versions of this code would add the .toPHP() member to the Object prototype. After trying to enumerate things, I found that this is a REALLY bad thing to do, as toPHP springs up where it's not wanted in ALL objects. As a result, I've opted to go the Prototype route and apply it as a member of the Object object.

Please note that if you pass a serialized string to PHP via GET or POST, you'll need to stripslashes() before unserialization.

Javascript sample of use:

var myObject = {
name:'value',
test:['Array','of','strings'],
bool:false,
timestamp: new Date(),
float: 3.1415926539,
number: 42,
func: function () {
alert('Member functions are always omitted from serialization');
}
}
alert(Object.toPHP(myObject));

Output:

a:7:{s:4:"name";s:5:"value";s:4:"test";a:3:{i:0;s:5:"Array";i:1;s:2:"of";i:2;s:7:"strings";}s:4:"bool";b:0;s:9:"timestamp";i:1190897619824;s:5:"float";d:3.1415926539;s:6:"number";i:42;s:4:"func";null}

Sample of subsequent unserialization in PHP (passed via POST as 'myobject')

$myObject=unserialize(stripslashes($_POST['myobject']));
var_dump($myObject);

Output:

array(7) {
["name"]=>
string(5) "value"
["test"]=>
array(3) {
[0]=>
string(5) "Array"
[1]=>
string(2) "of"
[2]=>
string(7) "strings"
}
["bool"]=>
bool(false)
["timestamp"]=>
int(1192296601)
["float"]=>
float(3.1415926539)
["number"]=>
int(42)
["func"]=>
NULL
}

Here's a quick lil' addon for Prototype that I use often.

Prototype.Promise(condition, action, interval)
condition is a string that you want met before an action is run.
action is a function that does the action.
interval is the polling rate for condition in seconds, and defaults to 1

So, for example, you may want function foo to run, but only once bar has been set:

function foo(a,b) {
this.retVal=a+b;
}
var thingy = {
retVal:0
};
Prototype.Promise(
'thingy.retVal=5',
foo.bind(thingy,5,10),
5
);


Then, in some point in the mysterious future, thingy.retVal gets set to 5, at which point, the Promise goes into effect, and thingy.retVal becomes 10.

Where I find this particularly useful is in making sure that a document is loaded before doing something (condition="$$('body').length>=1), as you can see it used for the include functions.

Speaking of which, the following include functions are great for getting scripts and stylesheets into your page. I won't bother with examples, as they're pretty straightforward.

Meanwhile, Prototype.scriptPath will point to wherever in your server's heirarchy Prototype was loaded from. The regex, you'll note allows for names like prototype.compressed.js, prototype.modified.js, 2007-09-28.prototype.js, etc - just in case you want to keep track of your various hacks of Prototype, as I do.

This is a version of my previous entrant, Promise, that will work with prototype, but does not require it. It's slightly more advanced, using .apply here and there.

It's basically a function to allow any other function to poll. Polling is generally regarded as bad practice in object oriented code, but can make very simple the matter of, for example, running a bit of code only after a single-run event (like onload) occurs (whether that be in the future or past), another unrelated bit of code needs to be hack-tracked, or any other generic condition.

Note that when the function runs, it doesn't necessarily run within the scope that's called it. Its context is set to itself, rather than its normal context, and it's asynchronous, so you'll not get a return value. If you use Prototype, you can bind the function and it'll behave as it should in terms of context, but I'm unaware of a way to cause an asynch function to block execution - and you'd really rather that not happen anyway, trust me.

Ok, so I was getting frustrated with the inability to hide data in javascript, as well as a number of other concerns with existing class structures (the requirement to use this.constructor in lieu of self, for example).

As a result, I've thrown together a nice little class constructor that accepts a class definition, as well as a class extender that allows you to build a new class from an existing one, with full access to all that private stuff.