I had exactly the the same problem but as I was very short on time and needed really good code coverage I decided to write a PHP wrapper around Pygments library.
It's called PHPygmentizator. It's really simple to use. I wrote a very basic manual. As PHP is Web Development language primarily, I subordinated the structure to that fact and made it very easy to implement in almost any kind of website.
It supports configuration files and if that isn't enough and somebody needs to modify stuff in the process it also fires events.
Demo of how it works can be found on basically any post of my blog which contains source code, this one for example.
With default config you can just provide it a string in this format:
Any text here.
[pygments=javascript]
var a = function(ar1, ar2) {
return null;
}
[/pygments]
Any text.
So it highlights code between tags (tags can be customized in configuration file) and leaves the rest untouched.
Additionally I already made a Syntax recognition library (it uses algorithm which would probably be classified as Bayesian probability) which automatically recognizes which language code block is written in and can easily be hooked to one of PHPygmentizator events to provide automatic language recognition. I will probably make it public some time this week since I need to beautify the structure a bit and write some basic documentation. If you supply it with enough "learning" data it recognizes languages amazingly well, I tested even minified javascripts and languages which have similar keywords and structures and it has never made a mistake.
