Develop Reference

How to utilize a js minimization for a web site of a big organization

I am looking for a JS minimization (maybe CSS as well) tool to use in our website. The site is fairly big and we cant manually minify files individually. We are also planning to use Long term caching for files and need to append like a version number to each file. I am afraid that this is very hard to keep track of when publishing frequently.
I know of tools like YUI Compressor, etc.. is there, but I am not sure how they are used for a big project like I have. Technically, I am looking for a script or an app that can be called after our development is finished to utilize it with the minified versions of files.
What are the common practices big companies use/follow these days ?? Any help is appreciated. I am just not sure what to search for.
Thank you.

I advise you to use a kind of makefile toolchain (there are many, for example ant or maven) to :
concatenate your js files in one file
then minify the resulting files (I use Google Closure Compiler, called with an ant target)
Note that making one file is the most important operation as on modern networks the latency due to the number of requests is much more a burden than the total size. This way you can easily work with dozens or hundreds of js (or css) files and don't hesitate to make a new one as soon as it helps the code source being readable and maintainable.
And this eliminates the need for the (manual or not) management of visible versionning of files for caching reasons.
As said recently in another answer, to help debug, my deployement scripts always make two versions in parallel : one non concatenated/minified and one concatenated/minified. The uncompressed version enables the development/test onsite without any deployement operation.

Minification for Css/Js - right way?

In my project each page has a bunch of dependent Javascript and Css. Whilst developing I just dumped this code right into the page but now I'm looking to clean it up...
it appears that the general approach out there is to package all the Javascript/CSS for an application into two big files that get minimised.
This approach has the benefit that it reduces bandwidth since all the front-end code gets pulled in just once from the server... however, I'm concerned I will be increasing the memory footprint of the application by defining a whole ton of functions for each page that I don't actually need - which is why I had them on a per-page basis to begin with.
is that something anyone else cares about or is there some way to manage this issue?
yes, I have thought of doing conditional function creation since I need to run code conditionally for each page anyway - though that starts to get a bit hackish in my view.
also, is there much cost to defining a whole ton of Css that is never used?

Serving the javascript/CSS in one big hit for the application, allows the browser to cache all it needs for all your pages. If the standard use case for your site is that users will stay and navigate around for a while then this is a good option to use.
If, however, you wish your landing page to load quickly, since there is a chance that the user will navigate away, consider only serving the CSS/javascript required for this page.
In terms of a performance overhead of a large CSS file - there will be none that is noticeable. All modern browsers are highly optimised for applying styles.
As for your javascript - try not to use conditional function creation, conditional namespace creation is acceptable and required, but your functions should be declared only in one place.

The biggest thing you can do for bandwidth is make sure your server is compressing output. Any static document type should be compressed (html, js, css, etc.).
For instance the jQuery Core goes from approx. 90KB to 30KB only because of the compressed output the server is sending to browsers.
If you take into account the compression, then you have to create some mammoth custom JS includes to really need to split-up your JS files.
I really like minifying and obfuscating my code because I can put my documentation right into the un-minified version and then the minification process removes all the comments for the production environment.

One approach would be to have all the shared javascript minified and compressed into one file and served out on each page. Then the page-specific javascript can be compressed/minified to its own files (although I would consider putting any very common page's javascript into the main javascript file).
I've always been in the habit of compressing/minifying all of the CSS into one file, rather than separate files for each page. This is because some of the page-specific files can be very small, and ideally we share as much css across the site as possible.
Like Jasper mentioned the most important thing would be to make sure that your sever is GZIPing the static resources (such as javascript and css).

If you have a lot of javascript code you can take a look on asynchronous loading of js files.
Some large project like ExtJs or Qooxdoo have build in loaders to load only required code, but here is a lot of libs which simplify this, and you can use in your project (e.g. head.js, LAB.js).
Thanks to them you can build application which loads only necessary files, not whole javascript code which in case of big apps can be a heavy stuff for browser.

jQuery file name

This one should be easy, and I think I know the right answer, but here goes.
For compatibility reasons, should I leave the filename of jQuery as "jquery-1.3.2.min.js" or just rename it to jquery.js?
My guess is leave it as is to avoid conflicts in case another app uses a different version of jQuery. If they've renamed it to "jquery.js" and I do the same, I see potential version conflicts.
Am I wrong or way off base?
Jeff

It's a very good idea to have version-numbered JS (and CSS) files, because that lets you configure your web server to use a far-future Expires header on such files without running into caching problems. When the file gets updated, it gets a new version number, so the browser always fetches the new version, not the old cached one.
You should do this on your other JS and CSS files, too. You want this to be automated, not something you manage by hand. Your development work happens on unversioned files, and your versioning system creates versioned copies and works out the details of updating the references to the CSS and JS files in the HTML files to point to the versioned copies. This can be a bit of work, but well worth it when it comes to speeding up your site. It took me about a day to set my system up. The improvement wasn't subtle.

I would go with jquery-1.3.2.min.js because it's more specific and you can immediately tell if you're reviewing this site in months to come, as well as avoiding any filename confliction in the future.
You shouldn't have any issues with updating, if you're relying on something like an include/template file for the javascript.

In my opinion, its just a personal preference. If you have version in your file name, It helps you easily identify which one you are using with out actually opening the file. It also provides an indirect way of clients downloading the new version file (as it is never cached). If you don't use the ext, upgrading to newer version is easy in coding perspective, but takes the pain of force downloading the new file by all users.
Recommended way to use jQuery in app is using the google's hosting..
google.load("jquery", "1.3.2");
google.setOnLoadCallback(function() {
// Place init code here instead of $(document).ready()
});
Why and how to use jQuery hosted on google

I prefer to leave the version in the file name because there are times when you are changing versions and this is very helpful. At a glance I can see which version I am using on any given webpage.

JavaScript build options/tools

These days I find myself shifting out more and more work to the client side and hence my JS files tend to get bigger and bigger. I have come to the point where most HTML pages have half a dozen or more JS imports in the header and I realised that this is hurting loading times.
I have recently discovered this script which lets you download several JS files with one HTTP request. It is written in PHP and being a Django fan I'm planning to rewrite it in Python. I'm planning to use a HTTP redirect to the pre-minified and concatenated file and was wondering what the cost of a 301 would be. Please let me know if that is a stupid idea.
On the other hand, am little worried about introducing scripting logic into the serving of static files and I was wondering if there is a viable build alternative like, say, an ant task that concatenates and minifies JS files and replaces multiple JS downloads in a HTML header with one big one, like the script does.

For PHP I certainly favour doing it dynamically just because if you introduce a build step you're losing one of the main benefits of using PHP. In fact, at the risk of self-promotion I've written Supercharging Javascript in PHP about this very issue.
Of course other technologies may vary.
Again it is PHP but it's not just a lump of code for you to use (although you can jump straight to Part 6 if you just want some fully working code) and may have value to you in terms of identifying the issues and doing things the right way and why you do them that way.
I favour having bundles of Javascript files (maybe only one for the entire application) and then each page simply activates the behaviour it needs through standard means but all the code bodies are in the larger cached and minified JS file. It works out fastest this way and is a good way to go.
If you do want it as part of a build process, which is a reasonable solution if you have a build process anyway, then I suggest you minify your code. There are lots of tools to do this. Have a look at YUI Compressor.
If you do a static combine of JS files, the other stuff mentioned above such as gzipping and associated issues is still relevant.

YUI compressor is a good choice. If you want to learn how to set up an Ant-based build process, have a look at this Tutorial: http://www.javascriptr.com/2009/07/21/setting-up-a-javascript-build-process/
As a Ruby-based alternative, I would recommend Sprockets

At what stage do you compress/minimize javascript?

When building, or "on the fly" (perhaps with caching) when the users request pages.
And what are the dis/advantages of each.

When the site moves from dev to the live server.
I always have an uncompressed version of the JS on the dev server and a minimized version on the live server.
The advantage of that is when developing i can run into JS trouble and fix it very simply, but i need to run each changed script through a minimizer, but for me its not that much.

When building or deploying to a stage environment is a good time to compress javascript. That way you will have a chance to test it in the stage environment and catch any errors that could occur.
Occasionally, there are errors that do come up when compressing. You may want to include a command-line version of jslint that runs before compressing, to make sure that the js passes. That will minimize, but not eliminate, all compression errors.

I'd imagine that on-the-fly would be an unnesessary unless you were adding dynamic data to the JavaScript (in which case there are better ways around it). IT's just a needless expenditure that will only slow down the page load.
Personally, I'd do so when deploying/building the app, it's a one-time thing really.

I'd say you have the js files as you'd code them in source control, when you kick off an automated build that's when, as part of your build script, it runs all the javascript files through a compressor. This way when you deploy it to a test / staging environment you've got the latest script but also compressed for performance testing and as they would be once they go to production.

I agree that on-the-fly is probably not really necessary (and eats up some cpu cycles) if the JS does not change.
There might be some middleware involved though which can check if the JS has changed and compress it only if requested (and maybe even group various JS files into one resulting one).
A good thing when deploying might also be to add some time stamp or random string as parameter to the JS link (e.g. .../scripts.js?t=cdkjnsccsds7sc8cshcsjhbcs). That way when the JS changes you use a different string and there will be no caching problems because it's a new URL. Same for CSS.