{% load utils %}
<link rel="stylesheet" type="text/css" href="{% version '/static/css/style.css' %}">

from django import template
register = template.Library()
import os, re        

STATIC_PATH="/path/to/templates/"
version_cache = {}

rx = re.compile(r"^(.*)\.(.*?)$")
def version(path_string):
    try:
        if path_string in version_cache:
            mtime = version_cache[path_string]
        else:
            mtime = os.path.getmtime('%s%s' % (STATIC_PATH, path_string,))
            version_cache[path_string] = mtime

        return rx.sub(r"\1.%d.\2" % mtime, path_string)
    except:
        return path_string 

register.simple_tag(version) 

/static/css/style.css?v=1207433992
/static/css/style.1207433992.css

RewriteRule ^/static/(.*?)\.[0-9]+\.(css|js|jpe?g|gif|png) /static/$1\.$2 [L]

Comments: Add yours

1. On April 8th, 2008 at 7:43 am Brad Wright said:

Hey Stu,

It seems Opera and Safari don’t cache file paths with query strings, as per the HTTP spec, and given that it’s part of the spec we have no guarantee that other browsers won’t behave the same in the future. Probably best to do something like:

return path_string.replace('.css', '.%d.css' % mtime)

And have something like this in mod_rewrite:

RewriteRule ^/static/([a-z-]+)\.[0-9]+\.(css|js) /static/$1\.$s [L]

This keeps the filenames looking like: /static/css/style.1207433992.css, but rewrites them to the latest versions in your static server.

2. On April 8th, 2008 at 8:35 am Brad Wright said:

Oops, my mod_rewrite had fail:

RewriteRule ^/static/([a-z-]+)\.[0-9]+\.(css|js) /static/$1\.$2 [L]

Is better.

3. On April 8th, 2008 at 10:34 am Stuart Colville said:

@Brad: Excellent points well made. I’ve updated the script accordingly.

4. On April 10th, 2008 at 7:32 am Ross Bruniges said:

I have a feeling that this will provide very very useful to a number of people :>

5. On April 28th, 2008 at 5:50 am Rich said:

I was just looking around for some background to write something just like this for Django!

Thanks very much for saving me several hours!

- rich

6. On June 10th, 2008 at 8:44 pm Lukasz Korzybski said:

Great stuff, thanks a lot

One note thought:

Atually as Steve pointed out in his comment to High Performance Web Sites: Rule 3 – Add an Expires Header the HTTP 1.1 specs says that URLs with query strings should not be cached only when there is no explicit expiration information:

From HTTP 1.1 spec, section 13.9:

e note one exception to this rule: since some applications have
traditionally used GETs and HEADs with query URLs (those containing a
“?” in the rel_path part) to perform operations with significant side
effects, caches MUST NOT treat responses to such URIs as fresh unless
the server provides an explicit expiration time.

Also, RoR have some kind of template tag for auto versioning and it uses QS as far as I know.

– Lukasz

7. On December 8th, 2008 at 1:28 pm Łukasz Korzybski said:

Recently I found this very interesting post, by Steve, about proxy and query strings issues. Steve proved that Squid’s default configuration is not to cache URLs with query strings.

http://www.stevesouders.com/blog/2008/08/23/revving-filenames-dont-use-querystring/

So at the end it seems that using conservative file name versioning is better and more safe than QS method.

8. On October 22nd, 2009 at 11:51 pm Ben said:

For the regexp substitution, I think you mean:
rx.sub(r"\1.%d.\2" % mtime, path_string)

Without the raw string you’ll get binary in your return value the dots will be wildcards.

This was a helpful article, though. Thanks!

9. On October 27th, 2009 at 9:37 pm Stuart Colville said:

@Ben: Thanks – I’ve updated the code example