WSGI caching middleware.
PageCache is a WSGI middleware that can be used to cache complete responses from WSGI applications. PageCache works well with memcached but can be used with other caching backends as well.
PageCache has not been widely tested, use with caution.
Here's a simple example to get you started:
from pagecache import PageCacheMiddleware app = ... cache = memcache.Client(...) cached_urls = (('/foo', 30, 120), # url prefix, TTL, grace period ('/bar', 3600, 60)) app = PageCacheMiddleware(app, cached_urls, cache)
The list of urls to cache is given in the
cached_urls list (or tuple).
Each entry of this list is a tuple in the following format:
(<url prefix>, <TTL in seconds>, <grace period in seconds>)
Pages are served from the cache in the TTL period. When the TTL period has expired the grace period begins. The first request coming in in the grace period will recalculate the page and store the new result in the cache. While the new result is being calculated stale results are served from the cache (until the grace period expires). This is to protect against the dog-pile effect
The cache object
PageCache was only tested with memcached, but it may work with other cache
backends as well. The only requirement is that the cache object should have
the following methods with reasonably similar semantics as in memcached:
A note on cookies
All Set-Cookie headers returned by the application will be ignored. Also, depending on your application you'll most probably want to make sure that the application doesn't use any incoming cookie information to calculate responses that will be cached.
You can install PageCache from PyPI:
pip install pagecache
or you can install the latest version from the github repository:
git clone git://github.com/abiczo/pagecache.git cd pagecache python setup.py install
- more unit testing
- better syntax for the cached urls configuration
- regexp based url matching
- configurable cache keys (so that multiple applications can use the same memcached instance without having to worry about having the same cached urls)
- configurable request charset
Patches / pull-requests are welcome.