A Redis cache backend for Django
Docs can be found at http://django-redis-cache.readthedocs.org/en/latest/.
- Drops support for Django < 1.8 and Python 3.2.
- Confirms support for Django 1.9 and 1.10.
- Adds ability to compress/decompress cache values using pluggable compressors
including zlib, bzip2, or a custom implementation.
- Adds support for providing a socket timeout on the redis-py client.
- Adds support for pluggable serializers including pickle(default), json,
msgpack, and yaml.
- Deprecate support for Python 2.6. The cache should still work, but tests
will fail and compatibility will not be guaranteed going forward.
Backward incompatibilities:
- The
HashRingbehavior has changed to maintain a proper keyspace balance.
This will lead to some cache misses, so be aware.
- Now requires redis-py >= 2.10.3
- Deprecate support for django < 1.3 and redis < 2.4. If you need support for those versions,
- pin django-redis-cache to a version less than 1.0, i.e. pip install django-redis-cache<1.0
- Application level sharding when a list of locations is provided in the settings.
- Delete keys using wildcard syntax.
- Clear cache using version to delete only keys under that namespace.
- Ability to select pickle protocol version.
- Support for Master-Slave setup
- Thundering herd protection
- Add expiration to key using expire command.
- Add persistence to key using persist command.
- Adds custom has_key implementation that uses Redis's exists command.
- This will speed has_key up drastically if the key under question is extremely large.
- Keys can now be kept alive indefinitely by setting the timeout to None,
- e.g. cache.set('key', 'value', timeout=None)
- Adds ttl method to the cache. cache.ttl(key) will return the number of
- seconds before it expires or None if the key is not volitile.
- Adds support for specifying the connection pool class.
- Adds ability to set the max connections for the connection pool.
Adds Support for Python 3.3 and Django 1.5 and 1.6. Huge thanks to Carl Meyer for his work.
Redis cache now allows you to use either a TCP connection or Unix domain socket to connect to your redis server. Using a TCP connection is useful for when you have your redis server separate from your app server and/or within a distributed environment. Unix domain sockets are useful if you have your redis server and application running on the same machine and want the fastest possible connection.
You can now specify (optionally) what parser class you want redis-py to use
when parsing messages from the redis server. redis-py will pick the best
parser for you implicitly, but using the PARSER_CLASS setting gives you
control and the option to roll your own parser class if you are so bold.
redis-py >= 2.10.3 redis >= 2.4 hiredis python >= 2.7
- Run
pip install django-redis-cache. - Modify your Django settings to use
redis_cache.
# When using TCP connections
CACHES = {
'default': {
'BACKEND': 'redis_cache.RedisCache',
'LOCATION': [
'<host>:<port>',
'<host>:<port>',
'<host>:<port>',
],
'OPTIONS': {
'DB': 1,
'PASSWORD': 'yadayada',
'PARSER_CLASS': 'redis.connection.HiredisParser',
'CONNECTION_POOL_CLASS': 'redis.BlockingConnectionPool',
'CONNECTION_POOL_CLASS_KWARGS': {
'max_connections': 50,
'timeout': 20,
},
'MAX_CONNECTIONS': 1000,
'PICKLE_VERSION': -1,
},
},
}
# When using unix domain sockets
# Note: ``LOCATION`` needs to be the same as the ``unixsocket`` setting
# in your redis.conf
CACHES = {
'default': {
'BACKEND': 'redis_cache.RedisCache',
'LOCATION': '/path/to/socket/file',
'OPTIONS': {
'DB': 1,
'PASSWORD': 'yadayada',
'PARSER_CLASS': 'redis.connection.HiredisParser',
'PICKLE_VERSION': 2,
},
},
}
# For Master-Slave Setup, specify the host:port of the master
# redis-server instance.
CACHES = {
'default': {
'BACKEND': 'redis_cache.RedisCache',
'LOCATION': [
'<host>:<port>',
'<host>:<port>',
'<host>:<port>',
],
'OPTIONS': {
'DB': 1,
'PASSWORD': 'yadayada',
'PARSER_CLASS': 'redis.connection.HiredisParser',
'PICKLE_VERSION': 2,
'MASTER_CACHE': '<master host>:<master port>',
},
},
}django-redis-cache shares the same API as django's built-in cache backends, with a few exceptions.
cache.delete_pattern
Delete keys using glob-style pattern.
example:
>>> from news.models import Story
>>>
>>> most_viewed = Story.objects.most_viewed()
>>> highest_rated = Story.objects.highest_rated()
>>> cache.set('news.stories.most_viewed', most_viewed)
>>> cache.set('news.stories.highest_rated', highest_rated)
>>> data = cache.get_many(['news.stories.highest_rated', 'news.stories.most_viewed'])
>>> len(data)
2
>>> cache.delete_pattern('news.stores.*')
>>> data = cache.get_many(['news.stories.highest_rated', 'news.stories.most_viewed'])
>>> len(data)
0
cache.clear
Same as django's cache.clear, except that you can optionally specify a
version and all keys with that version will be deleted. If no version is
provided, all keys are flushed from the cache.
cache.reinsert_keys
This helper method retrieves all keys and inserts them back into the cache. This is useful when changing the pickle protocol number of all the cache entries. As of django-redis-cache < 1.0, all cache entries were pickled using version 0. To reduce the memory footprint of the redis-server, simply run this method to upgrade cache entries to the latest protocol.
A common problem with caching is that you can sometimes get into a situation where you have a value that takes a long time to compute or retrieve, but have clients accessing it a lot. For example, if you wanted to retrieve the latest tweets from the twitter api, you probably want to cache the response for a number of minutes so you don't exceed your rate limit. However, when the cache entry expires you can have mulitple clients that see there is no entry and try to simultaneously fetch the latest results from the api.
The way to get around this problem you pass in a callable and timeout to
get_or_set, which will check the cache to see if you need to compute the
value. If it does, then the cache sets a placeholder that tells future clients
to serve data from the stale cache until the new value is created.
Example:
tweets = cache.get_or_set('tweets', twitter.get_newest, timeout=300)
./install_redis.sh
make test