django-cache-utils

django-cache-utils provides some utils for make cache-related work easier:

• cached decorator. It can be applied to function, method or classmethod and can be used with any django cache backend (built-in or third-party like django-newcache).

  Supports fine-grained invalidation for exact parameter set (with any backend) and bulk cache invalidation (only with group_backend). Cache keys are human-readable because they are constructed from callable's full name and arguments and then sanitized to make memcached happy.

  Wrapped callable gets invalidate methods. Call invalidate with same arguments as function and the cached result for these arguments will be invalidated.

• group_backend. It is a django memcached cache backend with group O(1) invalidation ability, dog-pile effect prevention using MintCache algorythm and project version support to allow gracefull updates and multiple django projects on same memcached instance. Long keys (>250) are auto-truncated and appended with md5 hash.

Installation

pip install django-cache-utils

and then (optional):

# settings.py
CACHE_BACKEND = 'cache_utils.group_backend://localhost:11211/'

Usage

cached decorator can be used with any django caching backend (built-in or third-party like django-newcache):

from cache_utils.decorators import cached

@cached(60)
def foo(x, y=0):
    print 'foo is called'
    return x+y

foo(1,2) # foo is called
foo(1,2)
foo(5,6) # foo is called
foo(5,6)
foo.invalidate(1,2)
foo(1,2) # foo is called
foo(5,6)
foo(x=2) # foo is called
foo(x=2)

class Foo(object):
    @cached(60)
    def foo(self, x,y):
        print "foo is called"
        return x+y

obj = Foo()
obj.foo(1,2) # foo is called
obj.foo(1,2)

With group_backend cached decorator supports bulk O(1) invalidation:

from django.db import models
from cache_utils.decorators import cached

class CityManager(models.Manager):

    # cache a method result. 'self' parameter is ignored
    @cached(60*60*24, 'cities')
    def default(self):
        return self.active()[0]

    # cache a method result. 'self' parameter is ignored, args and
    # kwargs are used to construct cache key
    @cached(60*60*24, 'cities')
    def get(self, *args, **kwargs):
        return super(CityManager, self).get(*args, **kwargs)


class City(models.Model):
    # ... field declarations

    objects = CityManager()

    # an example how to cache django model methods by instance id
    def has_offers(self):
        @cached(30)
        def offer_count(pk):
            return self.offer_set.count()
        return history_count(self.pk) > 0

# cache the function result based on passed parameter
@cached(60*60*24, 'cities')
def get_cities(slug)
    return City.objects.get(slug=slug)


# cache for 'cities' group can be invalidated at once
def invalidate_city(sender, **kwargs):
    cache.invalidate_group('cities')
pre_delete.connect(invalidate_city, City)
post_save.connect(invalidate_city, City)

Notes

If decorated function returns None cache will be bypassed.

django-cache-utils use 2 reads from memcached to get a value if 'group' argument is passed to 'cached' decorator:

@cached(60)
def foo(param)
    return ..

@cached(60, 'my_group')
def bar(param)
    return ..

# 1 read from memcached
value1 = foo(1)

# 2 reads from memcached + ability to invalidate all values at once
value2 = bar(1)

Running tests

cd test_project
./runtests.py