Skip to content
Fetching contributors…
Cannot retrieve contributors at this time
218 lines (169 sloc) 8.05 KB

django-tools.cache contains two cache related things:

  • smooth cache backends
  • per-site cache middleware

smooth cache backends

Same as django cache backends, but you can sooth renew the content by call cache.smooth_update() depend on the current system load.

Using the django cache framework can save processing-overhead. The problem is Validation/Invalidation of cache items.

One simple solution: Do 'cache.clear()' every time the data changed. This ensures that everything is always up to date. But an empty cache causes very high load when many requests are at the same time.

With the "smooth cache backends" it's not needed to clear the complete cache. It renew out dated entries variable based on the system load. The cache would be updated more quickly if system load is low than under heavy load.

The idea is simple:

  • save the timestamp when item added to the cache
  • save the "last change" timestamp with cache.smooth_update()
  • check in cache.get() if and how long this item is out-dated
  • re-use or delete a out-dated item depend on os.getloadavg()[0]

The smooth cache backends works like the origin Django backends. The API is the same. There is only the public method smooth_update() added to the origin cache backends.


If something changed (e.g. a cms page) call cache.smooth_update(), e.g:

def edit_foobar(request):
    if request.method == 'POST':
        form = MyForm(request.POST)
        if form.is_valid():
                cache.smooth_update() # Save "last change" timestamp in django-tools SmoothCacheBackend
            except AttributeError:
                # Fallback if no SmoothCacheBackend used -> clean the complete cache
            # redirect somewhere


setup backends

    'default': {
        'BACKEND': 'django_tools.cache.smooth_cache_backends.SmoothFileBasedCache',
        'LOCATION': '/var/tmp/django_cache',
        'TIMEOUT': 3600 # current value is only a example, it should be a very long time ;)
CACHE_MIDDLEWARE_SECONDS = 3600 # current value is only a example, it should be a very long time ;)

Existing backends:

FileBasedCache django_tools.cache.smooth_cache_backends.SmoothFileBasedCache
DatabaseCache django_tools.cache.smooth_cache_backends.SmoothDatabaseCache
LocMemCache django_tools.cache.smooth_cache_backends.SmoothLocMemCache
MemcachedCache django_tools.cache.smooth_cache_backends.SmoothMemcachedCache
PyLibMCCache django_tools.cache.smooth_cache_backends.SmoothPyLibMCCache


(tuple) This is the information between out-date time and system load. The defaults are:

    # load value, max age in sec.
    (0, 5), #          < 0.1 ->  5sec
    (0.1, 10), #   0.1 - 0.5 -> 10sec
    (0.5, 30), #   0.5 - 1.0 -> 30sec
    (1.0, 60), #   1.0 - 1.5 ->  1Min
    (1.5, 120), #  1.5 - 2.0 ->  2Min
    (2.0, 300), #  2.0 - 3.0 ->  5Min
    (3.0, 900), #  3.0 - 4.0 -> 15Min
    (4.0, 3600), #     > 4.0 ->  1h

These defaults are just probably devised ;)


(String, default: DJANGO_TOOLS_SMOOTH_CACHE_CHANGE_TIME) Cache key value to store the "last change timestamp"


(Integer, default: 10) Time in seconds to update the "last change timestamp" from cache. Note: The timestamp would be stored in the class, too. But this value would not be synced in multi-process environments. So we put/get the timestamp into the cache for sync over all processes. But to safe processing-overhead, we don't fetch the current value from the cache every time.

per-site cache middleware

Similar to django UpdateCacheMiddleware and FetchFromCacheMiddleware, but has some enhancements:

  • Skip cookies and attributes like response.csrf_processing_done
  • ignores response['Vary']
  • Check if 'csrfmiddlewaretoken' is in content
  • stores information about request/response count and cache hits (see cache information below)

The cache key would be generated with:

  • request.get_full_path()
  • request.LANGUAGE_CODE
  • settings.SITE_ID


add this to your settings:



  • "update" must be the first in the list and "fetch" at the last. (see also Order of MIDDLEWARE_CLASSES)
  • django.middleware.locale.LocaleMiddleware must be insert before cache middleware



(Boolean, default: False) If ==True: Don't cache for authenticated users.


(String, e.g.: "my_project.cache.cache_callback", default: None) Path to a call back function for checking if request/response should be caches. e.g.:

from django_tools.cache.site_cache_middleware import logger # Maybe use a own logger

def cache_callback(request, response=None):
    if request.GET:
        logger.debug("Don't cache request with GET query: %s" % repr(request.GET))
        return False

    if response and getattr(response, 'disable_cache', False):
        logger.debug("Don't cache because response.disable_cache==True")
        return False

    return True


(Boolean, default: "runserver" in sys.argv) Are we run with the django developer server? If yes, we do some more checks:

  • Don't cache STATIC files. (request.path starts with settings.STATIC_URL)
  • raise error if a {% csrf_token %} would be stored into the cache (e.g. view dosn't use @csrf_protect decorator)


(Boolean, default: False) creates more logger.debug() output


(Add these values are Boolean and the defaults are False)

The middlewares can store some informations:

  • total number of requests
  • total number cache hits
  • COUNT_UPDATE_CACHE activate counting in UpdateCacheMiddleware.
  • COUNT_FETCH_FROM_CACHE activate counting in FetchFromCacheMiddleware.
  • COUNT_IN_CACHE count global via cache, too. (see below)

Theses counter are stored in two ways:

  • into a local dict (not valid values in multi-process environments)
  • into the cache

into a local dict is the default (settings.COUNT_IN_CACHE==False), because it's very fast and costs almost no performance. The disadvantage is, that the values are only valid for the current process.

In the into the cache solution the valued would be stored into the cache. This costs more or less performances, depend on the used cache backend.

To get the information e.g.:

from django.core.cache import cache

from django_tools.cache.site_cache_middleware import LOCAL_CACHE_INFO,\

def my_view(request):
    context = {
        # from FetchFromCacheMiddleware (if settings.COUNT_FETCH_FROM_CACHE != True: all values are None):
        "local_cache_requests": LOCAL_CACHE_INFO["requests"],
        "local_cache_request_hits": LOCAL_CACHE_INFO["request hits"],
        "global_cache_requests": cache.get(CACHE_REQUESTS),
        "global_cache_request_hits":  cache.get(CACHE_REQUEST_HITS),

        # from UpdateCacheMiddleware (if settings.COUNT_UPDATE_CACHE != True: all values are None):
        "local_cache_responses": LOCAL_CACHE_INFO["responses"],
        "local_cache_response_hits": LOCAL_CACHE_INFO["response hits"],
        "global_cache_responses": cache.get(CACHE_RESPONSES),
        "global_cache_response_hits":  cache.get(CACHE_RESPONSE_HITS),


  • relevant count values are None if counting is deactivated.
  • LOCAL_CACHE_INFO is a normal dict, so you can put it directly into the context etc.
Something went wrong with that request. Please try again.