Cache decorators


How to use the Python decorator pattern to cache the result values of your computationally expensive method calls.


Cache decorators are convenient methods caching of function return values.

Use them like this:

def my_slow_function():
    # This is run only once and all subsequent calls get value from the cache


Cache decorators do not work with methods or functions that use generators (yield). The cache will end up storing an empty value.

The plone.memoize package offers helpful function decorators to cache return values.

See also using memcached backend for memoizers.

Cache result for process lifecycle


from plone.memoize import forever

def getFields(area, subject):
    """ Get all fields inside area / subject.

    Results is cached for process lifetime.

    @return: List of internal fields
    schema = getSchema(area)
    return [field for field in schema if field["subject"] == subject]

Timeout caches

The @ram.cache decorator takes a function argument and calls it to get a value. So long as that value is unchanged, the cached result of the decorated function is returned. This makes it easy to set a timeout cache:

from plone.memoize import ram
from time import time

@ram.cache(lambda *args: time() // (60 * 60))
def cached_query(self):
    # very expensive operation,
    # will not be called more than once an hour

time.time() returns the time in seconds as a floating point number. "//" is Python's integer division. So, the result of time() // (60 * 60) only changes once an hour. args passed are ignored.

Caching per request

This pattern shows how to avoid recalculating the same value repeatedly during the lifecycle of an HTTP request.

Caching on BrowserViews

This is useful if the same view/utility is going to be called many times from different places during the same HTTP request.

The plone.memoize.view package provides necessary decorators for BrowserView-based classes.

from plone.memoize.view import memoize, memoize_contextless

class MyView(BrowserView):

    def getValue():
        """ This value is recalculated for every new BrowserView context
            per request.
        return "something"

    def getValueNoContext():
        """ This value is recalculated for all context objects once per
        return "something"

Caching on Archetypes accessors

If you have a custom Archetypes accessor method, you can avoid recalculating it during the request processing.


def getParsedORADataCached(self):
    """ Same as above, but does not run through JSON reader every time.

    # Manually store the result on HTTP request object annotations

    # Use informative string + Archetypes unique identified as the key
    key = "parsed-ora-data-" + self.UID()

    cache = IAnnotations(self.REQUEST)
    data = cache.get(key, None)
    if data is not None:
        data = self.getParsedORAData()
        cache[key] = data

    return data

Caching using global HTTP request

This example uses the five.globalrequest package for caching. Values are stored on the thread-local HTTPRequest object which lasts for the transaction lifecycle:

from zope.globalrequest import getRequest
from zope.annotation.interfaces import IAnnotations

    def _getProductList(self, type, language):
        """ Private implementation, builds list of products.
        """"Getting product list %s %s" % (type, language))
        return result

    def getProductListCached(self, type, language):
        """ Public cached method, delegates to _getProductList.

        request = getRequest()

        key = "cache-%s-%s" % (type, language)

        cache = IAnnotations(request)
        data = cache.get(key, None)
        if not data:
            data = self._getProductList(type, language)
            cache[key] = data

        return data

Testing memoized methods inside browser views

While testing browser views memoized methods you could find out that calling a method multiple times inside a test could result in getting the same result over and over, no mater what the parameters are, because you have the same context and request inside the test and the result is being cached.

One approach to by-pass this is to put your code logic inside a private method while memoizing a public method with the same name that only calls the private one:

from plone.memoize import view
from Products.Five import BrowserView

class MyView(BrowserView):

    def _my_expensive_method():
        """Code logic goes here.
        return "something"

    def my_expensive_method():
        """We just call the private method here and memoize the result.
        return self._my_expensive_method()

In your tests you can call the private method to avoid memoization.