skip to navigation
skip to content

Not Logged In

plone.memoize 1.1.1

Decorators for caching the values of functions and methods

plone.memoize

plone.memoize provides Python function decorators for caching the values of functions and methods.

The type of cache storage is freely configurable by the user, as is the cache key, which is what the function's value depends on.

plone.memoize has support for memcached and is easily extended to use other caching storages. It also has specialized decorators for use with Zope views. However, plone.memoize can be used without Zope.

Introduction

plone.memoize provides Python function decorators for caching the values of functions and methods.

The type of cache storage is freely configurable by the user, as is the cache key, which is what the function's value depends on.

plone.memoize has support for memcached and is easily extended to use other caching storages. It also has specialized decorators for use with Zope views. However, plone.memoize can be used without Zope.

volatile

The 'volatile' module defines a versatile caching decorator that gives you total control of how the cache key is calculated and where it is stored. This decorator is explained in more in depth with example usage in 'volatile.py'.

'volatile' caches can be stored in different places. A common use is a zope RAMCache. There are convenience methods in the 'ram' module to support that.

A quick example of a view that uses 'volatile' caching through the 'ram' module:

>>> from zope.publisher.browser import BrowserView
>>> from plone.memoize import ram
>>> def _render_details_cachekey(method, self, brain):
...    return (brain.getPath(), brain.modified)
>>> class View(BrowserView):
...    @ram.cache(_render_details_cachekey)
...    def render_details(self, brain):
...        obj = brain.getObject()
...        view = obj.restrictedTraverse('@@obj-view')
...        return view.render()

The results of our hypothetical 'render_details' method are cached across requests and independently of the (not) logged in user. The cache is only refreshed when the brain's path or modification date change, as defined in '_render_details_cachekey'.

This is how you could use the same decorator to cache a function's value for an hour:

>>> from time import time
>>> @ram.cache(lambda *args: time() // (60 * 60))
... def fun():
...     return "Something that takes awfully long"
>>> fun()
'Something that takes awfully long'

view and instance

See view.txt and instance.txt for usage of cache decorators that have a fixed cache key and cache storage. The most common usage pattern of these view and instance caching decorators is:

>>> from plone.memoize import instance
>>> class MyClass(object):
...   @instance.memoize
...   def some_expensive_function(self, arg1, arg2):
...       return "Some expensive result"

The first time some_expensive_function() is called, the return value will be saved. On subsequent calls with the same arguments, the cached version will be returned. Passing in different arguments will cause the function to be called again.

Note that this only works if the arguments are hashable!

If you are writing a Zope 3 view, you can do:

>>> from plone.memoize import view
>>> class MyView(BrowserView):
...   @view.memoize
...   def some_expensive_function(self, arg1, arg2):
...       return "Some expensive result"

This has the same effect, but subsequent lookup of the same view in the same context will be memoized as well.

You can also use @view.memoize_contextless to have the memoization not take the context into account - the same view looked up during the same request (but possibly on another context) with the same parameters will be memoized.

Note that this requires that the request is annotatable using zope.annotation!

generic

The generic decorator uses the GenericCache module as storage. By default it'll store into a global cache of its own, with default parameters of 1000 maximal objects and 1 hour maximal lifespan.

You can create your own storage area with its specific parameters using the new_storage method.

Look at the docstring for a few examples.

keys and parameters marshaling

An important issue about caches is how to generate the cache key. In all the decorators above, you can create your own function.

The marshallers module provide with useful default marshallers. args_marshaller will compute a key from function name, module and parameters, applying a hash if asked for. Look into the docstring for usage example.

Changelog

1.1.1 (2011-05-12)

  • Add MANIFEST.in. [WouterVH]

1.1 (2010-07-18)

  • Update license to GPL version 2 only. [hannosch]
  • Solve intermittent error during testing of CleanupDict class, as a cleanup period of zero seconds would not always result in a cleanup, if the tests were run fast. [maurits]

1.1b1 (2009-11-13

  • Updated package documentation. [hannosch]

1.1a4 (2009-07-23)

  • Use the new zope.ramcache package instead of zope.app.cache. This reduces our dependencies by quite a bit. [hannosch]

1.1a3 (2009-05-10)

  • Modernized and cleaned up the code a bit. Also drop BBB support for zope.app.annotation and made the tests work again. [hannosch]

1.1a2 (2009-05-08)

  • Removed all testing dependencies on zope.app packages. [hannosch]
  • Changed a test to use zope.publisher instead of a Five BrowserView. This removes the entire Zope2 dependency. [hannosch]
  • Correct Zope2 dependency, it's indeed only a test dependency. [hannosch]

1.1a1 (2009-04-04)

  • Clarified license statements. [hannosch]
  • Moved declaration of test dependencies into a test extra. [hannosch]
  • Avoid deprecation warnings for the md5 and sha modules in Python 2.6. [hannosch]
  • Specify package dependencies. [hannosch]
  • Added check for Unicode values in cache keys before calculating md5 checksums, as the md5 module doesn't seem to like Unicode. [hannosch]
  • Removed BBB code for zope.app.annotation. [hannosch]
  • The clearbefore decorator was mistakenly not tested. [maurits]

1.0.4 (2008-03-31)

  • Documentation and release notes cleanup. [hannosch]

1.0.3 (2007-11-09)

  • Remove features from Plone 3.0 branch. [nouri]

  • Maintenance branch for Plone 3.0. [nouri]

  • Get rid of sys.modules hack, which according to this changeset: http://dev.plone.org/plone/changeset/15030 was added because I advised it generally. With help from Kapil for the PloneGetPaid project I figured out a better way. [maurits]

  • Revise docs and project description. [nouri]

  • Merge patch from Gael Le Mignot:

    • Do not use hash anymore when making cache keys. This is to avoid cache collisions, and to avoid a potential security problem where an attacker could manually craft collisions. Also, stop recommending the use of hash() in tests.
    • Add support for using Pilot System's GenericCache as a backend for 'plone.memoize.volatile.cache'.
    • Add an arguments marshaller that gives you a more convenient way to declare that your cache is dependent on arguments. See 'plone.memoize.marshallers'.

    [nouri, gael]

1.0.1 (2007-09-10)

  • Simplify forever by reuse of stuff from plone.memoize.volatile. [nouri]

1.0 (2007-08-17)

  • Add a forever memo - lives until Zope restart. [optilude]
  • hash((1, 2)) returns something different on ree's 64-bit Python :) [nouri]
  • Don't treat None in a special way. Avoid one dict lookup. [nouri]
  • Extended the xhtml_compress method to use a utility lookup for IXHTMLCompressor utilities instead. Now you can turn the slimmer based compression on via a simple utility registration. See compress.py. [hannosch, fschulze]

1.0rc2 (2007-07-27)

  • Added simple xhtml_compress method which can be used to plug in whitespace removal libraries. Peter Bengtsson's slimmer library is configured but not enabled by default. [hannosch]

1.0b4 (2007-07-09)

  • Use a md5 hash of the provided key in RAMCacheAdapter, reducing the memory footprint and speeding up lookup time. [hannosch]
  • Reword the volatile section a bit to indicate why the example does not use anything from the volatile module. [wichert]
  • Use an exception DontCache instead of the DONT_CACHE marker return value. Allow for no ICacheChooser to be registered. [nouri]
  • Add cache decorator for request (which can in fact be used for all sorts of annotatable objects). [nouri]
  • Added decorator for storing cache values on the request as annotations. [nouri]
  • Always include the function's dotted name in the key. [nouri]
  • Added a new cache decorator which can memoize a the result of a method call on the request and lets you specify which argument on the function is the request. [hannosch]
  • Add MemcacheAdapter as an alternative to RAMCacheAdapter. [nouri]
  • Generalize IRAMCacheChooser to ICacheChooser, which doesn't return an IRAMCache but a simple dict. [nouri]
  • Use a more sensible default for the maxAge of the new RAMCache. [hannosch]
  • Add cache storage for plone.memoize.volatile for use with zope.app.cache.ram.RAMCache. [nouri]
  • Rolled in changes from memojito to fix recursively memoized methods(fix by Rob Miller and Whit Morriss) [whit]
  • Made plone.memoize backwardly compatible with zope2.9 and remain usable w/out zope.annotation. Minor wording changes to some docs. [whit]
  • Per default, use a volatile dict that cleans up after itself. [nouri]
  • This 'volatile' module defines a versatile caching decorator that gives you total control of how the cache key is calculated and where it is stored. [nouri]

1.0b3 (2007-05-05)

  • Initial package structure and implementation. [hannosch, nouri, optilude, whit, zopeskel]
 
File Type Py Version Uploaded on Size
plone.memoize-1.1.1.zip (md5) Source 2011-05-12 51KB
  • Downloads (All Versions):
  • 113 downloads in the last day
  • 660 downloads in the last week
  • 3582 downloads in the last month