supycache

supycache

Simple yet capable caching decorator for python

Source code: https://github.com/lonetwin/supycache

Install using pip: pip install supycache or download from https://pypi.python.org/pypi/supycache

What is supycache ?

supycache is a decorator that enables caching of return values for time-consuming functions, either in memory or on a cache server such as memcached <http://memcached.org/>_ or redis <http://redis.io/>_.

The cache keys can either be indedependent or dependent (completely or partially) on the arguments passed to the function.

This is different from other similar caching decorators, for instance, functools.lru_cache <https://docs.python.org/3/library/functools.html#functools.lru_cache>_ which is dependent on all the arguments passed to the function and requires the arguments to be hashable.

If you use the default cache backend (ie: supycache.backends.DictCache), you can also provide an age for the cached values

Here's an example of how you might use supycache

.. code:: python

import time
import supycache

@supycache.supycache(cache_key='result', max_age=5)
def execute_expensive():
    print 'original function called'
    time.sleep(15)
    return 42

print execute_expensive()  # This will take 15 seconds to execute ...
original function called
42
print execute_expensive()  # ...not this tho', because the value is cached ...
42
print supycache.default_backend.get('result') # ..keyed as `result`
42
time.sleep(5)              # wait for the cache to expire...
execute_expensive()        # This will again take 15 seconds to execute ...
original function called
42
print execute_expensive()  # ...not this tho', because the value is re-cached ...
42

Sometimes you might want to be aware of the arguments that are passed to the function:

.. code:: python

@supycache(cache_key='sum_of_{0}_and_{1}')   # Cache the sum of x and y creating a
def cached_sum(x, y):                        # key based on the arguments passed
    return x + y

print cached_sum(28, 14)
42
print supycache.default_backend.get('sum_of_28_and_14')
42

You can also create the key based on partial arguments or on the attributes/items within the arguments.

.. code:: python

class User:
    def __init__(self, name, session_key):
        self.name = name
        self.session_key = session_key

@supycache(cache_key='{user_obj.name}')   # build the cache key dependent on *just*
def get_username(user_obj):               # the `.name` attribute
    time.sleep(15)
    return user_obj.name

a = User(name='steve', session_key='0123456789')
b = User(name='steve', session_key='9876543210') # same name, different session

print get_username(user_obj=a)   # This will take 15 seconds to execute ...
steve
print get_username(user_obj=a)   # ...not this tho'...
steve
print get_username(user_obj=b)   # ...and neither will this !
steve


@supycache(cache_key='{choices[0]}_{menu[lunch]}')         # build the cache
def supersized_lunch(ignored, choices=None, menu=None):    # key dependent on
    time.sleep(15)                                         # partial arguments
    return 'You get a %s %s' % (choices[-1], menu['lunch'])

menu = {'breakfast' : 'eggs',
        'lunch'     : 'pizza',
        'dinner'    : 'steak'}

sizes = ['small', 'medium', 'large', 'supersize']

print supersized_lunch('ignored', choices=sizes, menu=menu)
You get a supersize pizza       # This will take 15 seconds to execute ...

print supersized_lunch('changed', choices=sizes, menu=menu)
You get a supersize pizza       # ...not this tho'...

If that format specification for the cache_key looks familiar, you've discovered the secret of supycache !

.. code:: python

@supycache(backend=memcached_backend, cache_key='{0}_{kw[foo]}_{obj.x}')
def custom_key_built_from_args(positional, kw=None, obj=None):
    # now, supycache will build the `cache_key` from the arguments passed and
    # use the memcached_backend instance to `set` the key with the return value
    # of this function
    return 'cached'

The secret of supycache is quite simple -- it calls .format() on the cache_key/expire_key with the passed args and kwargs to build the actual key.

However, if you'd like to have more control on the way the cache_key/expire_key are created, simply pass in a callable !

.. code:: python

def extract_path(url=None, *args, **kwags):
    return urlparse.urlparse(url).path

@supycache(cache_key=extract_path, ignore_errors=False)
def do_something_with(url):
    # will call `extract_path` at runtime passing `url` as parameter and
    # will use the returned value as the cache key. Also, don't ignore any
    # errors in the entire process if something fails (the default is to
    # ignore any caching errors and just return the result as tho' this
    # function was undecorated.
    return 'cached'

do_something_with('http://www.example.com/foo/bar')
'cached'
supycache.default_backend.get('/foo/bar')
'cached'

The backend interface is abstarcted out neatly so that backends can be swapped out without too much hassle. As long as the passed in object has a get(), set() and delete() methods, it can be passed to supycache as a backend or can be set as the default_backend.

Right now though, this project has only the code and tests, no docs (barring some docstrings !). I'll be adding them soon. If interested take a look at the tests to see the typical usage and try it out. Feedback, bug reports and pull requests would be great !

Help required

I would really appreciate any help you could offer, not just in implementation but also in validating the packaging and distribution of this module via pypi since I've not distributed any packages before.

Besides that I plan on adding a few more things:

* Ability to specify a ``max_age`` for all backends.
* I'm not sure whether I am doing the right thing for the not the packaging
  of the memcached dependency. I'd like to automatically include the
  support for ``memcached`` or ``redis`` backends if the python memcached
  or redis modules are installed.
* logging support