A memoize library which can be used standalone, or plugged into key/value stores such as Redis. Also contains functionality to invalidate cache based on function name and arguments.

There are already several python libraries providing memoizing in some form, ranging from very basic dict caches, more performance focused, to very advanced, comprehensive libraries. However, I've caught myself writing this particular implementation in two separate projects, and I gathered I should extract it to a separate library. It has a couple of strengths:

• It is flexible - the two arguments unique_on and conditions provide a powerful configurability in how and what you actually want to memoize, making it ideal in situations where memoization is introduced to speed up legacy code. Paired with the invalidate_memoize function, there's no limit to how specific you can have your caching be.

• It is extendable - currently shipped only with support for dict-type cache and Redis, it is very easy to write new extensions for other storage alternatives. The test blueprints are even already written!

Using Memoizit completely bare with in-memory caching:

pip install memoizit

If you want to (be able to) use Redis as cache storage:

pip install memoizit[redis]

from memoizit import Memoizer

m = Memoizer()

@m.memoize()
def heavy_calculation(n):
    return n**n

Default configuration looks for Redis at localhost:6379. If this matches your setup, simply run

m = Memoizer(backend="redis")

Otherwise, Redis can be configured using either keyword arguments in the class constructor, or by setting environment variables:

REDIS_HOST
REDIS_PORT
REDIS_USERNAME
REDIS_PASSWORD

m = Memoizer(
    backend="redis",
    host="redis",
    port="63791",
    username="johndoe",
    password="hunter2"
)

The memoize decorator has three arguments:

• expiration - seconds until the memoized value is invalidated (default 300 - 5 minutes).
• unique_on - arguments to include in the cache key. Can be both args and kwargs (mainly useful to avoid self arguments in class methods). If ignored, all arguments will be used.
• conditions - list of arguments and conditional values that need to match in order to apply memoization.

class Office:

    @m.memoize(
        expiration=60*60*24,
        unique_on=['department', 'include_management'],
        conditions=[('with_sideeffects', False)]
    )
    def get_employees(self, department, include_management=False, with_sideeffects=False):
        <expensive external requests>
        <possible sideeffects>
        return employees

⚠ NB: Note that omitting self from unique_on might pollute other class instances with incorrect cache.

This library includes the strange function invalidate_memoize. This implements a limited wildcard functionality, where arguments and keyword arguments can be provided to specify which cache entries to delete, but only in a left-to-right order. Example:

@m.memoize()
perform_miracle(country_id, miracle_type, target="cats"):
    <miracle>

This function is then called with the following arguments:

perform_miracle(2, "divine", target="dogs")

This will find the cache entry and delete it (specifying argument name is optional):

m.invalidate_memoize('perform_miracle', 2, "divine")

While this will not work:

m.invalidate_memoize("perform_miracle", 2, target="dogs")

This function can be particulary useful when memoizing read functions from an API wrapper, where you want to invalidate the cache on update or delete.

Clone repo:

git clone [email protected]:blixhavn/memoizit.git
cd memoizit

Set up virtual environment and activate it:

python3 -m virtualenv venv
source venv/bin/activate

Install with dev requirements:

pip install -e .[dev]

Run tests and code checks as follows:

black .
mypy .
flake8
pytest 

The Redis tests are skipped unless Redis is available at localhost:6379.

Then, create a pull request with your work! For your convenience there is also a pre-commit file that can be copied into .git/hooks/, which will run black, mypy and flake8 before allowing commits. Note however that this does not work in Windows.

Memoizit is released under the MIT license. Have at it.