Pickle function cache

Allows for caching of the results of function calls in a file using pickle (builtin shelve did not work for some reason and marshal was not considered; json was not able to serialize all the things I needed).

• the file read happens only once per process per filename (all the function caches in the same cache file are loaded once)
• if the result is not found in the cache, the result is computed using the original function and is written into the cache file and to the runtime cache.
• if the result for the same set of arguments is requested and found the original function is not called but the cached result is returned.

Usage

from pickle_function_cache import cache_responses


@cache_responses("cache.dat")
def f1(param1, param2):
    ...  # take a lot of time


@cache_responses("cache.dat")
def f2(param1, param2, param3):
    ...  # take a lot of time

# use function normally ...

• when first called, f1(1,2) computes the result using the original, f1
• afterwards f1(1,2) returns the cached result, omitting the function call
• if you call f2(1,2,3), it will be no trouble because function caches are stored separately per function name in the same file (note: cached function names should be unique)

Configuration

You can set the env var OMIT_PICKLE_FUNCTION_CACHE to either 1 or true to disable the cache at compile time without changing the code.

Performance

The cache was tested on a small set of quick functions and a small test list of examples and did not introduce a significant overhead. Probably, if the chosen functions for cache were more computationally intensive, we would see a speed-up because of the cache.