Python library to cache data and function results in temporary files

This library offers a simple way to cache data and function results using temporary files, including a mechanism for automatic expiration after a certain time.

Dy default, the package uses the pickle module to serialize data and hash key values.

Note For more advanced use cases you may want to look at the Memory class in joblib.

Basic Usage

An instance of the TempCache class be used as a decorator to automatically cache the results of a function.

from tempcache import TempCache

CACHE_MAX_AGE = 24 * 60 * 60 * 2    # two days
cache = TempCache("mycache", max_age=CACHE_MAX_AGE)

@cache
def long_running(...):
    ...

result = long_running(...)

Caching results at the call site

You can also use a TempCache object to cache a result at the call site with the cache_result method.

from tempcache import TempCache

CACHE_MAX_AGE = 24 * 60 * 60 * 2    # two days
cache = TempCache("mycache", max_age=CACHE_MAX_AGE)

def long_running(...):
    ...

result = cache.cache_result(long_running, ...)

Advanced usage

In cases where the function or some of its arguments are defined in the __main__ namespace or in a jupyter notebook and cannot be pickled by pickle you can use a different pickle module like cloupickle.

import cloudpickle

from tempcache import TempCache

CACHE_MAX_AGE = 24 * 60 * 60 * 2    # two days
cache = TempCache("mycache",
                    pickler=cloudpickle,
                    max_age=CACHE_MAX_AGE)

key = ...
# key object can be complex as long as it is pickle-able

item = cache.item_for_key(key)
# cache item for the given key whether it exists or not

# load item if it exists
if item.exists():
    value = item.load()

# save item
item.save(value)

Examples

Examples notebooks are in the extras folder.

Installation

You can install this package with pip.

pip install tempcache

Related Projects

• joblib Computing with Python functions
• percache Persistently cache results of callables
• disckcache Disk and file backed cache library compatible with Django
• cloudpickle Extended pickling support for Python objects