shelved-cache

from shelved_cache import PersistentCache

Shelved Cache

Persistent cache implementation for Python cachetools.

Behaves like any Cache implementation, but entries are persisted to disk.

Original repository: https://github.com/mariushelf/shelved_cache

Usage example

from shelved_cache import PersistentCache
from cachetools import LRUCache

filename = 'mycache'

# create persistency around an LRUCache
pc = PersistentCache(LRUCache, filename=filename, maxsize=2)

# we can now use the cache like a normal LRUCache.
# But: the cache is persisted to disk.
pc["a"] = 42
pc["b"] = 43

assert pc["a"] == 42
assert pc["b"] == 43

# close the file
pc.close()

# Now in the same script or in another script, we can re-load the cache:
pc2 = PersistentCache(LRUCache, filename=filename, maxsize=2)
assert pc2["a"] == 42
assert pc2["b"] == 43

Use as a decorator

Just like a regular cachetools.Cache, the PersistentCache can be used with cachetools' cached decorator:

import cachetools
from shelved_cache import PersistentCache
from cachetools import LRUCache

filename = 'mycache'
pc = PersistentCache(LRUCache, filename, maxsize=2)

@cachetools.cached(pc)
def square(x):
    print("called")
    return x * x

assert square(3) == 9
# outputs "called"
assert square(3) == 9
# no output because the cache is used

Note: decorating multiple functions

If you want to decorate multiple functions, you need to use a new instance of PersistentCache for each function. Make sure that each cache uses a different file name.

import cachetools
from shelved_cache import PersistentCache
from cachetools import LRUCache

@cachetools.cached(PersistentCache(LRUCache, "square.cache", maxsize=100))
def square(x):
    return x * x

@cachetools.cached(PersistentCache(LRUCache, "cube.cache", maxsize=100))
def cube(x):
    return x * x * x

assert square(2) == 4
assert cube(2) == 8

Features

persistent cache

See usage examples above.

Async decorators

The package contains equivalents for cachetools' cached and cachedmethod decorators which support wrapping async methods. You can find them in the decorators submodule.

They support both synchronous and asynchronous functions and methods.

Examples:

from shelved_cache import cachedasyncmethod
from cachetools import LRUCache

class A:
    # decorate an async method:
    @cachedasyncmethod(lambda self: LRUCache(2))
    async def asum(self, a, b):
        return a + b

a = A()
assert await a.asum(1, 2) == 3
    
class S:
    @cachedasyncmethod(lambda self: LRUCache(2))
    def sum(self, a, b):
        return a + b

s = S()
assert s.sum(1, 2) == 3

Support for lists as function arguments

Using the autotuple_hashkey function, list arguments are automatically converted to tuples, so that they support hashing.

Example:

from cachetools import cached, LRUCache
from shelved_cache.keys import autotuple_hashkey

@cached(LRUCache(2), key=autotuple_hashkey)
def sum(values):
    return values[0] + values[1]

# fill cache
assert sum([1, 2]) == 3

# access cache
assert sum([1, 2]) == 3

Changelog

0.4.0

• drop support for Python 3.7 and 3.8
• add support for Python 3.12 and 3.13
  • note: shelved_cache does not seem to work with Python 3.13 on Windows
• note about decorating multiple functions in the README
• improvement in async decorators

0.3.1

• fix for Windows users
• add Windows and MacOS to test suite

0.3.0

• add support for Python 3.10 and 3.11
• better error message when trying to use the same file for multiple caches
• CI/CD pipeline
• fixes for documentation

0.2.1

• improved error handling

Acknowledgements

• cachetools by Thomas Kemmer
• asyncache by hephex

License

Author: Marius Helf ([e-mail])

License: MIT -- see LICENSE