Motivation

Why not use core.memoize?

Multiple eviction strategies

I've decided to use the Guava Cache and it allows similar cache eviction strategies than core.memoize like ttl, fifo, size-based evicion, lu, lru. But unlike core.cache caches, Guava allows multiple policies at the same time. And it also has more policies, so this is definitely an additional killer feature.

Shared caches

When you have many memoized functions in the project, and each has its own cache, it is hard to predict the full impact of your eviction policies, e.g. if you're using size based policy, it is hard to know exactly how many cache entries there are in the system as a whole. Sometime you just want to say I want to cache max 10000 entries across all my memoized functions. This is not possible in core.memoize. Or you simply want one cache for most functions and a different one with high concurrency number for a specific function. Being able to specify same cache for multiple functions is an important feature.

Lexically scoped (request) cache

A lot of the time the validity of cache data has nothing to do with time or count, but you want to get the cached data within a certain live scope such as request.

This is a very important feature.

Tagged evictions

I need the capability to evict items based on a tag. If I have 100 memoized functions in a project that cache parts of a Person entity, then I want a command like 'evict all info in connection with person 4', which should evict multiple entries in multiple memoized caches.

Nice to have features

Alternate keys

General memoization utilities consider all function arguments part of the key... different arguments different entry. But sometimes some arguments don't really affect the result of the function, but rather drive something else (like a log statement, or an error function). It would be nice to be able to specify an alternative key function.

Extensibility

It would be nice to leave open access to the underlying cache object, in order to be able to use the full power of implementation. Guava cache provides stats keeping, and other thing like refreshes, removal listeners.

Library Non-goals

Performance

It is not the goal of this library to provide an implementation of function memoization with a massive throughput. The use-case for this memoization is calling external APIs or database and the performance target is to stay a couple of orders of magnitude faster than a network access.