caches is a Ruby gem, providing a small collection of caches with good performance and hash-like access patterns. Each is named for its cache expiration strategy - when it will drop a key.

caches provides the following classes.

TTL (Time To Live) remembers values for as many seconds as you tell it. The default is 3600 seconds (1 hour). Sub-seconds are supported; the tests will fail if the value is too low for your machine.

require 'caches/ttl'

h = Caches::TTL.new(ttl: 0.01)
h[:a] = 'aardvark'
h[:a] #=> 'aardvark'
sleep(0.02)
h[:a] #=> nil

• With refresh: true, reading a value will reset its timer; otherwise, only writing will.
• With max_keys: 5, on insertion, it will evict the oldest item if necessary to keep from exceeding 5 keys.

keys and values work the same as for a hash. They do not check whether each key is current.

memoize method fetches a key if it exists and isn't expired; otherwise, it calculates the value using the block and saves it.

h = Caches::TTL.new(ttl: 5)

# Runs the block
h.memoize(:a) { |k| calculation_for(k) }

# Returns the previously-calculated value
h.memoize(:a) { |k| calculation_for(k) }

sleep(6)

# Runs the block
h.memoize(:a) { |k| calculation_for(k) }

LRU (Least Recently Used) remembers as many keys as you tell it to, dropping the least recently used key on each insert after its limit is reached. (Inserts and reads count as a usage; updates do not.)

require 'caches/lru'
h = Caches::LRU.new(max_keys: 3)
h[:a] = "aardvark"
h[:b] = "boron"
h[:c] = "cattail"
h[:d] = "dingo"
puts h[:a] # => nil

• #fetch works like Hash#fetch
• #memoize returns the key's value, if any; if not, it gets its return value from the block, and sets the key before it returns
• #stats tells how many cache hits and misses have occurred
• #clear resets all data and stats

Add this line to your application's Gemfile:

gem 'caches'

And then execute:

$ bundle

Or install it yourself as:

$ gem install caches

See CONTRIBUTING.md

• Thanks to FromAToB.com for giving me time to make this
• Thanks to satyap; the original specs for caches were adapted from volatile_hash.