Mentat
Provides a super simple cache with ttls.
Usage
A cache must be given a name when its started.
Mentat.start_link(name: :my_cache)
After its been started you can store and retrieve values:
Mentat.put(:my_cache, user_id, user)
user = Mentat.get(:my_cache, user_id)
TTLs
Both put
and fetch
operations allow you to specify the key's TTL. If no
TTL is provided then the TTL is set to :infinity
. TTL times are always
in milliseconds.
Mentat.put(:my_cache, :key, "value", [ttl: 5_000])
Mentat.fetch(:my_cache, :key, [ttl: 5_000], fn key ->
{:commit, "value"}
end)
Default TTLs
You can set a global TTL for all new keys
Limits
Mentat supports optional limits per cache.
Mentat.start_link(name: LimitedCache, limit: [size: 100])
When the limit is reached, the janitor will asynchronously reclaim a percentage of the keys.
Telemetry
Mentat publishes multiple telemetry events.
-
[:mentat, :get]
- executed after retrieving a value from the cache. Measurements are::status
- Can be either:hit
or:miss
depending on if the key was found in the cache.
Metadata are:
:key
- The key requested:cache
- The cache name
-
[:mentat, :put]
- executed when putting a key into the cache. No measurements are provided.Metadata are:
:key
- The key requested:cache
- The name of the cache
-
[:mentat, :janitor, :cleanup]
- executed after old keys are cleaned from the cache. Measurements are::duration
- the time it took to clean up the old keys. Time is in:native
units.total_removed_keys
- The count of keys removed from the cache.
Metadata are:
cache
- The cache name.
Contracts
Mentat supports Oath
contracts. This helps ensure that you're using Mentat correctly
and that Mentat is returning what you expect. You can enable contracts by setting
config :oath,
enable_contracts: true
And then recompiling Mentat in dev and test environments:
MIX_ENV=dev mix deps.compile mentat --force
MIX_ENV=test mix deps.compile mentat --force
Installation
def deps do
[
{:mentat, "~> 0.7"}
]
end
Should I use this?
There are (many) other caching libraries out there that provide many more features than Mentat. But, it turns out, I don't need most of those features. Mentat is intended to be very small while still providing the necessary components. The test suite is sparse, but we've been using this implementation in production for a while now so I feel pretty confident in it.