Skip to content

hpeebles/CacheMeIfYouCan

Repository files navigation

CacheMeIfYouCan

Build Status NuGet version (CacheMeIfYouCan)

Features

Installing via NuGet

Install-Package CacheMeIfYouCan

Cached Functions

A cached function is a new version of an input function, which when invoked first tries to find the return value in a cache and only calls into the original function if the value was not found.

Cached functions come with a wide variety of configuration options and the ability to use any local cache (such as System.Runtime.Caching.MemoryCache) or distributed cache (such as Redis) or a 2-tier caching strategy that uses a local cache backed by a distributed cache.

Func<int, Task<int>> originalFunction = ...

var cachedFunction = CachedFunctionFactory
    .ConfigureFor(originalFunction)
    .WithLocalCache(new DictionaryCache<int, int>())
    .WithDistributedCache(new RedisCache<int, int>(...))
    .WithTimeToLive(TimeSpan.FromHours(1))
    .OnResult(r => logSuccess(r), ex => logException(ex))
    .Build();

It is possible to create cached functions where the original function has up to 8 parameters (+ an optional cancellation token) and the cache key can be any function of the input parameters. The original function can be synchronous, asynchronous, or return a ValueTask (the underlying implementation uses ValueTasks). The resulting cached function will always have exactly the same input parameters and return type as the original function.

Func<string, bool, int, CancellationToken, Task<int>> originalFunction = ...

var cachedFunction = CachedFunctionFactory
    .ConfigureFor(originalFunction)
    .WithCacheKey((param1, param2, param3) => param3)
    .WithLocalCache(new DictionaryCache<int, int>())
    .WithTimeToLive(TimeSpan.FromHours(1))
    .Build();

If your function takes a list of keys as the last parameter and returns a dictionary of values where each key in the returned dictionary corresponds to a key in the input list then you can use the WithEnumerableKeys configuration option. This will cause each key in the input list to be cached separately. When the cached function is called, all the keys are looked up in the cache and the underlying function is only called for those keys that were not found.

Note: The WithEnumerableKeys option is available whenever the last parameter is an IEnumerable<TKey> and the return type is an IEnumerable<KeyValuePair<TKey, TValue>>.

Func<List<int>, Dictionary<int, string>> originalFunction = null;

var cachedFunction = CachedFunctionFactory
    .ConfigureFor(originalFunction)
    .WithEnumerableKeys<List<int>, Dictionary<int, string>, int, string>() // The generic arguments are each of the parameter types, the return type (only the inner type if Task<T> or ValueTask<T>), then the type of the keys, then the type of the values
    .WithLocalCache(new DictionaryCache<int, string>())
    .WithTimeToLive(TimeSpan.FromHours(1))
    .Build();

You can use the WithEnumerableKeys configuration option for functions with multiple input parameters. In these cases the WithOuterCacheKey option is available. Use WithOuterCacheKey to make each key in the cache be made up of 2 parts, for each value in the last input parameter an 'inner key' is created, and then the 'outer key' is made by passing the earlier parameters into the outer key generator function. In order for there to be a cache hit both the inner key and the outer key need to be matched.

Func<string, string, List<int>, Dictionary<int, string>> originalFunction = null;

var cachedFunction = CachedFunctionFactory
    .ConfigureFor(originalFunction)
    .WithEnumerableKeys<string, string, List<int>, Dictionary<int, string>, int, string>()
    .WithOuterCacheKey((param1, param2) => $"{param1}_{param2}")
    .WithLocalCache(new DictionaryCache<string, int, string>())
    .WithTimeToLive(TimeSpan.FromHours(1))
    .Build();

Cached Interfaces

A cached interface is an implementation of an interface where each function is an individually configured cached function.

public interface IMyCoolInterface
{
    string GetOne(int id);
    Dictionary<int, string> GetMany(List<int> ids);
}
IMyCoolInterface originalImpl = ...;

var cache = new DictionaryCache<int, string>();

IMyCoolInterface cachedImpl = CachedInterfaceFactory.For(originalImpl)
    .Configure<int, string>(x => x.GetOne, c => c
        .WithLocalCache(cache)
        .WithTimeToLive(TimeSpan.FromHours(1)))
    .Configure<List<int>, Dictionary<int, string>>(x => x.GetMany, c => c
        .WithEnumerableKeys<List<int>, Dictionary<int, string>, int, string>()
        .WithLocalCache(cache) // Note that I am sharing the same cache between the 2 methods
        .WithTimeToLive(TimeSpan.FromHours(1)))
    .Build();

Each call to Configure takes 2 input parameters, the first is an expression which (along with the generic type arguments) is used to identify the method to be configured, the second is the configuration function which is used to configure the cached function.

Cached Objects

A cached object is an object which exposes a single value which can be refreshed or updated in the background. Accessing the value is fast and reliable since any failures that happen while trying to update the value do not propagate to the threads accessing the value.

Func<Task<Configuration>> getConfigurationFunc = ...

var cachedConfiguration = CachedObjectFactory
    .ConfigureFor(getConfigurationFunc)
    .WithRefreshInterval(TimeSpan.FromMinutes(10))
    .WithJitter(10) // Use WithJitter if you want the refresh intervals to flucuate slightly
    .Build();

cachedConfiguration.Initialize(); // The first access will trigger a call to initialize if not already initialized

// To access the value
var configuration = cachedConfiguration.Value;

Benchmarks

Cached function with single key using DictionaryCache -

Method Mean Error StdDev Gen 0 Gen 1 Gen 2 Allocated
CacheHit 90.10 ns 0.269 ns 0.252 ns - - - -
CacheMiss 180.04 ns 1.192 ns 1.115 ns - - - -

Cached function with enumerable keys using DictionaryCache -

Method Mean Error StdDev Gen 0 Gen 1 Gen 2 Allocated
OneHit 410.9 ns 3.19 ns 2.98 ns 0.0391 - - 248 B
OneMiss 683.2 ns 1.96 ns 1.63 ns 0.0439 - - 280 B
OneHitAndOneMiss 810.8 ns 4.53 ns 4.24 ns 0.0734 - - 464 B
OneHundredHits 2,704.3 ns 20.80 ns 19.45 ns 0.3700 0.0038 - 2328 B
OneHundredMisses 10,784.1 ns 42.62 ns 37.78 ns 0.3662 - - 2360 B
OneHundredHitsAndOneHundredMisses 15,643.8 ns 99.99 ns 88.64 ns 1.4954 0.0610 - 9456 B

About

High performance, highly configurable caching library

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages