Lightweight caching solution in Scala?

Question

I just wonder if there is any caching solution available in Scala. I'm looking for something like it is provided by Guava in Java.

Should I use Guava too in Scala? Is there a wrapper / pimp in Scalaz or something similar? Any alternative more appropriate for Scala devs?

What Guava provides:

LoadingCache<Key, Graph> CACHE= CacheBuilder.newBuilder()
       .maximumSize(1000)
       .expireAfterWrite(10, TimeUnit.MINUTES)
       .removalListener(MY_LISTENER)
       .build(
           new CacheLoader<Key, Graph>() {
             public Graph load(Key key) throws AnyException {
               return createExpensiveGraph(key);
             }
           });

Supplier<Animal> singleAnimalCache = Suppliers.memoizeWithExpiration(animalFromDbSupplier(), 365, TimeUnit.DAYS);

I need some basic cache management like in Guava.

Answer 1

Just adding an answer to plug my own project, but I recommend ScalaCache.

• Supports Guava, Ehcache, Memcached and Redis (or you can plugin your own implementation if you like)
• Simple, idiomatic Scala API
• Support for per-element Time To Live (even for Guava, which doesn't provide that out of the box)
• Support for auto-generation of cache keys using macro magic

https://github.com/cb372/scalacache

Answer 2

Using guava caching is straightforward in Scala.

import com.google.common.base._
import com.google.common.cache._

object Simple_Guava_Caches_in_Scala {

   def main(args: Array[String]): Unit = {

      // Simple scala guava cache
      val simpleCache1 =
         CacheBuilder
           .newBuilder()
           .build(new CacheLoader[String, String] {
              def load(key: String): String = {
                 println(s"Simple scala guava cache, heavy work calculating $key")
                 s"It's me $key"
              }
           })
      println(simpleCache1.get("1"))
      println(simpleCache1.get("1"))
      println(simpleCache1.get("2"))
      println(simpleCache1.get("2"))
      println(simpleCache1.get("2"))


      // Simple scala guava supplier cache / factory
      println()
      val supplier_cache: Supplier[String] = Suppliers.memoize(
         () => {
            println("Simple scala guava supplier cache / factory, heavy work creating singleton:")
            "It's me"
         }
      )
      println(supplier_cache.get)
      println(supplier_cache.get)
      println(supplier_cache.get)
   }
}

This is producing

Simple scala guava cache, heavy work calculating 1
It's me 1
It's me 1
Simple scala guava cache, heavy work calculating 2
It's me 2
It's me 2
It's me 2

Simple scala guava supplier cache / factory, heavy work creating singleton:
It's me
It's me
It's me

I've added this to build.sbt:

libraryDependencies += "com.github.cb372" %% "scalacache-guava" % "0.26.0"

Would be interesting to transform the coding above to cb372's scalacache-guava. Could be even simpler/more standardized.

Answer 3

We had the same requirements and ended up building wrappers around Guava. We recently open-sourced parts of the library called Mango. If you don’t mind the extra dependency you can use it like

import org.feijoas.mango.common.cache._
import org.feijoas.mango.common.base.Suppliers._

val MY_LISTENER = (remNot: RemovalNotification[Key, Graph]) => ()
// > MY_LISTENER : RemovalNotification[Key,Graph] => Unit = <function1>

val CACHE = CacheBuilder.newBuilder()
  .maximumSize(1000)
  .expireAfterWrite(10, TimeUnit.MINUTES)
  .removalListener(MY_LISTENER)
  .build { (key: Key) => new Graph() }
// > CACHE : LoadingCache[Key,Graph] = <function1>

val animalFromDbSupplier = () => {
  // load from db
  new Animal()
}
// > animalFromDbSupplier  : () => Animal = <function0>

val singleAnimalCache = memoizeWithExpiration(animalFromDbSupplier, 365, TimeUnit.DAYS)
// > singleAnimalCache  : () => Animal = Suppliers.memoizeWithExpiration(<function0>, 365, DAYS)

Answer 4

Is there a wrapper / pimp in Scalaz or something similar?

In Scalaz 7 there's Memo, which I covered a bit in learning Scalaz day 16.

It's the first thing Adam Rosien covered in scalaz "For the Rest of Us" talk, so check that out too. He's using Scalaz 6.