FileCache is a simple building block for the Scala programming language, managing a directory with cache files. It is (C)opyright 2013–2020 by Hanns Holger Rutz. All rights reserved. This project is released under the GNU Lesser General Public License v2.1+ and comes with absolutely no warranties. To contact the author, send an e-mail to contact at sciss.de.

To link to this library, use either of the following artifacts:

"de.sciss" %% "filecache-mutable" % v
"de.sciss" %% "filecache-txn"     % v

The current version v is "1.1.2". The -mutable variant provides a (thread-safe) mutable cache object, whereas the -txn variant uses the scala-stm software transactional memory.

This project builds with sbt against Scala 2.13, 2.12, 3. The last version to support Scala 2.11 was v0.5.1.

The project is documented through scaladoc run sbt doc to create the API docs.

Here is an example of the mutable package, the transactional package works analogously. A cache is represented by the MutableProducer[Key, Value] type. Therefore, a resource (the value) is identified by and can be looked up with a key. A producer is created by calling apply on the companion object, requiring implicit values for key and value serializers (de.sciss.serial.ConstFormat). Serializers are readily available for primitive types and some combinators (e.g. Tuple2 or IndexedSeq).

For example, let us assume the key type is String, and the value type is IndexedSeq[Int]. First of all, we need to create a configuration for the cache. A mutable ConfigBuilder is obtained by calling Config[Key, Value](). It will be implicitly frozen as a Config object when the builder is passed into the producer constructor. We'll go with the defaults except changing the cache capacity to hold at maximum two entries (the default is unlimited), and we'll assign an eviction function that prints to the console when a cache entry is removed from disk.

    import de.sciss.filecache._

    type Vec[+A] = collection.immutable.IndexedSeq[A]
    val c = Config[String, Vec[Int]]()
    c.capacity = Limit(count = 2)
    c.evict = { (key, value) => println(s"Evicted key $key") }
    val p = MutableProducer(c)

The producer is like an exclusive handle to resources. A resource can be obtained with method

    def acquire(key: A)(source: => B): Future[B]

Where the source thunk is executed only if the resource for key was not found in the cache. When the resource is not used any longer, it must be released using

    def release(key: A): Unit

If your application wants to be able to have multiple handles on a resource, a second structure MutableConsumer can be used which simply wraps a producer with an internal use count. Here the acquirement is a more simple method def acquire(key: A): Future[B] which may be called multiple times. Each acquirement should be matched with an eventual release call.

For our example, we will just use the exclusive producer. Cache production happens inside a Future, so if we want to see the result, we can use the future's foreach method. A suitable ExecutionContext can be found by importing p.executionContext (that context was defined in the original configuration).

    import p.executionContext
    import scala.concurrent._
    import duration.Duration

    val foo1 = p.acquire("foo") { println("Producing..."); Vector(1, 2, 3) }
    Await.result(foo1, Duration.Inf)  // Vector(1, 2, 3)

Let's release and re-acquire that resource. It should be cached:

    p.release("foo")
    val foo2 = p.acquire("foo")(???)
    Await.result(foo2, Duration.Inf)  // Vector(1, 2, 3)

If we release "foo" and produce new entries in the cache, we can witness the eviction:

    p.release("foo")
    val bar = p.acquire("bar") { Vector(4, 5, 6) }
    val baz = p.acquire("baz") { Vector(7, 8, 9) }  // causes eviction of "foo"

Note that an entry is never evicted while acquired.

Some more notes:

• the capacity Limit has two conditions, a count for the maximum number of entries and a space for the maximum total size in bytes. A value of -1 for each of these (default) means no limit. Both limits work together, so if both count and space are specified, the capacity will be constrained to satisfy these two properties.
• if using a space limit, the space function in the configuration must be provided. This function is given a key-value pair and must return the resource size in bytes. The reason this is not automatically equated with the size of the serialized value is that often a resource extends beyond the directly given value. For example, if the cache is used to produce waveform overviews for a sound file, the value might just point to the waveform file. But the size of the resource is then determined by the size of that waveform file, not the cache entry which is just the path name to that file.
• a producer can be disposed calling the dispose method. This will release all currently acquired keys.
• the cache directory is specified in the configuration using the folder variable. A second variable extension can be used to specify the file extension.
• the current cache size is available through the usage method on a producer
• cache values may become invalid for some reason. As an example, imagine again the sound file waveform cache. If the sound file whose waveform is produced has changed, then obviously the waveform overview must be re-computed. If that case may occur in your scenario, the function accept in the configuration should be specified. It is given a key-value pair and must return a boolean. A return value true indicates that the cache entry is still valid, a value of false will cause the re-run of the producer function and overwrite the cache entry with the new result. In the sound file example, we may store the last-modified time-stamp of the sound file in the cache value, so we can quickly determine whether the file has changed. In this scenario it is also important to use the configuration's evict function to delete the additional resources (the waveform file).