Skip to content

Latest commit

 

History

History
154 lines (115 loc) · 6.34 KB

README.md

File metadata and controls

154 lines (115 loc) · 6.34 KB

FSharpx

FSharpx is a library for the .NET platform implementing general functional constructs on top of the F# core library. Its main target is F# but it aims to be compatible with all .NET languages wherever possible.

It currently implements:

  • Several standard monads: State, Reader, Writer, Either, Continuation, Distribution
  • Validation applicative functor
  • General functions like flip
  • Additional functions around collections
  • Functions to make C# - F# interop easier
  • Async and Observable extensions

F# Async Extensions

This library implements various extensions for asynchronous programming using F# asynchronous workflows and F# agents (the MailboxProcessor type in the standard F# library). It defines asynchronous sequences that represent asynchronous operations returning multiple values (such as reading data from a stream in chunks), several reusable F# agents and numerous extensions.

  • Samples that demonstrate how to use most of the extensions can be found in the samples directory

Asynchronous sequences

Asynchronous sequences can be used to work with asynchronous computations that return multiple results. A value of type AsyncSeq<'T> can be started (just like an asynchronous workflow) and it eventually returns. The result is either a special value representing the end of the sequence or a value of type 'T (head) together with the rest of the asynchronous sequence (tail) of type AsyncSeq<'T>.

Unlike IObservable<'T>, asynchronous sequences are not push-based. The code that generates the next value of the asynchronous sequence starts only after previous elements have been processed. This makes it possible to easily write computations that return results as long as some component is using them.

However, IObservable<'T> values can be converted to asynchronous sequences. The AsyncSeq.ofObservable combinator creates an asynchronous sequence that discards values produced by the observable while the asynchronous sequence was blocked. The AsyncSeq.ofObservableBuffered combinator stores all produced values in an unbounded buffer and returns the values from the buffer as soon as the user of asynchronous sequence requestst the next element.

The library defines an F# computation expression for workfing with asynchronous sequences. For example, sequence that emits numbers in 1 second intervals can be defined as follows:

let rec numbers n = asyncSeq {
  yield n
  do! Async.Sleep(1000)
  yield! numbers (n + 1) }

Asynchronous workflows and asynchronous sequences can use the for construct to iterate over all elements of an asynchronous sequence. For example:

let rec evenNumbers = asyncSeq {
  for n in numbers 0 do
    if n%2=0 then yield n }

The library also provides numerous combinators (similar to functions from the Seq module). The result of operations that aggregate values of an asynchronous sequence is an asynchronous workflow that returns a single value:

let rec sumTenEvenSquares = 
  numbers 0
  |> AsyncSeq.filter (fun n -> n%2 = 0)
  |> AsyncSeq.map (fun n -> n*n)
  |> AsyncSeq.fold (+) 0

let n = 
  sumTenEvenSquares 
  |> Async.RunSynchronously

For some examples that use (earlier versions) of asynchronous sequences, see also the following two F# snippets: first and second.

Reusable agents

The library implements several reusable agents for building concurrent applications:

  • Agent is a simple type aliast for MailboxProcessor that is more convenient to use

  • AutoCancelAgent wraps the standard F# agent and adds support for stopping of the agent's body using the IDisposable interface (the type automatically creates a cancellation token, uses it to start the underlying agent and cancels it when the agent is disposed). For example, see this F# snippet.

  • BatchProcessingAgent can be used to implement batch processing. It creates groups of messages (added using the Enqueue method) and emits them using the BatchProduced event. A group is produced when it reaches the maximal size or after the timeout elapses.

  • BlockingQueueAgent implements an asynchronous queue with blocking put and blocking get operations. It can be used to implement the producer-consumer concurrent pattern. The constructor of the agent takes the maximal size of the buffer.

Observable extensions

The library implements extensions for using IObservable<'T> type from F# asynchronous workflows. An overloaded extension method Async.AwaitObservable can be used to wait for an occurrence of an event (or other observable action):

let counter n = async {
  printfn "Counting: %d" n
  let! _ = form.MouseDown |> Async.AwaitObservable
  return! counter (n + 1) }

Overloaded version of the method allows waiting for the first of multiple events. The method asynchronously returns Choice<'T1, 'T2> value that can be used to determine which of the events has occurred.

For examples using this method see Chapter 16 of Real World Functional Programming (some examples are available in a free excerpt from the chapter). The Async.AwaitObservable method should be used instead of Async.AwaitEvent to avoid memory leaks (see also related StackOverflow discussion)

Type Providers

FSharpx.TypeProviders is a library for the .NET platform implementing common type providers on top of the FSharpx.Core. main target is F# 3.0 but it aims to be compatible with all .NET languages wherever possible.

It currently implements type safe variants of:

  • Regex
  • FileSystem
  • Csv
  • Excel
  • JSON
  • XML
  • Registry
  • XAML
  • AppSettings.

Building the project

Read the wiki page about Build