[F#] Parallelize code in nested loops

Posted by Juliet on Stack Overflow See other posts from Stack Overflow or by Juliet
Published on 2009-01-05T03:52:55Z Indexed on 2010/05/10 23:44 UTC
Read the original article Hit count: 282

Filed under:
|
|

You always hear that functional code is inherently easier to parallelize than non-functional code, so I decided to write a function which does the following:

Given a input of strings, total up the number of unique characters for each string. So, given the input [ "aaaaa"; "bbb"; "ccccccc"; "abbbc" ], our method will returns a: 6; b: 6; c: 8.

Here's what I've written:

(* seq<#seq<char>> -> Map<char,int> *)
let wordFrequency input =
    input
    |> Seq.fold (fun acc text ->
        (* This inner loop can be processed on its own thread *)
        text
        |> Seq.choose (fun char -> if Char.IsLetter char then Some(char) else None)
        |> Seq.fold (fun (acc : Map<_,_>) item ->
            match acc.TryFind(item) with
            | Some(count) -> acc.Add(item, count + 1)
            | None -> acc.Add(item, 1))
            acc
        ) Map.empty

This code is ideally parallelizable, because each string in input can be processed on its own thread. Its not as straightforward as it looks since the innerloop adds items to a Map shared between all of the inputs.

I'd like the inner loop factored out into its own thread, and I don't want to use any mutable state. How would I re-write this function using an Async workflow?

© Stack Overflow or respective owner

Related posts about F#

Related posts about multithreading