Benchmark of groupBy perf

code.fsx

open System.Collections.Generic
open System.Diagnostics

let groupBy keyMaker elements  =
    elements |> Seq.groupBy keyMaker |> Seq.length

let loopGroup keyMaker elements =
    let d = Dictionary<string, seq<'a>>()
    for p in elements do 
        let key = keyMaker p
        if d.ContainsKey(key) then d.[key] <- (Seq.append d.[key] [|p|])
        else d.Add(key, [|p|])
    d |> Seq.cast<KeyValuePair<_,_>> |> Seq.length


let timeit f = Stopwatch.StartNew() |> (fun sw -> (f (), sw.Elapsed.TotalSeconds))

let rnd = System.Random()
let data = Array.init 1000000 (fun _ -> rnd.Next(1000))

[1..5] |> List.iter (fun _ -> timeit (fun _ -> groupBy (fun n -> n.ToString()) data) |> printfn "groupBy %A" )
[1..5] |> List.iter (fun _ -> timeit (fun _ -> loopGroup (fun n -> n.ToString()) data) |> printfn "loopGroup %A" )


// F# 3.1 results
// groupBy (1000, 0.3984047)
// groupBy (1000, 0.346236)
// groupBy (1000, 0.3270108)
// groupBy (1000, 0.3249828)
// groupBy (1000, 0.3991331)
// loopGroup (1000, 0.8852538)
// loopGroup (1000, 0.8283647)
// loopGroup (1000, 0.9569686)
// loopGroup (1000, 0.8888318)
// loopGroup (1000, 0.9748136)

// F# 4.0 results
// groupBy (1000, 0.1995975)
// groupBy (1000, 0.1632741)
// groupBy (1000, 0.1873295)
// groupBy (1000, 0.1631585)
// groupBy (1000, 0.1628704)
// loopGroup (1000, 0.8894845)
// loopGroup (1000, 0.9352174)
// loopGroup (1000, 0.9339083)
// loopGroup (1000, 0.8045598)
// loopGroup (1000, 1.0230667)