I\'m trying to learn F# by rewriting some C# algorithms I have into idiomatic F#.
One of the first functions I\'m trying to rewrite is a batchesOf where:
This can be done without recursion if you want
[0..20]
|> Seq.mapi (fun i elem -> (i/size),elem)
|> Seq.groupBy (fun (a,_) -> a)
|> Seq.map (fun (_,se) -> se |> Seq.map (snd));;
val it : seq<seq<int>> =
seq
[seq [0; 1; 2; 3; ...]; seq [5; 6; 7; 8; ...]; seq [10; 11; 12; 13; ...];
seq [15; 16; 17; 18; ...]; ...]
Depending on how you think this may be easier to understand. Tomas' solution is probably more idiomatic F# though
I found this to be a quite terse solution:
let partition n (stream:seq<_>) = seq {
let enum = stream.GetEnumerator()
let rec collect n partition =
if n = 1 || not (enum.MoveNext()) then
partition
else
collect (n-1) (partition @ [enum.Current])
while enum.MoveNext() do
yield collect n [enum.Current]
}
It works on a sequence and produces a sequence. The output sequence consists of lists of n elements from the input sequence.
You can solve your task with analog of Clojure partition library function below:
let partition n step coll =
let rec split ss =
seq {
yield(ss |> Seq.truncate n)
if Seq.length(ss |> Seq.truncate (step+1)) > step then
yield! split <| (ss |> Seq.skip step)
}
split coll
Being used as partition 5 5
it will provide you with sought batchesOf 5
functionality:
[1..17] |> partition 5 5;;
val it : seq<seq<int>> =
seq
[seq [1; 2; 3; 4; ...]; seq [6; 7; 8; 9; ...]; seq [11; 12; 13; 14; ...];
seq [16; 17]]
As a premium by playing with n
and step
you can use it for slicing overlapping batches aka sliding windows, and even apply to infinite sequences, like below:
Seq.initInfinite(fun x -> x) |> partition 4 1;;
val it : seq<seq<int>> =
seq
[seq [0; 1; 2; 3]; seq [1; 2; 3; 4]; seq [2; 3; 4; 5]; seq [3; 4; 5; 6];
...]
Consider it as a prototype only as it does many redundant evaluations on the source sequence and not likely fit for production purposes.
Here's a simple implementation for sequences:
let chunks size (items:seq<_>) =
use e = items.GetEnumerator()
let rec loop i acc =
seq {
if i = size then
yield (List.rev acc)
yield! loop 0 []
elif e.MoveNext() then
yield! loop (i+1) (e.Current::acc)
else
yield (List.rev acc)
}
if size = 0 then invalidArg "size" "must be greater than zero"
if Seq.isEmpty items then Seq.empty else loop 0 []
let s = Seq.init 10 id
chunks 3 s
//output: seq [[0; 1; 2]; [3; 4; 5]; [6; 7; 8]; [9]]