Question

I have defined a record type for some client data in F# as follows:-

  type DataPoint = {
       date: string; 
       dr: string; 
       Group: string; 
       Product: string; 
       Book: int; 
       Revenue: int} with 
          static member fromFile file =
               file
               |> File.ReadLines
               |> Seq.skip 1 //skip the header
               |> Seq.map (fun s-> s.Split ',') // split each line into array
               |> Seq.map (fun a -> {date = string a.[0]; dr = string a.[1];
                              Group = string a.[2]; Product = string a.[3];
                                Book = int a.[4]; Revenue = int a.[5] });;  

    // creates a record for each line
    let pivot (file) = DataPoint.fromFile file
              |> ??????????

For the rows where date, dr, Group and Product are all equal, I want to then sum all of the Book and Revenue entries, producing a pivoted row. So some kind of if else statement should be fine. I suspect I need to start at the first data point and recursively add each matching row and then delete the matching row to avoid duplicates in the output.

Once I have done this I will be easily able to write these pivoted rows to another csv file.

Can anyone get me started?

Answer 1

Seq.groupBy and Seq.reduce are what you’re looking for:

let pivot file = 
    DataPoint.fromFile file
    |> Seq.groupBy (fun dp -> dp.date, dp.dr, dp.Group, dp.Product)
    |> Seq.map (snd >> Seq.reduce (fun acc dp -> 
                          { date = acc.date; dr = acc.dr; 
                            Group = acc.Group; Product = acc.Product;
                            Book = acc.Book + dp.Book; 
                            Revenue = acc.Revenue + dp.Revenue; }))