Question

Can I have a class with one member either be float array or array of float array like

type testArr(data: float[],...) =
  member x.data = data
  member x.others = ...
  new( data: float[][],...) = ??? 

I tried to define
type Value = Value1D of float[] | Value2D of float[][], then define type testArr(data: Value,... ), but I then need to rewrite a lot of code for type Value to repeat the array type methods like .Item etc. Alternatively, How do I define type Value inherited from 'a[] but limit the elment type 'a to be float or float[] only?

Answer 1

I’m not entirely sure why would you need this, but you can define a TestArr class with private constructor and add two static methods to create the two types (float[] and float[][]) that you want to allow:

type TestArr<'T> private(data:'T[]) = 
  member x.Data = data 
  member x.Others = 42
  static member Array1D(data:float[]) = TestArr(data)
  static member Array2D(data:float[][]) = TestArr(data)

Using an array of arrays always, but with just a single element for the 1D scenario as suggested by Gabe in a comment sounds like a good and simpler alternative.