ADT design choices

Question

How do you go about deciding which of the following two representations (in F# syntax) is the right choice in a particular situation?

type Choice = A of string | B of string

Or:

type ChoiceKind = A | B
type Choice = { Kind: ChoiceKind; Value: string }

I'm deliberately avoiding giving a more specific example, as either approach can "feel more natural" in a particular scenario; I'm interested in the general reasoning that goes into the design choice.

Answer 1

In my opinion the former makes more sense if there's not much you can sensibly do without knowing whether you have an A or B. If you wanted to unconditionally get the string with the former declaration, you'd end up with redundant cases in the pattern match. Of course, you could easily write a function to do it.

I would imagine the latter is easier for other .NET languages to work with since they lack pattern matching, but you can fake it well enough in C# using named arguments and lambda expressions.

Disclaimer: I've never coded in F# but I'm familiar with Standard ML, which is related (through OCaml).