Question

I have a list of

type my_sum = {a : type_a} / {b : type_b}
mylist = [{field_only_in_a = "test"} : type_a,{haha=3 ; fsd=4} : type_b]

I want to do that :

result = List.find( a -> match a with 
                                      | {a = _} -> true
                                      | _ -> false
                                     end,
                                mylist)
 if Option.is_some(result) then
     Option.some(Option.get(result).field_only_in_a)
 else
     Option.none

Like you can see, after the find I’m sure to get something of type_a but on compilation time I get that :

    Record has type
{ a : type_a } / { b : type_b }  but field access expected it to have type
{ field_only_in_a: 'a; 'r.a }
Hint:
  You tried to access a sum type with several cases as a record.

How could I say to the compilator, that I have extract only one type of the sum type and that I have the good type to acces the record… ?

Answer 1

Well, you cannot really inform the compiler that only a subtype will exist in the list… but you can explicitly create a list with this subtype only. Actually, what you are looking for is List.find_map which find a first element matching a certain criteria and maps it (you use this mapping to project from my_sum to its case type_a). Below is a fully working code (compiles on its own):

type type_a = {fld_a : string}
type type_b = {fld_b : int}
type my_sum = {a : type_a} / {b : type_b}

mylist = [{a = {fld_a = "test"}}, {b = {fld_b = 10}}] : list(my_sum)

get_first_a(l : list(my_sum)) : option(type_a) =
  List.find_map(
    (el -> match el
           | ~{a} -> some(a)
           | _ -> none
    ), l)

_ = prerr("First a-typed element of {mylist} is {get_first_a(mylist)}\n")

If there was no List.find_map function in the stdlib there would still be a ton of ways to do it. Probably the simplest would be to use List.filter_map to obtain a list(type_a) and then get its head with List.head_opt.