A Fortran function/subroutine that could return either a real, an integer or a string.

Question

I would like to know how to create a function that either returns a real, an integer or a string.

For example, the call would be write(*,*)dt%get() where

Answer 1

That is simply impossible in Fortran.

You can use a generic interface with different specific functions, but these functions must have arguments of different types (see how several intrinsic functions, like transfer() use a mold argument). This is called the TKR (type, kind, rank) resolution. Generic functions cannot be distinguished based on a value of an argument.

type :: dt
    real(dp) :: realValue
    integer :: integerValue
    character(*) :: strValue             !!!! <= THIS IS WRONG !!!!
    logical :: isReal, isInteger, isStr
contains
    generic :: get => getReal, getInteger, getStr
    procedure :: getReal, getInteger, getStr
end type

function getReal(self, mold)
  class(dt) :: self
  real, intent(in) :: mold
end function

function getInteger(self, mold)
  class(dt) :: self
  integer, intent(in) :: mold
end function

function getString(self, mold)
  class(dt) :: self
  character(*), intent(in) :: mold
end function

As you see, you have to know the correct type when calling get(). You call it like

real_variable = object%get(1.0)
integer_variable = object%get(1)

Be also careful about strings of different lengths. I marked it above. You probably want character(:), allocatable.

---

You can make also function which returns a generic container and then extract the value from the container. The extraction could even be done directly using an overloaded assignment for the container.

You could also just return an unlimited polymorphic variable (class(*)).