Like several other languages, Fortran 90 and newer supports the ability to select the appropriate routine from a list of routines based on the arguments passed. This selection is done at compile time and is thus unencumbered by run-time performance penalties. This feature is accessed by use of modules and the interface block.
In the following example, a module is specified with contains a single logical function which can handle arguments of various types.
module extension_mod implicit none private interface f module procedure f_i module procedure f_r module procedure f_z end interface public::f contains function f_i(x) result(y) integer,intent(in)::x integer::y y = x**2-1 end function f_i function f_r(x) result(y) real,intent(in)::x real::y y = x**2-1.0 end function f_r function f_z(x) result(y) complex,intent(in)::x complex::y y = x**2-1.0 end function f_z end module extension_mod
A program which uses this module now has access to a single logical function f which accepts arguments which are of the integer, real, or complex data types. The return type of the function is the same as the input type. In this way the routine is much like many of the intrinsic functions defined as part of the Fortran standard. An example program is given below:
program main use extension_mod implicit none integer::xi,yi real::xr,yr complex::xz,yz xi = 2 xr = 2.0 xz = 2.0 yi = f(xi) yr = f(xr) yz = f(xz) end program main
One can extend intrinsic functions. This is similar to overload operators.
Here we will demonstrate this by extending the
The intrinsic function is not implemented for arguments of integer type.
This is because there is no clear idea how to define the result of non integer type (e.g. , but how to define ).
We implement a method here where the result is always the nearest integer.
module sqrt_int_module implicit none private ! use intrinsic sqrt for data types which are not overloaded intrinsic :: sqrt ! extend sqrt for integers public :: sqrt interface sqrt procedure sqrt_int end interface sqrt contains pure function sqrt_int( int ) result( res ) integer, intent(in) :: int integer :: res res = nint( sqrt( real(int) ) ) end function sqrt_int end module sqrt_int_module program main use sqrt_int_module implicit none integer :: i do i = 1, 7 print *, "i:", i, "sqrt(i):", sqrt(i) end do end program main
Derived Data Types
Fortran 90 and newer supports the creation of new data types which are composites of existing types. In some ways this is similar to an array, but the components need not be all of the same type and they are referenced by name, not index. Such data types must be declared before variables of that type, and the declaration must be in scope to be used. An example of a simple 2d vector type is given below.
type::vec_t real::x,y end type
Variables of this type can be declared much like any other variable, including variable characteristics such are pointer or dimension.
Using derived data types, the Fortran language can be extended to represent more diverse types of data than those represented by the primitive types.
Operators can be overloaded so that derived data types support the standard operations, opening the possibility of extending the Fortran language to have new types which behave nearly like the native types.
The assignment operator = can be overloaded. We will demonstrate this by the following example. Here, we define how the assignment of a logical type on the left and an integer on the right should be performed.
module overload_assignment implicit none private public :: assignment(=) interface assignment(=) procedure logical_gets_integer end interface assignment(=) contains subroutine logical_gets_integer( log, i ) logical, intent(out) :: log integer, intent(in) :: i if ( i==0 ) then log = .true. else log = .false. end if end subroutine logical_gets_integer end module overload_assignment program main use overload_assignment implicit none logical :: log log = 0 print *, "log=0:", log ! yields: T log = 1 print *, "log=1:", log ! yields: F end program main
One can overload intrinsic operators, such as
In the following example we will overload the
* operator to work as the logical
module overload_operator implicit none private public :: operator(*) interface operator(*) procedure logical_and end interface operator(*) contains pure function logical_and( log1, log2 ) result( log_res) logical, intent(in) :: log1, log2 logical :: log_res if ( log1 .and. log2 ) then log_res = .true. else log_res = .false. end if end function logical_and end module overload_operator program main use overload_operator implicit none logical, parameter :: T = .true., F = .false. print *, "T*T:", T*T ! yields: T print *, "T*F:", T*F ! yields: F print *, "F*T:", F*T ! yields: F print *, "F*F:", F*F ! yields: F end program main
One can create newly self-created operators.
We demonstrate this by the following example: We create an operator
.even. <int> which outputs a
logical if the given
integer is even.
module new_operator implicit none private public :: operator(.even.) interface operator(.even.) procedure check_even end interface operator(.even.) contains pure function check_even( int ) result( log_res ) integer, intent(in) :: int logical :: log_res if ( modulo( int, 2 )==0 ) then log_res = .true. else log_res = .false. end if end function check_even end module new_operator program main use new_operator implicit none integer :: i do i = 1, 6 print *, "i:", i, "even?", .even. i end do end program main