Fortran Data Types

From WikiOD

Precision of floating point numbers[edit | edit source]

Floating point numbers of type real cannot have any real value. They can represent real numbers up to certain amount of decimal digits.

FORTRAN 77 guaranteed two floating point types and more recent standards guarantee at least two real types. Real variables may be declared as

real x
double precision y

x here is a real of default kind and y is a real of kind with greater decimal precision than x. In Fortran 2008, the decimal precision of y is at least 10 and its decimal exponent range at least 37.

 real, parameter             :: single = 1.12345678901234567890
 double precision, parameter :: double = 1.12345678901234567890d0

 print *, single
 print *, double



in common compilers using default configuration.

Notice the d0 in the double precision constant. A real literal containing d instead of e for denoting the exponent is used to indicate double precision.

! Default single precision constant
! Double precision constant

Fortran 90 introduced parameterized real types using kinds. The kind of a real type is an integer named constant or literal constant:

real(kind=real_kind) :: x

or just

real(real_kind) :: x

This statement declares x to be of type real with a certain precision depending on the value of real_kind.

Floating point literals can be declared with a specific kind using a suffix


The exact value of real_kind is not standardized and differs from compiler to compiler. To inquire the kind of any real variable or constant, the function kind() can be used:

print *, kind(1.0), kind(1.d0)

will typically print

4 8


1 2

depending on the compiler.

Kind numbers can be set in several ways:

Single (default) and double precision:

integer, parameter :: single_kind = kind(1.)
integer, parameter :: double_kind = kind(1.d0)

Using the intrinsic function selected_real_kind([p, r]) to specify required decimal precision. The returned kind has precision of at least p digits and allows exponent of at least r.

integer, parameter :: single_kind = selected_real_kind( p=6, r=37 )
integer, parameter :: double_kind = selected_real_kind( p=15, r=200 )

Starting with Fortran 2003, pre-defined constants are available through the intrinsic module ISO_C_Binding to ensure that real kinds are inter-operable with the types float, double or long_double of the accompanying C compiler:

use ISO_C_Binding

integer, parameter :: single_kind = c_float
integer, parameter :: double_kind = c_double
integer, parameter :: long_kind = c_long_double

Starting with Fortran 2008, pre-defined constants are available through the intrinsic module ISO_Fortran_env. These constants provide real kinds with certain storage size in bits

use ISO_Fortran_env

integer, parameter :: single_kind = real32
integer, parameter :: double_kind = real64
integer, parameter :: quadruple_kind = real128

If certain kind is not available in the compiler, the value returned by selected_real_kind() or the value of the integer constant is -1.

Intrinsic types[edit | edit source]

The following are data types intrinsic to Fortran:


integer, real and complex are numeric types.

character is a type used to store character strings.

logical is used to store binary values .true. or .false..

All numeric and logical intrinsic types are parametrized using kinds.


or just


where specific_kind is an integer named constant.

Character variables, as well as having a kind parameter, also have a length parameter:

character char

declares char to be a length-1 character variable of default kind, whereas

character(len=len) name

declares name to be a character variable of default kind and length len. The kind can also be specified

character(len=len, kind=specific_kind) name
character(kind=specific_kind) char

declares name to be a character of kind kind and length len. char is a length-1 character of kind kind.

Alternatively, the obsolete form for character declaration

character*len  name

may be seen in older code, declaring name to be of length len and default character kind.

Declaration of a variable of intrinsic type may be of the form above, but also may use the type(...) form:

integer i
real x
double precision y

is equivalent to (but greatly preferred over)

type(integer) i
type(real) x
type(double precision) y

Derived data types[edit | edit source]

Define a new type, mytype:

type :: mytype
  integer :: int
  real    :: float
end type mytype

Declare a variable of type mytype:

type(mytype) :: foo

The components of a derived type can be accessed with the % operator1:

foo%int = 4
foo%float = 3.142

A Fortran 2003 feature (not yet implemented by all compilers) allows to define parameterized data types:

type, public :: matrix(rows, cols, k)
  integer, len :: rows, cols
  integer, kind :: k = kind(0.0)
  real(kind = k), dimension(rows, cols) :: values
end type matrix

The derived type matrix has three type parameters which are listed in parentheses following the type name (they are rows, cols, and k). In the declaration of each type parameter it must be indicated whether they are kind (kind) or length (len) type parameters.

Kind type parameters, like those of the intrinsic types, must be constant expressions whereas length type parameters, like the length of an intrinsic character variable, may vary during execution.

Note that parameter k has a default value, so it may be provided or omitted when a variable of type matrix is declared, as follows

type (matrix (55, 65, kind=double)) :: b, c ! default parameter provided
type (matrix (rows=40, cols=50)     :: m    ! default parameter omitted

The name of a derived type may not be doubleprecision or the same as any of the intrinsic types.

  1. Many people wonder why Fortran uses % as the component-access operator, instead of the more common .. This is because . is already taken by the operator syntax, i.e. .not., .and., .my_own_operator..

Literal constants[edit | edit source]

Program units often make use of literal constants. These cover the obvious cases like

print *, "Hello", 1, 1.0

Except in one case, each literal constant is a scalar which has type, type parameters and value given by the syntax.

Integer literal constants are of the form

*1_1   ! For valid kind parameter 1
1_ik   ! For the named constant ik being a valid kind paramter

Real literal constants are of the form

1.0    ! Default real
1e0    ! Default real using exponent format
1._1   ! Real with kind parameter 1 (if valid)
1.0_sp ! Real with kind paramter named constant sp
1d0    ! Double precision real using exponent format
1e0_dp ! Real with kind named constant dp using exponent format

Complex literal constants are of the form

(1, 1.)       ! Complex with integer and real components, literal constants
(real, imag)  ! Complex with named constants as components

If the real and imaginary components are both integer, the complex literal constant is default complex, and the integer components are converted to default real. If one component is real, the kind parameter of the complex literal constant is that of the real (and the integer component is converted to that real kind). If both components are real the complex literal constant is of kind of the real with the greatest precision.

Logical literal constants are

.TRUE.     ! Default kind, with true value
.FALSE.    ! Default kind, with false value
.TRUE._1   ! Of kind 1 (if valid), with true value
.TRUE._lk  ! Of kind named constant lk (if valid), with true value

Character literal values differ slightly in concept, in that the kind specifier precedes the value

"Hello"       ! Character value of default kind
'Hello'       ! Character value of default kind
ck_"Hello"    ! Character value of kind ck
"'Bye"        ! Default kind character with a '
'''Bye'       ! Default kind character with a '
""            ! A zero-length character of default kind

As suggested above, character literal constants must be delimted by apostrophes or quotation marks, and the start and end marker must match. Literal apostrophes can be included by being within quotation mark delimiters or by appearing doubled. The same for quotation marks.

BOZ constants are distinct from the above, in that they specify only a value: they have no type or type parameter. A BOZ constant is a bit pattern and is specified as

B'00000'    ! A binary bit pattern
B"01010001" ! A binary bit pattern
O'012517'   ! An octal bit pattern
O"1267671"  ! An octal bit pattern
Z'0A4F'     ! A hexadecimal bit pattern
Z"FFFFFF"   ! A hexadecimal bit pattern

BOZ literal constants are limited in where they may appear: as constants in data statements and a selection of intrinsic procedures.

Assumed and deferred length type parameters[edit | edit source]

Variables of character type or of a derived type with length parameter may have the length parameter either assumed or deferred. The character variable name

character(len=len) name

is of length len throughout execution. Conversely the length specifier may be either

character(len=*) ...  ! Assumed length


character(len=:) ...  ! Deferred length

Assumed length variables assume their length from another entity.

In the function

function f(dummy_name)
  character(len=*) dummy_name
end function f

the dummy argument dummy_name has length that of the actual argument.

The named constant const_name in

character(len=*), parameter :: const_name = 'Name from which length is assumed'

has length given by the constant expression on the right-hand side.

Deferred length type parameters may vary during execution. A variable with deferred length must have either the allocatable or pointer attribute

character(len=:), allocatable :: alloc_name
character(len=:), pointer :: ptr_name

Such a variable's length may be set in any of the following ways

allocate(character(len=5) :: alloc_name, ptr_name)
alloc_name = 'Name'         ! Using allocation on intrinsic assignment
ptr_name => another_name    ! For given target

For derived types with length parameterization the syntax is similar

  type t(len)
    integer, len :: len
    integer i(len)
  end type t

  type(t(:)), allocatable :: t1
  type(t(5)) t2

  call sub(t2)
  allocate(type(t(5)) :: t1)


  subroutine sub(t2)
    type(t(*)), intent(out) :: t2
  end subroutine sub


Accessing character substrings[edit | edit source]

For the character entity

character(len=5), parameter :: greeting = "Hello"

a substring may be referenced with the syntax

greeting(2:4)  ! "ell"

To access a single letter it isn't sufficient to write

greeting(1)    ! This isn't the letter "H"


greeting(1:1)  ! This is "H"

For a character array

character(len=5), parameter :: greeting(2) = ["Hello", "Yo!  "]

we have substring access like

greeting(1)(2:4)  ! "ell"

but we cannot reference the non-contiguous characters

greeting(:)(2:4)  ! The parent string here is an array

We can even access substrings of literal constants


A portion of a character variable may also be defined by using a substring as a variable. For example

integer :: i=1
character :: filename = 'file000.txt'

filename(9:11) = 'dat'
write(filename(5:7), '(I3.3)') i

Accessing complex components[edit | edit source]

The complex entity

complex, parameter :: x = (1., 4.)

has real part 1. and complex part 4.. We can access these individual components as

real(x)  ! The real component
aimag(x) ! The complex component
x%re     ! The real component
y%im     ! The complex component

The x%.. form is new to Fortran 2008 and not widely supported in compilers. This form, however, may be used to directly set the individual components of a complex variable

complex y
y%re = 0.
y%im = 1.

Declaration and attributes[edit | edit source]

Throughout the topics and examples here we'll see many declarations of variables, functions and so on.

As well as their name, data objects may have attributes. Covered in this topic are declaration statements like

integer, parameter :: single_kind = kind(1.)

which gives the object single_kind the parameter attribute (making it a named constant).

There are many other attributes, like

  • target
  • pointer
  • optional
  • save

Attributes may be specified with so-called attribute specification statements

integer i    ! i is an integer (of default kind)...
pointer i    ! ... with the POINTER attribute...
optional i   ! ... and the OPTIONAL attribute

However, it is generally regarded to be better to avoid using these attribute specification statements. For clarity the attributes may be specified as part of a single declaration

integer, pointer, optional :: i

This also reduces the temptation to use implicit typing.

In most cases in this Fortran documentation this single declaration statement is preferred.