Variable Declarations

Like most compiled languages, Fortran is statically typed . All variables must be declared to be of a specific type before they can be used. A variable’s type cannot be changed once it is declared.

Fortran is (nearly) strongly typed. Mixed-mode expressions are limited and most conversions must be explicit.

Unlike most languages, Fortran is not case sensitive. Variables Mean, mean, and even mEan are the same to the compiler.

Variable names may consist of alphanumeric (letter or digit) characters, plus underscores. No other characters, including spaces, are permitted. The first character must be an alphabetical character. The maximum length of a variable name for modern Fortran as of Fortran 95 is 31 characters. The 6-character limit of Fortran 77 is long gone. Some compilers permit up to 127 characters as an extension, though excessively long variable names is not a good programming practice.

A good descriptive variable name often consists of multiple words or parts of words. Since Fortran is not case-sensitive, underscores can be used to separate the components.

is_valid
start_date
num_species

Separation through capitalization is possible with the understanding that different variables cannot be distinguished by “camel case.”

subroutine BioGeoChem
type myType

Variables are declared by indicating the type followed by a comma-separated list of variables. In older code no separator was used.

INTEGER i, j, k

In newer code, use the double colon to separate the type from the variable list

INTEGER  :: i, j, k

If there are other attributes on the line the :: will be required .

It is not necessary to write keywords, or any source at all, in all capital letters, but they may be written in capitals here for clarity.

Declarations by Type

Other types may be specified through [KIND].

Implicit and Explicit Typing

For historical reasons, Fortran used implicit typing for numerical types. Any variable starting with the letters A-H or O-Z were floating point. Variables beginning with the letters I-N were integers. Note that IN are the first two letters of the word “integer.” That is a longstanding mathematical tradition and Fortran was developed to translate mathematical notation (FORmula TRANslation).

Older code often changes the default float to double:

IMPLICIT DOUBLE PRECISION(a-h,o-z)

However, in modern usage, all variables should be explicitly typed. This will enable the compiler to catch typographical errors. If implicit typing is used, a new variable would not need to be declared and would assume the type based on its name, so a misspelling of an existing variable would create a different variable. Bugs like this can be difficult to track down.

The statement

IMPLICIT NONE

negates implicit typing. It must be the first line after a unit declaration unless a USE is present.

Example

PROGRAM simple
IMPLICIT NONE
INTEGER              ::   I, J
REAL                 ::   R, S, T
DOUBLE PRECISION     ::   D
DOUBLE COMPLEX       ::   Z
LOGICAL              ::   FLAG
CHARACTER (len=20)   ::   C

Line up declarations neatly.

Initializing at Compile Time

Variables can be declared and initialized at the same time:

real  :: x=1.e-8, y=42.

When variables are initialized in this manner it happens only once , at compile time. If this takes place in a subprogram it will not happen again upon repeated invocations.

It is equivalent to the older DATA statement:

DATA x,y/1.e-8,42./

In Fortran 2003 it became possible to initialize using intrinsic functions:

real  :: pi = 4.0*atan(1.0)

Example

Start your choice of IDE or editor. Type

program first
! My first program
! Author:  Your Name
  implicit none
  real     ::x,y
  integer  ::i,j=11
     x=1.0
     y=2.0
     i=j+2
     print *, "Reals are ",x,y
     print *, "Integers are ",i,j
end program

PARAMETER

In compiled languages, programmers can declare a variable to have a fixed value that cannot be changed. In Fortran this is indicated by the PARAMETER attribute.

REAL, PARAMETER  ::  pi=4.0*ATAN(1.0)

Attempting to change the value of a variable declared to be a parameter will result in a fatal compiler error.

In older code the declaration and parameter statement will be on different lines

real  pi
parameter (pi=3.14159)