8.5.2 Functions

A function subprogram is invoked in an expression and returns a single value (a function result) that is used to evaluate the expression.

The FUNCTION statement is the initial statement of a function subprogram. It takes the following form:

[prefix] FUNCTION name ([d-arg-list]) [RESULT (r-name)]

prefix
Is one of the following:

type [keyword]
keyword [type]

type
Is a data type specifier.

keyword
Is one of the following:

Keyword	Meaning
RECURSIVE	Permits direct recursion to occur. If a function is directly recursive and array valued, RESULT must also be specified (see Section 8.5.1.1).
PURE	Asserts that the procedure has no side effects (see Section 8.5.1.2).
ELEMENTAL	Restricted form of pure procedure that acts on one array element at a time (see Section 8.5.1.3).

You can also specify one of the High Performance Fortran (HPF) keywords (see Section 8.2). These keywords are functional only in parallel programs on Tru64 UNIX systems.

name
Is the name of the function. If RESULT is specified, the function name must not appear in any specification statement in the scoping unit of the function subprogram.

The function name can be followed by the length of the data type. The length is specified by an asterisk (*) followed by any unsigned, nonzero integer that is a valid length for the function's type. For example, REAL FUNCTION LGFUNC*8 (Y, Z) specifies the function result as REAL(8) (or REAL*8).

This optional length specification is not permitted if the length has already been specified following the keyword CHARACTER.

d-arg-list
Is a list of one or more dummy arguments.

r-name
Is the name of the function result. This name must not be the same as the function name.

Rules and Behavior

The type and kind parameters (if any) of the function's result can be defined in the FUNCTION statement or in a type declaration statement within the function subprogram, but not both. If no type is specified, the type is determined by implicit typing rules in effect for the function subprogram.

Execution begins with the first executable construct or statement following the FUNCTION statement. Control returns to the calling program unit once the END statement (or a RETURN statement) is executed.

If you specify CHARACTER*(*), the function assumes the length declared for it in the program unit that invokes it. This type of character function can have different lengths when it is invoked by different program units; it is an obsolescent feature in Fortran 95.

If the length is specified as an integer constant, the value must agree with the length of the function specified in the program unit that invokes the function. If no length is specified, a length of 1 is assumed.

If the function is array-valued or a pointer, the declarations within the function must state these attributes for the function result name. The specification of the function result attributes, dummy argument attributes, and the information in the procedure heading collectively define the interface of the function.

The value of the result variable is returned by the function when it completes execution. Certain rules apply depending on whether the result is a pointer, as follows:

• If the result is a pointer, its allocation status must be determined before the function completes execution. (The function must associate a target with the pointer, or cause the pointer to be explicitly disassociated from a target.)

  The shape of the value returned by the function is determined by the shape of the result variable when the function completes execution.

• If the result is not a pointer, its value must be defined before the function completes execution. If the result is an array, all the elements must be defined; if the result is a derived-type structure, all the components must be defined.

A function subprogram cannot contain a SUBROUTINE statement, a BLOCK DATA statement, a PROGRAM statement, or another FUNCTION statement. ENTRY statements can be included to provide multiple entry points to the subprogram.

You can use a CALL statement to invoke a function as long as the function is not one of the following types:

• REAL(8)
• REAL(16) (VMS, U*X)
• COMPLEX(8)
• COMPLEX(16) (VMS, U*X)
• CHARACTER

Section 8.5.2.1 describes the RESULT keyword and Section 8.5.2.2 describes function references.

Examples

The following example uses the Newton-Raphson iteration method (F(X) = cosh(X) + cos(X) - A = 0) to get the root of the function:

FUNCTION ROOT(A)
  X  = 1.0
  DO
    EX = EXP(X)
    EMINX = 1./EX
    ROOT  = X - ((EX+EMINX)*.5+COS(X)-A)/((EX-EMINX)*.5-SIN(X))
    IF (ABS((X-ROOT)/ROOT) .LT. 1E-6) RETURN
    X  = ROOT
  END DO
END

In the preceding example, the following formula is calculated repeatedly until the difference between X_i and X_i+1 is less than 1.0E-6:

                        cosh(Xi) + cos(Xi) - A
            Xi+1 = Xi - ------------------------
                          sinh(Xi) - sin(Xi)

The following example shows an assumed-length character function:

CHARACTER*(*) FUNCTION REDO(CARG)
  CHARACTER*1 CARG
  DO I=1,LEN(REDO)
    REDO(I:I) = CARG
  END DO
  RETURN
END FUNCTION

This function returns the value of its argument, repeated to fill the length of the function.

Within any given program unit, all references to an assumed-length character function must have the same length. In the following example, the REDO function has a length of 1000:

CHARACTER*1000 REDO, MANYAS, MANYZS
MANYAS = REDO('A')
MANYZS = REDO('Z')

Another program unit within the executable program can specify a different length. For example, the following REDO function has a length of 2:

CHARACTER HOLD*6, REDO*2
HOLD = REDO('A')//REDO('B')//REDO('C')

The following example shows a dynamic array-valued function:

FUNCTION SUB (N)
  REAL, DIMENSION(N) :: SUB
  ...
END FUNCTION

For More Information:

• On general rules that apply to function subprograms, see Section 8.5.1.
  
  
• On argument keywords in function references, see Section 8.5.2.2.
  
  
• On the ENTRY statement, see Section 8.11.
  
  
• On the RETURN statement, see Section 7.9.