forterpbeta

Documentation

5. Data & procedure identification

5. Data & procedure identification

Everything a program works on is identified by a name or written as a constant (§5). This chapter is the catalogue of those identifiers: the kinds of constant, variables, arrays and their subscripts, and how a name acquires a data type. Two verbs recur throughout the standard:

  • to reference a datum is to make its current value available — you reference a variable by using it in an expression;
  • to define a datum is to give it a value — assignment defines a variable. A datum with no value yet is undefined, and using an undefined value is not allowed (§10).

Constants

A constant is a datum that is always defined and can never be redefined (§5.1.1). Each data type has its own written form (introduced in Chapter 4):

Type Example constants Notes
integer 0, 42, -7 a digit string, optionally signed (§5.1.1.1)
real 3.14, .5, 6.022E23 decimal point and/or E exponent (§5.1.1.2)
double precision 1.5D0, 3.0D-2 D exponent instead of E (§5.1.1.3)
complex (1.0,2.0) a parenthesized pair of reals (§5.1.1.4)
logical .TRUE., .FALSE. the two truth values (§5.1.1.5)
Hollerith 4HABCD count, H, then exactly that many characters (§5.1.1.6)

A real constant is written as an integer part, a decimal point, and a fractional part — either part may be empty, but not both (§5.1.1.2): 3., .5, and 3.14 are all valid; a bare . is not. A Hollerith constant may appear only in the argument list of a CALL statement and in a DATA initialization statement (§5.1.1.6).

Variables and arrays

A variable is a datum identified by a symbolic name; it may be referenced and defined (§5.1.2). An array is an ordered set of data of one, two, or three dimensions, identified by an array name (§5.1.3); the array is declared by an array declarator (DIMENSION, or a type or COMMON statement — Chapter 7).

Array elements and subscripts

A single array element is identified by following the array name with a parenthesized subscript (§5.1.3.1, §5.1.3.2). The number of subscript expressions must match the array's declared dimensionality, and array storage is column-major — the first subscript varies fastest (the array element successor function, §7.2.1.1.1).

Crucially, FORTRAN 66 restricts a subscript expression to one of just seven simple forms (§5.1.3.3), where c and k are integer constants and v is an integer variable:

      c*v+k        c*v-k        c*v        v+k        v-k        v        k

      DIMENSION A(20)
      I = 2
      A(2*I+1) = 9.0
C     -> subscript 2*I+1 = 5, a legal "c*v+k" form

A subscript like A(I+J) (two variables) or A(I*I) (a variable times a variable) is not permitted in FORTRAN 66 — you must compute the index into an integer variable first and subscript with that:

      M = I + J
      A(M) = 9.0

Procedures

A procedure is identified by a symbolic name (§5.1.4) and is one of: a statement function, an intrinsic function, a basic external function, an external function, or an external subroutine (all in Chapter 8). A function supplies a result to be used at the point of reference; a subroutine does not — it is invoked by CALL and acts through its arguments and side effects.

A function reference is the function name followed by an actual argument list in parentheses (§5.2): SQRT(X), MAX0(I,J,K).

Type rules

The data type of a name is fixed within a program unit (§5.3):

  • a constant's type is implicit in how it is written;
  • a subroutine or common block name has no data type;
  • a variable, array, or statement-function name takes its type from an explicit type-statement if present; otherwise by the implicit rule — initial letter I, J, K, L, M, N → integer, any other letter → real;
  • an intrinsic or basic external function has the type given in Tables 3 and 4 (Appendix A);
  • an external function's type is set in its own subprogram (implicitly by name, or explicitly on the FUNCTION statement).

Dummy arguments

A dummy argument of an external procedure identifies a variable, array, subroutine, or external function (§5.4). When the procedure is referenced, each dummy is associated with the corresponding actual argument; a value of the same type must be available through that association. Argument association is detailed in Chapters 8 and 10.

forterp notes.

  • forterp enforces the restricted subscript forms under F66: a subscript outside the seven shapes of §5.1.3.3 (plus their obvious commutations, e.g. I*2 and 2+I) is a hard error (?FTNNRC … F66 array subscript must be of the form …). The FORTRAN10 and F77 dialects lift this and allow a general integer expression as a subscript.
  • forterp allows arrays of up to seven dimensions on every dialect; the X3.9-1966 limit of three is not enforced, as accepting more cannot break a conforming program. (The cap of seven is enforced — see the F77 manual.)