Ultimate CASE Statement#

by Wil Baden, Costa Mesa, California

appeared in VD February 1987

Many citizens of the Forth community have lamented the lack of a case statement in standard Forth language specifications. Since the first rule of Forth programming is, "If you don't like it, change it," there have been many proposals, and Forth Dimensions even held "The Great CASE Contest in Volume II". Although the winning entry of that contest, submitted by Charles Eaker, has been widely implemented and even offered as part of many vendors' systems, the flood of proposals has not ceased. There have been many articles and letters on the subject in Forth Dimensions.

All proposals to date have had problems. Portability is one. Another is that they all have been too specialized and restricted in their area of application. Generalization is accomplished by designing another special case of case.

Strictly speaking, a case statement is unnecessary. It is " syntactic sugar" to make a program easier to write, read and understand. It is so helpful in doing this that it is a standard feature of all other modern programming languages.

Figure 1a

craps (n) 
int n;
{   if (n==7) printf ("You win");
    else if (n == 11) printf ("You win");
    else if (n == 2)  printf ("You lose"); 
    else if (n == 3)  printf ("You lose"); 
    else if (n == 12) printf ("You lose"); 
    else printf ("%d is your point",n);
}

Figure 1a is a rather futile program written in C to illustrate a common pattern of logical decisions in many programs.("==" is "equal to" for comparing two things, to distinguish it from "=" for assignment as in Fortran or Basic.) An equivalent Forth version would look something like Figure 1b.

Figure 1b

: CRAPS ( n -- ) 
  DUP 7 =
  IF DROP ." You win" 
  ELSE DUP 11 =
    IF DROP ." You win" 
    ELSE DUP 2 =
      IF DROP ." You lose" 
      ELSE DUP 3 =
        IF DROP ." You lose" 
        ELSE DUP 12 =
          IF DROP ." You win"
          ELSE . ." is your point" THEN 
          THEN THEN THEN THEN THEN ;

Most people will agree that Figure 1a would be better written as in Figure 2a. An even better way is found in some dialects of C, illustrated by Figure 2a. In this extension, following syntax from Pascal, values separated by "," indicate a set of values, and values separated by ".." indicate a range.

Figure 2a

craps (n) 
int (n);
{
	switch(n) { 
	case  7: printf("You win"); break; 
	case 11: printf("You win"); break; 
	case  2: printf("You lose"); break; 
	case  3: printf("You lose"); break; 
	case 12: printf("You lose"); break; 
	default: printf("%d is your point",n); 
	}
}
Some Forth proposals have one definition for individual values and another definition for a range of values. There would have to be another definition for a set of values. No earlier Forth proposal that I know of allows sets and ranges together, as in:
case 2..3, 12:
What is proposed here is a single case statement for Forth which will include all these variations, and many more, that can be implemented in fig-FORTH, Forth-79, Forth-83 and any other Forth.

Figure 2b

craps(n)
int n;
{  switch(n)  {
   case 7, 11:    printf("You win"); break;
   case 2..3, 12: printf("You lose"); break;
   default:       printf("%d is you point",n);
   }
}

Figure 3

: CASE  DUP ;

: CRAPS ( n -- )
   CASE   7 = IF DROP ." You win"  EXIT THEN
   CASE  11 = IF DROP ." You win"  EXIT THEN
   CASE   2 = IF DROP ." You lose" EXIT THEN
   CASE   3 = IF DROP ." You lose" EXIT THEN
   CASE  12 = IF DROP ." You lose" EXIT THEN
   . . is your point" ;

Figure 2a would look as shown in Figure 3. Let's add two more spoons of syntactic sugar, as in Figure 4. As has been noted elsewhere, too much syntactic sugar causes semantic diabetes. Our case is sweet enough. Figure 5 is an example to show some of the possibilities.

Figure 3

: OF  ( n flag -- ) [COMPILE] IF  COMPILE DROP ; IMMEDIATE
: =OR ( n flag n -- n flag ) 2 PICK = DROP ;

: CRAPS ( n -- )
  CASE  7 =  11 =OR             OF ." You win"  EXIT THEN
  CASE 2 3 BETWEEN  12 =OR      OF ." You lose" EXIT THEN
  . ." is your point" ;

Figure 4

: WHATEVER  ( n -- )
  CASE 0=                       OF ." zero" EXIT THEN
  CASE 0<                       OF ." Negative" EXIT THEN
  CASE DUP 1- AND 0=            OF ." Power of 2" EXIT THEN
  CASE ASCII 0  ASCII 9 BETWEEN OF ." Digit" EXIT THEN
  CASE ASCII ,  ASCII / BETWEEN
       ASCII : =OR              OF ." Punctuation ,-./: " EXIT THEN
  DROP ." Whatever" ;

Now for a real life example. Figure 6 is a recension of a word in John James "Universal Text File Reader" (Forth Dimensions VII/3). One of my favorite examples is "Thirty days hath September, April, June and November ..." See Figure 7. If NUMBER in your system is vectored, you may want to replace it in some applications with a version that selects the numerical radix according to the first character. Figure 8 implements a convention used on Motorola systems (e.g., 68000). Laxen's CLASSIFY example (FD VII/I) can be written without redundant classes with no additional definitions, as in Figure 9.

Since DUP is assembler code, in most systems you can optimize its definition with something like that in Figure 10a. The Forth-79 definition of =OR is given in Figure 10b. If you do not have PICK, as in fig-FORTH, or if PICK is not an assembler code definition, see Figure 10c.

A case statement in any programming language is intended for a series of tests to classify a value. To do this in other languages without using a case structure would require repeating the value at each test, giving a tedious appearance to the source ln Forth, the data stack allows us to avoid such explicit references to the value In Forth, a CASE statement has the pattern DUP ... IF DROP ... We have sweetened this to CASE ... OF ... .

The trivial nature of the implementation emphasizes that a case statement is not essential to Forth. Those Forth practitioners who pride themselves on how lean and mean their Forth is will find it superfluous. My intent is not to propose this definition of case for standardization; but on the other hand, any further case proposal should be as simple to implement, as portable and as powerful.

Auxiliary Definitions#

You may already have some of these. Your definitions may be different from those shown in Figure 11a. #BLANK-LINES and ?NEW-LINE are words peculiar to the application. #BLANK-LINES is a variable counting the number of successive blank lines. ?NEW-LINE does a CR when the value of #BLANK-LINES is less than two.

Figure 11b provides definitions for several fundamental Forth words. It also presents a naive version of NUMBER? that ignores details such as sign and punctuation, and is not inĀ­ tended for actual use.