research!rsc

Steve Bourne* worked on an Algol 68 compiler at the Cambridge University Computer Lab before moving to Bell Labs, where he worked on Seventh Edition Unix. He is best known for writing V7's /bin/sh, now known as the Bourne shell.

Bourne's preference for Algol syntax is evident in shell scripts, which use Algol constructs like if ... then ... elif ... else ... fi and case ... esac instead of C equivalents that might be more familiar to Unix users. (The C shell and rc were both inspired, in part, by a desire for C-like syntax.)

Bourne's preference for Algol syntax is, surprisingly, also evident in the shell source. More comfortable with Algol but stuck with C, Bourne found a way to cope:

/usr/src/cmd/sh/mac.h:
#define IF      if(
#define THEN    ){
#define ELSE    } else {
#define ELIF    } else if (
#define FI      ;}

#define BEGIN   {
#define END     }
#define SWITCH  switch(
#define IN      ){
#define ENDSW   }
#define FOR     for(
#define WHILE   while(
#define DO      ){
#define OD      ;}
#define REP     do{
#define PER     }while(
#define DONE    );
#define LOOP    for(;;){
#define POOL    }

Using these macros, one can write programs that would certainly feel familiar to an Algol programmer:

/usr/src/cmd/sh/service.c:297:
LOCAL VOID      gsort(from,to)
        STRING          from[], to[];
{
        INT             k, m, n;
        REG INT         i, j;

        IF (n=to-from)<=1 THEN return FI

        FOR j=1; j<=n; j*=2 DONE

        FOR m=2*j-1; m/=2;
        DO  k=n-m;
            FOR j=0; j=0; i-=m
                DO  REG STRING *fromi; fromi = &from[i];
                    IF cf(fromi[m],fromi[0])>0
                    THEN break;
                    ELSE STRING s; s=fromi[m]; fromi[m]=fromi[0]; fromi[0]=s;
                    FI
                OD
            OD
        OD
}

(It's not clear why Bourne used braces around function bodies instead of BEGIN and END.)

The Bourne shell code, along with the 4.2BSD finger implementation, was the impetus for the International Obfuscated C Code Contest.

* Really, Stephen R. Bourne, but even then not quite specific enough. Years ago, I heard a story that at one point there were two Stephen R. Bournes working at SGI. Callers would ask the operator for Steve Bourne and would be asked, ‘which one?’ ‘Steven R. Bourne,’ they'd say. ‘Sure, which one?’ The ambiguity was apparently resolved by referring to them as ‘hardware Steve’ and ‘software Steve.’ I can find no justification for this story nor any trace of ‘hardware Steve’ other than the mention of two Steven R. Bournes in the sendmail README, which claims that they were both at Bell Labs, not SGI. If you can confirm or deny this story, please post a comment or mail me (rsc@swtch.com).

Posted by rsc on Friday, February 15, 2008 | 2 comments |
Labels: Bell Labs, history, macros