Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.7.0.10 $; site ccvaxa
Path: utzoo!watmath!clyde!cbosgd!ihnp4!inuxc!pur-ee!uiucdcs!ccvaxa!aglew
From: aglew@ccvaxa.UUCP
Newsgroups: net.lang.c
Subject: Re: An amusing piece of code
Message-ID: <2600044@ccvaxa>
Date: Sat, 5-Apr-86 23:13:00 EST
Article-I.D.: ccvaxa.2600044
Posted: Sat Apr  5 23:13:00 1986
Date-Received: Thu, 10-Apr-86 21:21:23 EST
References: <1370@ism780c.UUCP>
Lines: 79
Nf-ID: #R:ism780c.UUCP:1370:ccvaxa:2600044:000:2896
Nf-From: ccvaxa.UUCP!aglew    Apr  5 22:13:00 1986


>/* Written 10:10 pm  Apr  4, 1986 by tim@ism780c.UUCP in ccvaxa:net.lang.c */
>[Describes a code structure that looks reasonably ugly with if-then-elses]
>[I'll include it at the bottom of this letter for reference]
>We would like to use a switch for everything.  Here is a solution:
>
>	switch ( thing ) {
>case A:         A-code; break;
>case B:         B-code; if ( 0 ) {
>case C:         C-code; if ( 0 ) {
>case D:         D-code; }}
>		BCD-common-code; break;
>case E:         E-code;
>	}
>
>Noone here has been able to come up with a reasonable style for this.  The
>example above is not to bad, but if B-code, C-code, etc, are complicated,
>then it starts to get ugly.

I find it amusing to now be in the position of advocating `unstructured' 
goto code, since in my last job I was the evangelist of structured programming.
Oh, well, into the fray...

May I make a modest suggestion? The above code is a classic example of 
converging flows of control. Write it:

   switch ( thing ) {
       case A:	   A-code;     break;

       case B:	   B-code;     goto BCDcode;
       case C:	   C-code;     goto BCDcode;
       case D:	   D-code;     goto BCDcode;

         BCDcode:  BCD-common-code; break;

       case E:	   E-code;     break;
   }

Notes: (1) DO NOT cut out the goto BCDcode after D-code. Fall through should 
never be used in this type of situation, since it can cause awkward bugs
if the D case gets moved around. (2) I have tried to use the indentation 
above to indicate that BCDcode is subordinate to cases B, C, and D.
(3) Try to choose a reasonably meaningful name for the label.

Now, I ask you: which is more easily readable? The goto code, or those if(0) {
abominations in the example above? Or the original if-then-else code, which
I include here?

>	if ( thing == A )
>		A-code;
>	else
>	if ( thing == B || thing == C || thing == D ) {
>		switch ( thing ) {
>	case B:         B-code; break;
>	case C:         C-code; break;
>	case D:         D-code; break;
>		}
>		BCD-common-code;
>	} else
>	if ( thing == E )
>		E-code;
>
>A, B, C, D, and E are constant expressions, so this is not elegant.
>We would like to use a switch for everything.  Here is a solution:

A last word - gotos are often the most elegant means of implementing
complicated structures. Whether you should be using such complicated
structures is another question - for me, finding that I can code 
something more cleanly with gotos than with conventional structures
sets the flag <There is probably something wrong with this algorithm>.
As well it should. But gotos should not be thrown out.

Structured programming != gotoless programming

(Does anybody know Dr. Goto from Japan? I believe he's now a motor behind
the Fifth Generation Computer Project.)

Andy "Krazy" Glew. Gould CSD-Urbana.    USEnet:  ihnp4!uiucdcs!ccvaxa!aglew
1101 E. University, Urbana, IL 61801    ARPAnet: aglew@gswd-vms