goto thread wentto here

[xilman]

Originally Posted by rogue

Personally I'm not a fan of "goto"s in C code. IMO, well-designed code shouldn't need them. Where I work there is a project that requires all functions to have a single exit point, thus gotos are required to make that happen, but the only reason for the single exit point is a trace statement before returning. IMO, it should be quite acceptable to use return from within a function to reduce the need for unnecessary if statements.

I consider goto's harmful and should never be used under any circumstances. The are clearly unnecessary and serve only to hide the structure of a well-designed program. Consider the clarity of the following demonstration code which uses only control constructs from sound structured programing techniques --- a switch() and a while().

#include <stdio.h>

void hello (char *s)
{
  int i = 42;
  char c;

  while (1)
    {
      switch (i)
	{
	case 7:
	  putchar ('\n');
	  i = 93;
	  break;
	case 0:
	  putchar (c);
	  i = 5;
	  break;
	case 42:
	  c = *s;
	  i = 19;
	  break;
	default:
	  s++;
	  i = 42;
	  break;
	case 93:
	  return;
	case 19:
	  i = (c == 0) * 7;
	}
    }
}

main ()
{
  hello ("Hello world");
}

Paul

[rogue]

Originally Posted by xilman

I consider goto's harmful and should never be used under any circumstances. The are clearly unnecessary and serve only to hide the structure of a well-designed program. Consider the clarity of the following demonstration code which uses only control constructs from sound structured programing techniques --- a switch() and a while().
Paul

You're a little late for April Fool's Day, or am I missing your point?

[Brian Gladman]

Originally Posted by rogue

You're a little late for April Fool's Day, or am I missing your point?

Unless I am mistaken, Paul is gently pointing out that the use of structured programming constructs is no more a guarantee of good code any more than the use of goto's is a guarantee of bad code.

The real issue here is not to avoid 'break' or 'goto' clauses but rather to avoid their inappropriate use.

The use of break clauses to achieve early exits from loops is often appropriate - so much so that some high level languages have a specific construct to achieve this.

Brian Gladman

PS to Paul - I was about to disagree with you until I actually read your piece of code :-)

[xilman]

Originally Posted by rogue

You're a little late for April Fool's Day, or am I missing your point?

Yes, you are missing my point, probably because you need to get your irony detector serviced.

Paul

[tmorrow]

And here I was thinking Paul's "Hello World" was a beautifully elegant, transparent, tight, high performance implementation, exhibiting supreme mastery of the language. Damn!

[xilman]

Originally Posted by tmorrow

And here I was thinking Paul's "Hello World" was a beautifully elegant, transparent, tight, high performance implementation, exhibiting supreme mastery of the language. Damn!

But it is! Well, elegant and transparent, at least, though perhaps not optimized for performance.

In particular, it's a clear and simple example of the concept of Turing eqivalence -- in particular, of how to implement GOTO in a Turing-complete language which lacks an explicit representation of that concept. I even demonstrated how to implement a computed GOTO though that was, I confess, showing off.


Paul