hashing external names & other goodies.

[Mark Terribile]

> -- are break (n) and continue (n) in the ANSI Proposed Standard yet?  (If
> you couldn't guess, the one breaks (n) levels of code structures -- case's,
> while's, etc. -- while the other continues with the n'th enclosing loop.)
> Yes, I recognise the existence of setjmp() and longjmp(), but get queasy at
> the thought of any kind of jump.

I hope these NEVER get in!  This is a great way to write fragile code, and
there is a MUCH more durable way to get what you are trying to get

	Why is it fragile?  Well, it's easy to break it accidentally when
you rework code.

		while( ... )
			switch( ... )
			{
			  case ..:
				... lots of code ...
				break 2;
				... lots more code ...

				break;
			  case ..:
				...
			}

What happens when that ``lots of ... lots more code'' gets surrounded by
another loop or another switch in the course of the program's evolution?

This is a great pitfall for ``the next guy'' who is going to use this code.
This design is only pleasing to sociopaths.

One better way is with (horrors) a label.  Both ADA and (extended) PL/I
do this, and it seems to work pretty well.  The exact forms differ from
language to language, but in C it might take the form of


		while( ... )
		  command_char:
			switch( ... )
			{
			  case ..:
				... lots of code ...
				break command_char;
				... lots more code ...

				break;
			  case ..:
				...
			}

This is now solidly immune to the addition of new levels of control between
the structure to be ``broken'' and the decision to break it.
-- 

	from Mole End			Mark Terribile
		(scrape .. dig )	hou4b!mat
    ,..      .,,       ,,,   ..,***_*.