🚀 go-pugleaf

Greg Comeau (comeau@panix.com) wrote:
: In article <47s7l0$r18@news.mel.aone.net.au> andrewd@agnus.livewire.com.au (Andrew Dunbar) writes:
: >The only problem I can see is for do loops:
: >
: >do
: >{
: >    while inner (exp1)
: >    {
: >        if (exp2)
: >            break outer;
: >    }
: >} while outer;
: >
: 
: That's one issue.  That wouldn't work anyway because you do not have
: an expression for the do/while.  Which is part of the point of where
: things start collapsing with some proposals re things like this.
: Because now we can have:
: 
:     do .... while outer (expr);
: 
Oops! That was a typo on my part. It should have been as you have suggested
above.
: 
: while looks odder (IMO) that the for, but then so does:
:     do outer.... while (expr);

This is syntactically ambiguous anyway as 'outer' is simply an identifier
it is a legal statement and the syntax of a do loop is:
      do <statement> while (<expression>);

Although as there is no semicolon after the loop name it may be distinguishable
but I still doubt it. I might try it out with YACC soon.
:       
: and any other permutations.
: 
: Or to minimize impacts to the grammar, why not just leave as normal labels:
: 
:     outer: for(...) { ... }
: 
Well that's the Java approach so I've heard, but I didn't really think of the
loop names as labels. This would cause confusion for humans and parsers/
optimizers with real goto labels.

: but that doesn't fly as well, so how about:
: 
:     for(...) { ... } outer:
: 
: But that seems odd, and further, doesn't work in all cases w/o having
: to add a ; or something.
: 
When I thought it up I had multi-level break specifically in mind so:

    iteration-statement:
	while <loop-name> ( <expression> ) <statement>
	do <statement> while <loop-name> ( <expression> ) ;
	for <loop-name> ( <expression>opt ; <expression>opt ;
		<expression>opt ) <statement>

    selection-statement:
	switch <loop-name> ( <expression> ) <statement>

    jump-statement:
	break <loop-name> ;

would be the exact additions I would introduce into the grammar as
stated in 'The C Programming Language 2nd Edition'
(where 'loop-name' is simply a synonym for 'identifier')

: And while we are at it, why not add it to if statements, switches,
: (even return! <gasp>) and so on.
: 
I can't see a use for named 'if' statements but there may be a case for it?

: The idea is probably nice, and there is prior art from other languages,
: however, the impact and tradeoffs get bigger and bigger the more you
: look at it within a C/C++ world.
: 
Thanks for your views Greg

Andrew Dunbar
GPSoftware