transparent lambda

[Peter Levart]

On Friday 08 January 2010 01:55:15 Neal Gafter wrote:
> On Thu, Jan 7, 2010 at 4:10 PM, Peter Levart <peter.levart at gmail.com> wrote:
> 
> >  I would not put labeled block expression in the list of possible
> > StatementExpressions.
> >
> > So the only syntactically correct interpretation of the above example would be "a labeled block statement followed by an empty statement".
> >
> >  Am I missing something?
> >
> > Peter
> >
> 
> Yes.  Can an expression statement *start* with a labeled expression?
> 
>   label : { ... }
>   (expression) . f ();
> 
> Is the parenthesized expression an argument to the labeled block expression
> (which must be of function type), or the start of a new statement?
> 
> We can go on and on here, but the fundamental point is: it is a bad idea to
> have a single syntax be ambiguous between a statement form and an expression
> form.

Ok, I see the point. So what about a different syntax? 
Let's take the block expression as defined by CfJ 0.6b. That is one form of block expression, let's call it "unlabeled block expression":

  (S; E)

Could we introduce another form called "labeled block expression":

  label:(S;)    // I see the point in keeping expression/statement syntaxes separate ;-)

The later being a shortcut for:

  (Type retval; label:S; retval)

Where a special value-bearing break:

  break label : expression;

is a shortcut for

  { retval = expression; break label; }

and the Type being a reduction of all types of expressions present on value-bearing breaks.


for example this:

label:(
  if (a < b) break label : -1;
  if (a > b) break label : 1;
  break label : 0;
)

is equivalent to this:

(int retval; label:{
  if (a < b) { retval = -1; break label; }
  if (a > b) { retval = 1; break label; }
  { retval = 0; break label; }
} retval)


Ok, I see one hard to read expression:

a ? b : c : (...)


But is it ambigous? If it is not then it could be rewritten to be more obvious to human eye:

a ? b : (c : (...));


Many expressions do not need parentheses but are parenthesized for the purpose of clarity.

> 
> Cheers,
> Neal
> 

Regards, Peter