Re: Using Bison and Delphi

Mike Hahn wrote:
>
> "VBDis" <vbdis@aol.com> wrote:
>
> > The omission of punctuation between identifiers (dot operator,
> > argument list separator...) may cause more trouble than it's worth -
> > worth what at all? Since all tokens (identifiers...) must be separated
> > somehow, it doesn't make a big difference whether the separation
> > character is a space, dot, comma or whatsoever. Distinct separation
> > characters, instead of only whitespace, also allow for easier parsing
> > (less lookahead) and better diagnostics and error recovery.
>
> I disagree that the omission of the Java dot operator (making it
> implied) is a language flaw. If the first token (at the statement
> level) is an identifier, then the parser knows that a procedure call
> is coming. If the first token (at the expression level) is an
> identifier, then the parser knows that a function call is
> coming. Examples:
>
> foo myproc (arg1, arg2,... );
> = foo (bar myfunc (arg1, arg2,... ));
> = foo (bar func1 (arg1, arg2,... ) func2 (arg1, arg2,... ));
>
> The first statement calls the myproc procedure of the foo object. The second
> statement calls the myfunc function of the bar object, and assigns the
> result to the foo variable. The third statement calls func1, which returns
> an object having a method called func2. It then calls func2, and assigns the
> result to the foo variable.
>
> Maybe doing it this way is slightly harder to parse, but it enhances
> the simplicity of the language, in my opinion.
>
> Mike Hahn


Without distinct separators how does one parse the following


      foo(a -b)


Is this one argument, a-b, or two arguments, a, and -b


Same applies to subscripts.


Incidentally, Cobol does not require seperators between subscripts. As
Cobol 85 restricted subscripts to "var [ +/- constant ]" one can
disambiguate, buts its messy and annoying.

Return to the comp.compilers page.
Search the comp.compilers archives again.