On Thu, 2004-05-13 at 04:30, Rafael Garcia-Suarez wrote:
> Aaron Sherman wrote:
> > Is it a special type of calling convention, e.g.:
> >
> > sub s (Regex $pat, Str $replace, bool ?$i) is doublequotelike returns(Str) {
>
> Ooh, "doublequotelike" sounds so much 1984.
> (Moreover it doesn't describe accurately the reality, which allows to
> use different delimiters for the pattern and the replacement.)
Hmm... I think doublequotelike would work, as long as it were a superset
of both sets of behaviors and the object that you get is neither a Regex
nor a Str, but mutable into both (a sort of pre-parsed doublequotelike
thing) and then in this case, you could optimize the conversion of both
parameters, because you know what they'll be converted into.
For example:
role doublequotelike {
has Parser::Perl6::ProtoString $.string;
# Not sure how to say "conversion to Str calls this method"
method as_string() returns(Str) {...}
method as_regex() returns(Regex) {...}
}
doublequotelike.string would just contain a series of tokens (e.g.
m/\s+$pattern/ yields (literal => '\s+', interpolate => '$pattern')). In
this way, you can apply your Perl 5 or Perl 6 mode by simply deriving a
new doublequotelike class.
Is there some reason not to do that that I'm missing? Why use a macro
here when that prevents you from ever using it as a method?
More to the point, even if it is a macro, would this kind of facility be
allowed for creating a method with the same syntax exposed to the user.
Performing a regex on an object seems so common an idiom that I don't
think it's healthy to make it such a special case.
--
Aaron Sherman <[EMAIL PROTECTED]>
Senior Systems Engineer and Toolsmith
"It's the sound of a satellite saying, 'get me down!'" -Shriekback
