Front page | perl.perl5.porters |
Postings from May 2008
On May 28, 2008, at 12:36, A. Pagaltzis via RT wrote:
> Really this should be decided based on whether the closing brace
> is followed by a comma or not, so in some sense this is certainly
> a bug.
Yes, exactly, very well put.
> But my understanding is that the syntactic ambiguity of
> curly braces in Perl is difficult to handle in perl’s ultimately
> yacc-derived (but heavily mutated) parser, which uses a limited
> lookahead for curlies to work around the fact that this style of
> parser strongly favours tokens with only a single meaning. In
> practice, the heuristic works most of the time and the
> disambiguation cues are easy to employ.
Yes, but this one was so mystifying to me. Now that I know the issue,
I can fix it, but it took me over an hour to get it to work, and I was
mystified by the workaround. Your explanation has really helped, of
course, and I can see why it behaves that way (sort of), but it seems
pretty bizarre absent that knowledge.
> (You can equivalently disambiguate a hash constructor that gets
> misparsed as a block by prepending a unary plus: `+{ ... }`.
> Within such a block, statements are always a compile error.)
I use the unary plus to disambiguate class method calls in my tests,
FWIW:
ok +Some::Class->method, 'Call method';
> Someone else will have to tell you whether this is a WONTFIX tho.
I hope not! Thanks for the info, Aristotle, as always.
David
Thread Previous