Front page | perl.perl5.porters |
Postings from August 2013
Unicode properties are only defined for code points 0-0x10FFFF, yet Perl
allows the expression of code points up to UV_MAX, a much larger number.
That means that the Unicode behavior for the remainder is undefined.
We have chosen to make that behavior be as described in the warning. I
believe it was me who came up with the warning, and that it was
introduced in Perl for 5.16. (And I don't have the tuits/energy to
fully research it right now, as I believe that is tangential anyway to
this post.)
Most code will never deal with such large code points, and hence will
never encounter the warning. But if it does happen, the warning will
likely be displayed many times, even in the same regex match when
backtracking occurs over the large code point(s).
There are also bugs in the implementation. It was quite wrong for Perl
v5.16, as Tom Christiansen discovered; largely fixed for v5.18, but the
warning still doesn't get displayed if the regex node that contains the
\p{} or \P{} is optimized into something besides the normal one; and it
can be displayed twice for the same code point even if there is no
backtracking, the first time for the regex optimizer's synthetic start
class, before regular matching begins.
I have been thinking of what to do. One potential solution is to make
this a once-only display (per thread) message. That means that its
display would set a per-interpreter variable that would cause it to
never display again on the current thread.
But then it occurred to me. Is this message really necessary? Perhaps
we should just get rid of it altogether, and make sure the pod
documentation is very clear about this possibility for the very rare
program that is affected.
Thoughts?
Thread Next