Front page | perl.perl5.porters |
Postings from January 2010
I'm working on a patch to the tokenizer. In the past \N in regexes was
special, always followed by a {NAME}. Now, \N can also be a substitute
for dot, unaffected by the /s modifier. So, it is analogous to dot.
But consider:
./perl -Ilib -w -E "say 'B{A' =~ qr/.{A/"
1
But the corresponding:
./perl -Ilib -w -E "say 'B{A' =~ qr/\N{A/"
Missing right brace on \N{} in regex;
So they don't act analogously. (Escaping the brace with a backslash works.)
If the letter after the brace is a digit, they do act analogously:
./perl -Ilib -w -E "say 'B{1' =~ qr/.{1/"
1
./perl -Ilib -w -E "say 'B{1' =~ qr/\N{1/"
1
The behavior could be to leave it this way. No pre-existing code is
broken, as \N always had to be followed by a {NAME}, and still does
outside of patterns.
Or the behavior could be to not consider \N special unless it is
followed by a pair of braces with only things that could be a Unicode
name between them (the space character and \w restricted to ASCII, and
hyphen, I think without looking it up). But then, if someone forgets
the closing brace, it won't generate a warning like it has always done,
and won't do what they probably wanted, like:
./perl -Ilib -w -E "say 'B{SPACE =~ qr/\N{SPACE/" # HYPOTHETICAL
1
I'm inclined to go with the existing behavior, but thought I should
check it out with you all. We could also extend it so that any non-word
character following the left brace would behave as a digit does.
Thread Next