Front page | perl.perl5.porters |
Postings from February 2011
On Mon, Feb 21, 2011 at 10:49:31AM -0700, Tom Christiansen wrote:
> >> Here are the ones I can think of at the moment:
> >>
> >> /w to change the meaning of \b to Unicode's word boundary algorithm.
> >> /k to change /i to mean compatible decomposition
>
> > Please don't use /k. Having /k means '(?k:)' becoming valid, and that
> > will interfere with Regexp::Common, which uses '(?k:)'. /p started out
> > as /k, and it was changed to /p as to not interfere with Regexp::Common.
>
> The first thing that comes to mind is: do you have an alternate suggestion?
>
> As for Regexp::Common, UGH! I have rather mixed feelings about that.
>
> It's not even a module in the core Perl distribution, and yet its use of a
> syntax reserved to the core blocks progress in that direction. How many
> others like this exist?
>
> On the other hand, adding /k would suddenly change the behavior of
> existing programs through no fault of their owners. As frustrating
> though this may me, it seems a real blocker, and I fear you're right.
It's not going to change meaning of existing programs.
But what Regexp::Common does is take a string, process it for '(?k:)',
and turn it into a pattern. When Damian picked '(?k:)', it was safe,
as it didn't have a meaning. By allowing it, one creates an ambiguity:
does the user intent the Regexp::Common meaning of '(?k:)', or the
regexp engines meaning?
> But it is a very natural thing to use.
And so it was for ${^POSTMATCH} and friends before it got renamed to /p.
I'm not saying /k should not be used. But people should be aware of the
implications.
Abigail
Thread Previous
|
Thread Next