Unicode named groups for regular expressions

[BenjaminGalliot]

Hello,

This issue is strongly related to my old (more than 1 year) message here.

It is to allow Unicode named groups for regular expressions, it seems it was just related to a limitation of PCRE, which was updated to fix it last year.

I remember I tried near the end of last year the latest PCRE2 and it worked, and because I just tried again on the latest Julia 1.4 without success, I read again this topic and it is true they asked me to open an issue here! Done!

Sincerely.

[Micket]

Related #35322 ?

[BenjaminGalliot]

It is related, yes (I did not see this recent issue, at least I am not alone!), and PCRE was updated last year for this by Philip Hazel. I copy here the mails he sent to me (in January 2019), if it can help.

Thank you for raising this issue. When names were first invented for
PCRE, before Perl had them, it seemed sensible to restrict them to
letters, underscore, and digits. Perl eventually introduced named
groups, but I didn't make any changes to PCRE, and I haven't thought
about non-ASCII characters in names. I have just had a quick look at
Perl's current documentation. It says two things:

1. "name must not begin with a number, nor contain hyphens", which isn't
   very helpful.

2. "Currently NAME is restricted to simple identifiers only. In other
   words, it must match "/^[_A-Za-z][_A-Za-z0-9]*\z/" or its Unicode
   extension (see utf8), though it isn't extended by the locale (see
   perllocale)."

I have yet to find where in the copious Perl documentation it has a
definition for this extension. However, a quick test indicates that it
does allow, for example, ABáC as a name. I will do some more research
and then investigate upgrading PCRE2 in line with Perl. The next PCRE2
release (10.33) is likely to come out in a couple of months' time. I
will let you know when there is something to test.

Regards,
Philip

I have just committed a patch that upgrades PCRE2 to be more like Perl.
If it is running in UTF mode, capture group names are now allowed to
contain Unicode letters and Unicode decimal digits (and underscore). In
other words, names must match this pattern:

^[\p{L}][\p{L}\p{Nd}]*\z

The next release (10.33) is likely to happen in a couple of months'
time. In the meantime, you can check out the the latest code like this:

svn co svn://vcs.exim.org/pcre2/code/trunk pcre2

Regards,
Philip

[vtjnash]

Fixed by #39310