Re: gnat_regpat and unexpected handling of alnum and unicode needed

[Simon Wright <simon@pushface.org>]

19.krause.70@googlemail.com writes:

> The expression [[:alnum:]] matches the underscore in gnat_regpat but
> not in egrep. It feels much more natural to don't match the underscore
> like egrep does. And I think it is more posix compliant.
>
> Question is why?

Because, at s-regpat.adb:2325, we find

   function Is_Alnum (C : Character) return Boolean is
   begin
      return Is_Alphanumeric (C) or else C = '_';
   end Is_Alnum;

(Is_Alphanumeric is in Ada.Characters.Handling), presumably because the
author liked using underscores in identifiers.

> How do I handle unicode strings with gnat_regpat, because [[:alpha:]]
> seems to match only ascii a-zA-Z.

What GNAT does with -gnatW8 is to read UTF-8 from the source file and,
in the case of characters, convert then to the internal Latin-1
(approximately) character. So your 'ö' is converted to the single
character with value 246, LC_O_Diaeresis.

I tried just letters, and got

   fööbär Matched regexp3 ^[[:alpha:]]+$!

No idea what's going on here!