Re: "use utf8" in Locale::Maketext

Sean M. Burke wrote:

> At 02:37 PM 2003-06-06 +0000, Richard Evans wrote:
>>Having talked to a few folks about it, the answer was actually dead
>>simple:
>>   BEGIN
>>   {
>>     return unless $] >= 5.006;
>>    require utf8; "utf8"->import;
>>   }
>>I've tested it with perl 5.8, 5.6 and a 5.5 nsperl (can't remember the
>>exact release off hand).
> 
> 
> Hm.  But the utf8->import has no effects outside of that block.  To wit:
> 
> 
> C> \perl628\bin\perl -v
> 
> This is perl, v5.6.1 built for MSWin32-x86-multi-thread
> [...]
> Binary build 628 provided by ActiveState Tool Corp.
> Built 15:41:05 Jul  4 2001
> [...]
> 
> 
> C> type utf8_thingy.pl
> 
> my $x;
> BEGIN {
>    die "Nevermind!" unless $] >= 5.006;
>    require utf8; "utf8"->import();
>    $x = chr(338);
>    $y = $x;   $y =~ s/./!/g;   print "A: length: ", length $y, "\n"
> }
> 
> $y =   $x;   $y =~ s/./!/g;    print "B: length: ", length $y, "\n";
> 
> 
> 
> C> \perl628\bin\perl utf8_thingy.pl
> A: length: 2
> B: length: 1
> 
> (Whereas if the utf8->import had effect outside of that block, I'd expect
> to see "A: length: 2   B: length: 2".  Of course, this is noticeable only
> under those versions of Perl where use utf8 really did much at all, such
> as the old version I use above.)

Interesting - yes I get the same results as you with 5.6.1

I was getting bug reports from beta testers about catenated non-utf8 and
utf8 strings giving the wrong results, so I came up with this as a quick
test:

  BEGIN
  {
    return unless $] >= 5.006;

    require utf8;
    "utf8"->import();
  }

  my @month_abbreviations =
  (
    'janv.',
    'févr.',
    'mars',
    'avr.',
    'mai',
    'juin',
    'juil.',
    'août',
    'sept.',
    'oct.',
    'nov.',
    'déc.'
  );

  print $month_abbreviations[7] . "\x{999}\n";

If I comment out the utf8 lines, I get the following output:

  aoûtঙ

Uncommented, I get:

  aoûtঙ


./perl -v

This is perl, v5.6.1 built for i686-linux

This holds true for 5.8 and 5.6.1 on linux - I'm not sure exactly what's
happening then. Looks like the utf8 pragma has an effect at the package and
block level?

Cheers,
Rich
-- 
Richard Evans
[email protected]