I had a problem with an application that produced a horrible mixed UTF-8 and ISO-8859 encoded XML output. I found this way to transform it to pure UTF-8 without double-encoding the UTF-8 sequences that were already there. I know this will not work in all cases, but it has been helpful. What do you think?
#!/usr/bin/perl use strict; # mixed string with ISO 8859-1 und UTF-8: my $test_string = "Das Å (auch \"bolle-Å\" genannt, was soviel bedeute +t wie \"Kringel-Å\") ist mit der ". force_utf8("dänischen Rechtschreibreform von 1948 eingeführt worde +n."); print "Source: $test_string\n"; print "UTF : ".force_utf8($test_string)."\n"; print "ISO : ".force_latin($test_string)."\n"; sub force_utf8 { my $string = shift; $string =~ s/([\xc0-\xdf][\x80-\xbf]{1}|[\xe0-\xef][\x80-\xbf]{2}| +[\xf0-\xf7][\x80-\xbf]{3}|[\x80-\xff])/&encode_char_utf8($1)/ge; return $string; } sub force_latin { my $string = shift; $string =~ s/([\xc0-\xdf][\x80-\xbf]{1}|[\xe0-\xef][\x80-\xbf]{2}| +[\xf0-\xf7][\x80-\xbf]{3}|[\x80-\xff])/&decode_char_utf8($1)/ge; return $string; } sub encode_char_utf8 { my $char = shift; if($char =~ /^([\xc0-\xdf][\x80-\xbf]{1}|[\xe0-\xef][\x80-\xbf]{2} +|[\xf0-\xf7][\x80-\xbf]{3})$/) { return $char; } my $value = ord($char); return chr(($value>>6) | 0xc0).chr(0x80 | ($value & 0x3f)); } sub decode_char_utf8 { my $char = shift; if($char =~ /^([\xe0-\xef][\x80-\xbf]{2}|[\xf0-\xf7][\x80-\xbf]{3} +)$/) { return ''; } elsif($char =~ /^([\xc0-\xdf])([\x80-\xbf])$/) { my $value = ((ord($1) & 0x1f)<<6)+(ord($2) & 0x3f); if($value<256) { return chr($value); } else { return ''; } } else { return $char; } }

Replies are listed 'Best First'.
Re: Mixed ISO-8859/UTF-8 conversion
by Juerd (Abbot) on Oct 04, 2007 at 11:04 UTC

    Please use Perl's built in support for encodings. Read perlunitut to find out all about Perl's Unicode strings.

    sub decode_utf8_latin1 { my ($encoded) = @_; my $decoded; while (length $encoded) { $decoded .= decode("UTF-8", $encoded, Encode::FB_QUIET); $decoded .= substr($encoded, 0, 1, "") if length $encoded; } return $decoded; }

    If CHECK = coderef worked, you could just write decode("UTF-8", $encoded, sub { chr shift }, but alas, it doesn't work.

    Juerd # { site => 'juerd.nl', do_not_use => 'spamtrap', perl6_server => 'feather' }

      Wow. It's a bit hard to understand. I didn't expect Encode::FB_QUIET to change $encoded. But that's a better way, thanks a lot.

        I didn't expect Encode::FB_QUIET to change $encoded.

        All I can say is RTFM ;-)

Re: Mixed ISO-8859/UTF-8 conversion
by zby (Vicar) on Oct 04, 2007 at 10:52 UTC
    The horror ... the horror ...

      The jerk ... the jerk ...

      You could at least explain /why/ you think it is horrific, and perhaps even include suggestions for a better solution.

        That was more about the situation that the OP has to face than about the solution :)