langtags.pm

1. 记录每个帖子的访问人情况

  $lang =~ s<\.[-_a-zA-Z0-9\.]*><>s;  # "en_US.ISO8859-1" -> en-US

  return $lang if &is_language_tag($lang);

  return;
}

###########################################################################

=item * the function encode_language_tag($lang1)

This function, if given a language tag, returns an encoding of it such
that:

* tags representing different languages never get the same encoding.

* tags representing the same language always get the same encoding.

* an encoding of a formally valid language tag always is a string
value that is defined, has length, and is true if considered as a
boolean.

Note that the encoding itself is B<not> a formally valid language tag.
Note also that you cannot, currently, go from an encoding back to a
language tag that it's an encoding of.

Note also that you B<must> consider the encoded value as atomic; i.e.,
you should not consider it as anything but an opaque, unanalysable
string value.  (The internals of the encoding method may change in
future versions, as the language tagging standard changes over time.)

C<encode_language_tag> returns undef if given anything other than a
formally valid language tag.

The reason C<encode_language_tag> exists is because different language
tags may represent the same language; this is normally treatable with
C<same_language_tag>, but consider this situation:

You have a data file that expresses greetings in different languages.
Its format is "[language tag]=[how to say 'Hello']", like:

          en-US=Hiho
          fr=Bonjour
          i-mingo=Hau'

And suppose you write a program that reads that file and then runs as
a daemon, answering client requests that specify a language tag and
then expect the string that says how to greet in that language.  So an
interaction looks like:

          greeting-client asks:    fr
          greeting-server answers: Bonjour

So far so good.  But suppose the way you're implementing this is:

          my %greetings;
          die unless open(IN, "<in.dat");
          while(<IN>) {
            chomp;
            next unless /^([^=]+)=(.+)/s;
            my($lang, $expr) = ($1, $2);
            $greetings{$lang} = $expr;
          }
          close(IN);

at which point %greetings has the contents:

          "en-US"   => "Hiho"
          "fr"      => "Bonjour"
          "i-mingo" => "Hau'"

And suppose then that you answer client requests for language $wanted
by just looking up $greetings{$wanted}.

If the client asks for "fr", that will look up successfully in
%greetings, to the value "Bonjour".  And if the client asks for
"i-mingo", that will look up successfully in %greetings, to the value
"Hau'".

But if the client asks for "i-Mingo" or "x-mingo", or "Fr", then the
lookup in %greetings fails.  That's the Wrong Thing.

You could instead do lookups on $wanted with:

          use I18N::LangTags qw(same_language_tag);
          my $repsonse = '';
          foreach my $l2 (keys %greetings) {
            if(same_language_tag($wanted, $l2)) {
              $response = $greetings{$l2};
              last;
            }
          }

But that's rather inefficient.  A better way to do it is to start your
program with:

          use I18N::LangTags qw(encode_language_tag);
          my %greetings;
          die unless open(IN, "<in.dat");
          while(<IN>) {
            chomp;
            next unless /^([^=]+)=(.+)/s;
            my($lang, $expr) = ($1, $2);
            $greetings{
                        encode_language_tag($lang)
                      } = $expr;
          }
          close(IN);

and then just answer client requests for language $wanted by just
looking up

          $greetings{encode_language_tag($wanted)}

And that does the Right Thing.

=cut

sub encode_language_tag {
  # Only similarity_language_tag() is allowed to analyse encodings!

  ## Changes in the language tagging standards may have to be reflected here.

  my($tag) = $_[0] || return undef;
  return undef unless &is_language_tag($tag);

  # For the moment, these legacy variances are few enough that
  #  we can just handle them here with regexps.
  $tag =~ s/^iw\b/he/i; # Hebrew
  $tag =~ s/^in\b/id/i; # Indonesian
  $tag =~ s/^[ix]-lux\b/lb/i;  # Luxemburger
  $tag =~ s/^[ix]-navajo\b/nv/i;  # Navajo
  $tag =~ s/^ji\b/yi/i;  # Yiddish
  #
  # These go FROM the simplex to complex form, to get
  #  similarity-comparison right.  And that's okay, since
  #  similarity_language_tag is the only thing that
  #  analyzes our output.
  $tag =~ s/^[ix]-hakka\b/zh-hakka/i;  # Hakka
  $tag =~ s/^nb\b/no-bok/i;  # BACKWARDS for Bokmal
  $tag =~ s/^nn\b/no-nyn/i;  # BACKWARDS for Nynorsk

  $tag =~ s/^[xiXI]-//s;
   # Just lop off any leading "x/i-"

  return "~" . uc($tag);
}

#--------------------------------------------------------------------------

=item * the function alternate_language_tags($lang1)

This function, if given a language tag, returns all language tags that
are alternate forms of this language tag.  (I.e., tags which refer to
the same language.)  This is meant to handle legacy tags caused by
the minor changes in language tag standards over the years; and
the x-/i- alternation is also dealt with.

Note that this function does I<not> try to equate new (and never-used,
and unusable)
ISO639-2 three-letter tags to old (and still in use) ISO639-1
two-letter equivalents -- like "ara" -> "ar" -- because
"ara" has I<never> been in use as an Internet language tag,
and RFC 3066 stipulates that it never should be, since a shorter
tag ("ar") exists.

Examples:

          alternate_language_tags('no-bok')       is ('nb')
          alternate_language_tags('nb')           is ('no-bok')
          alternate_language_tags('he')           is ('iw')
          alternate_language_tags('iw')           is ('he')
          alternate_language_tags('i-hakka')      is ('zh-hakka', 'x-hakka')
          alternate_language_tags('zh-hakka')     is ('i-hakka', 'x-hakka')
          alternate_language_tags('en')           is ()
          alternate_language_tags('x-mingo-tom')  is ('i-mingo-tom')
          alternate_language_tags('x-klikitat')   is ('i-klikitat')
          alternate_language_tags('i-klikitat')   is ('x-klikitat')

This function returns empty-list if given anything other than a formally
valid language tag.

=cut

my %alt = qw( i x   x i   I X   X I );
sub alternate_language_tags {
  my $tag = $_[0];
  return() unless &is_language_tag($tag);

  my @em; # push 'em real goood!

  # For the moment, these legacy variances are few enough that
  #  we can just handle them here with regexps.
  
  if(     $tag =~ m/^[ix]-hakka\b(.*)/i) {push @em, "zh-hakka$1";
  } elsif($tag =~ m/^zh-hakka\b(.*)/i) {  push @em, "x-hakka$1", "i-hakka$1";

  } elsif($tag =~ m/^he\b(.*)/i) { push @em, "iw$1";
  } elsif($tag =~ m/^iw\b(.*)/i) { push @em, "he$1";

  } elsif($tag =~ m/^in\b(.*)/i) { push @em, "id$1";
  } elsif($tag =~ m/^id\b(.*)/i) { push @em, "in$1";

  } elsif($tag =~ m/^[ix]-lux\b(.*)/i) { push @em, "lb$1";
  } elsif($tag =~ m/^lb\b(.*)/i) {       push @em, "i-lux$1", "x-lux$1";

  } elsif($tag =~ m/^[ix]-navajo\b(.*)/i) { push @em, "nv$1";
  } elsif($tag =~ m/^nv\b(.*)/i) {          push @em, "i-navajo$1", "x-navajo$1";

  } elsif($tag =~ m/^yi\b(.*)/i) { push @em, "ji$1";
  } elsif($tag =~ m/^ji\b(.*)/i) { push @em, "yi$1";

  } elsif($tag =~ m/^nb\b(.*)/i) {     push @em, "no-bok$1";
  } elsif($tag =~ m/^no-bok\b(.*)/i) { push @em, "nb$1";
  
  } elsif($tag =~ m/^nn\b(.*)/i) {     push @em, "no-nyn$1";
  } elsif($tag =~ m/^no-nyn\b(.*)/i) { push @em, "nn$1";
  }

  push @em, $alt{$1} . $2 if $tag =~ /^([XIxi])(-.+)/;
  return @em;
}

###########################################################################

{
  # Init %Panic...
  
  my @panic = (  # MUST all be lowercase!
   # Only large ("national") languages make it in this list.
   #  If you, as a user, are so bizarre that the /only/ language
   #  you claim to accept is Galician, then no, we won't do you
   #  the favor of providing Catalan as a panic-fallback for
   #  you.  Because if I start trying to add "little languages" in
   #  here, I'll just go crazy.

   # Scandinavian lgs.  All based on opinion and hearsay.
   'sv' => [qw(nb no da nn)],
   'da' => [qw(nb no sv nn)], # I guess
   [qw(no nn nb)], [qw(no nn nb sv da)],
   'is' => [qw(da sv no nb nn)],
   'fo' => [qw(da is no nb nn sv)], # I guess
   
   # I think this is about the extent of tolerable intelligibility
   #  among large modern Romance languages.
   'pt' => [qw(es ca it fr)], # Portuguese, Spanish, Catalan, Italian, French
   'ca' => [qw(es pt it fr)],
   'es' => [qw(ca it fr pt)],
   'it' => [qw(es fr ca pt)],
   'fr' => [qw(es it ca pt)],
   
   # Also assume that speakers of the main Indian languages prefer
   #  to read/hear Hindi over English
   [qw(
     as bn gu kn ks kok ml mni mr ne or pa sa sd te ta ur
   )] => 'hi',
    # Assamese, Bengali, Gujarati, [Hindi,] Kannada (Kanarese), Kashmiri,
    # Konkani, Malayalam, Meithei (Manipuri), Marathi, Nepali, Oriya,
    # Punjabi, Sanskrit, Sindhi, Telugu, Tamil, and Urdu.
   'hi' => [qw(bn pa as or)],
   # I welcome finer data for the other Indian languages.
   #  E.g., what should Oriya's list be, besides just Hindi?
   
   # And the panic languages for English is, of course, nil!

   # My guesses at Slavic intelligibility:
   ([qw(ru be uk)]) x 2,  # Russian, Belarusian, Ukranian
   'sr' => 'hr', 'hr' => 'sr', # Serb + Croat
   'cs' => 'sk', 'sk' => 'cs', # Czech + Slovak

   'ms' => 'id', 'id' => 'ms', # Malay + Indonesian

   'et' => 'fi', 'fi' => 'et', # Estonian + Finnish

   #?? 'lo' => 'th', 'th' => 'lo', # Lao + Thai

  );
  my($k,$v);
  while(@panic) {
    ($k,$v) = splice(@panic,0,2);
    foreach my $k (ref($k) ? @$k : $k) {
      foreach my $v (ref($v) ? @$v : $v) {
        push @{$Panic{$k} ||= []}, $v unless $k eq $v;
      }
    }
  }
}

=item * the function @langs = panic_languages(@accept_languages)

This function takes a list of 0 or more language
tags that constitute a given user's Accept-Language list, and
returns a list of tags for I<other> (non-super)
languages that are probably acceptable to the user, to be
used I<if all else fails>.

For example, if a user accepts only 'ca' (Catalan) and
'es' (Spanish), and the documents/interfaces you have
available are just in German, Italian, and Chinese, then
the user will most likely want the Italian one (and not
the Chinese or German one!), instead of getting
nothing.  So C<panic_languages('ca', 'es')> returns
a list containing 'it' (Italian).

English ('en') is I<always> in the return list, but
whether it's at the very end or not depends
on the input languages.  This function works by consulting
an internal table that stipulates what common
languages are "close" to each other.

A useful construct you might consider using is:

  @fallbacks = super_languages(@accept_languages);
  push @fallbacks, panic_languages(
    @accept_languages, @fallbacks,
  );

=cut

sub panic_languages {
  # When in panic or in doubt, run in circles, scream, and shout!
  my(@out, %seen);
  foreach my $t (@_) {
    next unless $t;
    next if $seen{$t}++; # so we don't return it or hit it again
    # push @out, super_languages($t); # nah, keep that separate
    push @out, @{ $Panic{lc $t} || next };
  }
  return grep !$seen{$_}++,  @out, 'en';
}

###########################################################################
1;
__END__

=back

=head1 ABOUT LOWERCASING

I've considered making all the above functions that output language
tags return all those tags strictly in lowercase.  Having all your
language tags in lowercase does make some things easier.  But you
might as well just lowercase as you like, or call
C<encode_language_tag($lang1)> where appropriate.

=head1 ABOUT UNICODE PLAINTEXT LANGUAGE TAGS

In some future version of I18N::LangTags, I plan to include support
for RFC2482-style language tags -- which are basically just normal
language tags with their ASCII characters shifted into Plane 14.

=head1 SEE ALSO

* L<I18N::LangTags::List|I18N::LangTags::List>

* RFC 3066, C<ftp://ftp.isi.edu/in-notes/rfc3066.txt>, "Tags for the
Identification of Languages".  (Obsoletes RFC 1766)

* RFC 2277, C<ftp://ftp.isi.edu/in-notes/rfc2277.txt>, "IETF Policy on
Character Sets and Languages".

* RFC 2231, C<ftp://ftp.isi.edu/in-notes/rfc2231.txt>, "MIME Parameter
Value and Encoded Word Extensions: Character Sets, Languages, and
Continuations".

* RFC 2482, C<ftp://ftp.isi.edu/in-notes/rfc2482.txt>, 
"Language Tagging in Unicode Plain Text".

* Locale::Codes, in
C<http://www.perl.com/CPAN/modules/by-module/Locale/>

* ISO 639, "Code for the representation of names of languages",
C<http://www.indigo.ie/egt/standards/iso639/iso639-1-en.html>

* ISO 639-2, "Codes for the representation of names of languages",
including three-letter codes,
C<http://lcweb.loc.gov/standards/iso639-2/bibcodes.html>

* The IANA list of registered languages (hopefully up-to-date),
C<ftp://ftp.isi.edu/in-notes/iana/assignments/languages/>

=head1 COPYRIGHT

Copyright (c) 1998-2001 Sean M. Burke. All rights reserved.

This library is free software; you can redistribute it and/or
modify it under the same terms as Perl itself.

The programs and documentation in this dist are distributed in
the hope that they will be useful, but without any warranty; without
even the implied warranty of merchantability or fitness for a
particular purpose.

=head1 AUTHOR

Sean M. Burke C<[e-mail]>

=cut