#!/usr/bin/perl -w

use strict;
use Debian::DictionariesCommon q(:all);
use Getopt::Long;
Getopt::Long::Configure("pass_through","no_auto_abbrev");

# Autoflush output buffers
$|=1;

my $class        = "ispell";
my $dictionaries = loaddb ($class);
my $emacsen;
my $regexp;
my $dryrun;

# -------------------------------------------------
sub isoconv {
# -------------------------------------------------
# Function to convert ISO-8859-1 (latin1) accented characters to
# non-accented one.  Of course, this only works for west European
# languages.  We might try to find a more general solution based on
# the current locale character set.
# -------------------------------------------------
  my $s = shift;
  $s =~ y{A-Z��������������������������������������������������}
         {a-zaeiouyaeiouyaeiouaeiouaeiouaeiouaeiouaeiosuyacdnoacnoaeoaeo};
  return $s;
}

# -------------------------------------------------
sub try_emacsen () {
# -------------------------------------------------
# Function to try getting $lang after emacsen name
# -------------------------------------------------
  my $emacsen = shift;
  return unless $emacsen;
  my %available_emacsen = ();
  foreach my $lang (keys %$dictionaries) {
    my $language = $dictionaries->{$lang};
    my $hashname = $language->{"hash-name"};
    my $emacsenname = exists $language->{"emacsen-name"} ?
      $language->{"emacsen-name"} : $hashname;
    return $lang if ( lc($emacsen) eq lc($emacsenname) );
    $available_emacsen{$emacsenname}++;
  }
  die "$0: \"$emacsen\" does not match any available emacs dict value:\n  "
    . join("\n  ", sort keys %available_emacsen) . "\n";
}

# -------------------------------------------------
sub try_regexp () {
# -------------------------------------------------
# Function to try getting $lang after $regexp
# -------------------------------------------------
  my $regexp = shift;
  return unless $regexp;
  my $guess;
  my %regexp_matches    = ();

  $regexp = isoconv ($regexp);
  foreach my $key ( keys %$dictionaries ) {
    $_ = isoconv ( $key );
    if ( /$regexp/ ) {
      $regexp_matches{$key}++;
      $guess = $key;
    }
  }

  die "$0: No installed language matched `$regexp'\n" unless $guess;

  if ( scalar keys %regexp_matches == 1) {
    return $guess;
  } else {
    die ("$0: More than one installed languages matched `$regexp':\n  "
	 . join ("\n  ", sort keys %regexp_matches) . "\n");
  }
}

# --------------------------------------------------------------------
# Now the main program
# --------------------------------------------------------------------

GetOptions ('emacs=s'    => \$emacsen,
	    'language=s' => \$regexp,
	    'dry-run'    => \$dryrun);

die " ispell-wrapper is a wrapper to ispell, but ispell is not installed.\n"
  unless ( -x "/usr/bin/ispell" );

$regexp = $ENV{ISPELLDEFAULT} unless $regexp;

$regexp =~ s/([^\\]|^)(\(|\))/$1\\$2/g if $regexp;   # Make sure () are escaped

# In the POD section below there is an extensive description on the
# priority order for determining the ispell language.
my $lang = &try_emacsen($emacsen)
  ||  &try_regexp ($regexp)
  ||  &getuserdefault ()
  ||  &dico_getsysdefault ("ispell");

print STDERR " Warning: --language=$regexp will be overriden by
          --emacs=$emacsen setting\n\n"
  if ( defined $lang && $regexp && $emacsen );

my $ispell_wrapper_args = "";
$ispell_wrapper_args = dico_get_spellchecker_params($class,$dictionaries->{$lang})
  if ( $lang && defined $dictionaries->{$lang});

# Ignore $lang results if -d is explicitly set from commandline

foreach ( @ARGV ) {
  if ( /^\-d/ ){
    $ispell_wrapper_args = "";
    last;
  }
}

my $cli_opts = join(' ',@ARGV);

print STDERR "Warning: \'$lang\' values overriden with \'$cli_opts\'\n"
  if ( not $ispell_wrapper_args && defined $lang );

my $command_to_run = "ispell $ispell_wrapper_args $cli_opts";

if ( $dryrun ){
  print "--\n$command_to_run\n--\n";
} else {
  exec $command_to_run;
}

# Local Variables:
# perl-indent-level: 2
# End:

__END__

=head1 NAME

B<ispell-wrapper> - smart wrapper for ispell

=head1 SYNOPSIS

 ispell-wrapper [--emacs=name] [--language=regexp] [--dry-run] [ispell options] file

   Options (all long only options):
    --emacs=name           Set the language to use by emacs dict name
    --language=regexp      Set the language to use by name
    --dry-run              Only show what would have done

=head1 DESCRIPTION

B<ispell-wrapper> is a wrapper script for ispell intended to be used
in a Debian system in conjunction with the infrastructure introduced by
the dictionaries-common package. Option --dry-run will show the string
to be run without doing anything else.

It automatically sets the B<-d>, B<-w>, and B<-T> options to ispell as a
function of the chosen language.  Of course, this only works for dictionary
packages that comply with the above mentioned Policy.

Here is how the language is defined (in order of priority):

=over

=item 1)

By matching the emacs dict name given in --emacs option to the name of
one of the emacs dicts names provided by installed languages in the
system. This match must be exact (although is case insensitive).
Note that this will override any value given in the --language option.

=item 2)

By matching the regexp given in option --language to the list of
installed languages in the system.

=item 3)

By matching the regexp stored in the environment variable
ISPELLDEFAULT to the list of installed languages in the system.

=item 4)

By using the value stored in the user-specific file ~/.ispell-default
(use select-default-iwrap(1) to set it).

=item 5)

By using the value stored in the site-wide file
/etc/dictionaries-common/ispell-default (use select-default-ispell(8)
as superuser to set it).

=back

Note: regexp matches are case-insensitive and the ISO-8859-1 special
characters are transformed into their ASCII equivalents.  German
ess-zet is equivalent to the character "s" and the ae ligature to the
character "e".

=head1 EXAMPLE

Let us say that the following dictionaries are installed in the system
(as appearing in the Debconf question at installation time):

    castellano (Spanish TeX mode)
    castellano8 (Spanish 8 bit)
    portuguE<ecirc>s (European Portuguese)
    portuguE<ecirc>s brasileiro (Brazilian Portuguese)

Choosing the regexp (either in the --language option or in the
environment variable ISPELLDEFAULT) to be "span" will yield an error,
since two languages will match ("castellano" and "castellano8").
However, if the regexp is "span.*8", the language "castellano8
(Spanish 8 bit)" will be chosen.

=head1 ENVIRONMENT

=over

=item ISPELLDEFAULT

Regexp that matches the name of the default language to use, if no
--language option is given.

=back


=head1 FILES

=over

=item $HOME/.ispell-default

Contains the name of the language to use, if no --language option is
given or if the ISPELLDEFAULT environment variable is not set.  This
is a user-specific choice.

=item /etc/dictionaries-common/ispell-default

Name of the language to use when everything above is not set. This is
a system-wide setting.

=back


=head1 SEE ALSO

select-default-ispell(8), select-default-iwrap(1)

=head1 AUTHORS

Rafael Laboissiere

=cut