package Encode::CN::HZ;

use strict;

use vars qw($VERSION);
$VERSION = do { my @r = (q$Revision: 2.1 $ =~ /\d+/g); sprintf "%d."."%02d" x $#r, @r };

use Encode qw(:fallbacks);

use base qw(Encode::Encoding);
__PACKAGE__->Define('hz');

# HZ is a combination of ASCII and escaped GB, so we implement it
# with the GB2312(raw) encoding here. Cf. RFCs 1842 & 1843.

# not ported for EBCDIC.  Which should be used, "~" or "\x7E"?

sub needs_lines  { 1 }

sub decode ($$;$)
{
    my ($obj,$str,$chk) = @_;

    my $GB = Encode::find_encoding('gb2312-raw');
    my $ret = '';
    my $in_ascii = 1; # default mode is ASCII.

    while (length $str) {
	if ($in_ascii) { # ASCII mode
	    if ($str =~ s/^([\x00-\x7D\x7F]+)//) { # no '~' => ASCII
		$ret .= $1;
		# EBCDIC should need ascii2native, but not ported.
	    }
	    elsif ($str =~ s/^\x7E\x7E//) { # escaped tilde
		$ret .= '~';
	    }
	    elsif ($str =~ s/^\x7E\cJ//) { # '\cJ' == LF in ASCII
		1; # no-op
	    }
	    elsif ($str =~ s/^\x7E\x7B//) { # '~{'
		$in_ascii = 0; # to GB
	    }
	    else { # encounters an invalid escape, \x80 or greater
		last;
	    }
	}
	else { # GB mode; the byte ranges are as in RFC 1843.
	    if ($str =~ s/^((?:[\x21-\x77][\x21-\x7E])+)//) {
		$ret .= $GB->decode($1, $chk);
	    }
	    elsif ($str =~ s/^\x7E\x7D//) { # '~}'
		$in_ascii = 1;
	    }
	    else { # invalid
		last;
	    }
	}
    }
    $_[1] = '' if $chk; # needs_lines guarantees no partial character
    return $ret;
}

sub cat_decode {
    my ($obj, undef, $src, $pos, $trm, $chk) = @_;
    my ($rdst, $rsrc, $rpos) = \@_[1..3];

    my $GB = Encode::find_encoding('gb2312-raw');
    my $ret = '';
    my $in_ascii = 1; # default mode is ASCII.

    my $ini_pos = pos($$rsrc);

    substr($src, 0, $pos) = '';

    my $ini_len = bytes::length($src);

    # $trm is the first of the pair '~~', then 2nd tilde is to be removed.
    # XXX: Is better C<$src =~ s/^\x7E// or die if ...>?
    $src =~ s/^\x7E// if $trm eq "\x7E";

    while (length $src) {
	my $now;
	if ($in_ascii) { # ASCII mode
	    if ($src =~ s/^([\x00-\x7D\x7F])//) { # no '~' => ASCII
		$now = $1;
	    }
	    elsif ($src =~ s/^\x7E\x7E//) { # escaped tilde
		$now = '~';
	    }
	    elsif ($src =~ s/^\x7E\cJ//) { # '\cJ' == LF in ASCII
		next;
	    }
	    elsif ($src =~ s/^\x7E\x7B//) { # '~{'
		$in_ascii = 0; # to GB
		next;
	    }
	    else { # encounters an invalid escape, \x80 or greater
		last;
	    }
	}
	else { # GB mode; the byte ranges are as in RFC 1843.
	    if ($src =~ s/^((?:[\x21-\x77][\x21-\x7F])+)//) {
		$now = $GB->decode($1, $chk);
	    }
	    elsif ($src =~ s/^\x7E\x7D//) { # '~}'
		$in_ascii = 1;
		next;
	    }
	    else { # invalid
		last;
	    }
	}

	next if ! defined $now;

	$ret .= $now;

	if ($now eq $trm) {
	    $$rdst .= $ret;
	    $$rpos = $ini_pos + $pos + $ini_len - bytes::length($src);
	    pos($$rsrc) = $ini_pos;
	    return 1;
	}
    }

    $$rdst .= $ret;
    $$rpos = $ini_pos + $pos + $ini_len - bytes::length($src);
    pos($$rsrc) = $ini_pos;
    return ''; # terminator not found
}


sub encode($$;$)
{
    my ($obj,$str,$chk) = @_;

    my $GB = Encode::find_encoding('gb2312-raw');
    my $ret = '';
    my $in_ascii = 1; # default mode is ASCII.

    no warnings 'utf8'; # $str may be malformed UTF8 at the end of a chunk.

    while (length $str) {
	if ($str =~ s/^([[:ascii:]]+)//) {
	    my $tmp = $1;
	    $tmp =~ s/~/~~/g; # escapes tildes
	    if (! $in_ascii) {
		$ret .= "\x7E\x7D"; # '~}'
		$in_ascii = 1;
	    }
	    $ret .= pack 'a*', $tmp; # remove UTF8 flag.
	}
	elsif ($str =~ s/(.)//) {
	    my $s = $1;
	    my $tmp = $GB->encode($s, $chk);
	    last if !defined $tmp;
	    if (length $tmp == 2) { # maybe a valid GB char (XXX)
		if ($in_ascii) {
		    $ret .= "\x7E\x7B"; # '~{'
		    $in_ascii = 0;
		}
		$ret .= $tmp;
	    }
	    elsif (length $tmp) { # maybe FALLBACK in ASCII (XXX)
		if (!$in_ascii) {
		    $ret .= "\x7E\x7D"; # '~}'
		    $in_ascii = 1;
		}
		$ret .= $tmp;
	    }
	}
	else { # if $str is malformed UTF8 *and* if length $str != 0.
	    last;
	}
    }
    $_[1] = $str if $chk;

  # The state at the end of the chunk is discarded, even if in GB mode.
  # That results in the combination of GB-OUT and GB-IN, i.e. "~}~{".
  # Parhaps it is harmless, but further investigations may be required...

    if (! $in_ascii) {
	$ret .= "\x7E\x7D"; # '~}'
	$in_ascii = 1;
    }
    return $ret;
}

1;
__END__

=head1 NAME

Encode::CN::HZ -- internally used by Encode::CN

=cut