#Japanese dictionary-order sort
#
#0. ignore non-kana
#1. fold katakana to hiragana (if identical, hiragana comes first)
#2. replace tsu characters with base + N,
#   where N is 0 for standard, 1 for small, 2 for voiced
#3. replace voice-able chars with base + N.
#
#bydict is a sort function for arrays of hashes, each of which
#is assumed to contain a key SORTKEY, consisting of a [jstring,num]
#pair. All non-hiragana, non-katakana characters in jstring are
#ignored, the above modifications are applied, and a key named
#_sortkey is created, which contains the processed output. During
#comparison, num is used as the tie-breaker, so that there's some
#consistent way to order common words like "ai".
#
use strict;
package JDictSort;

my %S = (
	"\x{304b}" => ["\x{304b}","0"], # ka
	"\x{304c}" => ["\x{304b}","1"],
	"\x{304d}" => ["\x{304d}","0"], # ki
	"\x{304e}" => ["\x{304d}","1"],
	"\x{304f}" => ["\x{304f}","0"], # ku
	"\x{3050}" => ["\x{304f}","1"],
	"\x{3051}" => ["\x{3051}","0"], # ke
	"\x{3052}" => ["\x{3051}","1"],
	"\x{3053}" => ["\x{3053}","0"], # ko
	"\x{3054}" => ["\x{3053}","1"],
	"\x{3055}" => ["\x{3055}","0"], # sa
	"\x{3056}" => ["\x{3055}","1"],
	"\x{3057}" => ["\x{3057}","0"], # shi
	"\x{3058}" => ["\x{3057}","1"],
	"\x{3059}" => ["\x{3059}","0"], # su
	"\x{305a}" => ["\x{3059}","1"],
	"\x{305b}" => ["\x{305b}","0"], # se
	"\x{305c}" => ["\x{305b}","1"],
	"\x{305d}" => ["\x{305d}","0"], # so
	"\x{305e}" => ["\x{305d}","1"],
	"\x{305f}" => ["\x{305f}","0"], # ta
	"\x{3060}" => ["\x{305f}","1"],
	"\x{3061}" => ["\x{3061}","0"], # chi
	"\x{3062}" => ["\x{3061}","1"],
	"\x{3066}" => ["\x{3066}","0"], # te
	"\x{3067}" => ["\x{3066}","1"],
	"\x{3068}" => ["\x{3068}","0"], # to
	"\x{3069}" => ["\x{3068}","1"],
	"\x{306f}" => ["\x{306f}","0"], # ha
	"\x{3070}" => ["\x{306f}","1"],
	"\x{3071}" => ["\x{306f}","2"],
	"\x{3072}" => ["\x{3072}","0"], # hi
	"\x{3073}" => ["\x{3072}","1"],
	"\x{3074}" => ["\x{3072}","2"],
	"\x{3075}" => ["\x{3075}","0"], # fu
	"\x{3076}" => ["\x{3075}","1"],
	"\x{3077}" => ["\x{3075}","2"],
	"\x{3078}" => ["\x{3078}","0"], # he
	"\x{3079}" => ["\x{3078}","1"],
	"\x{307a}" => ["\x{3078}","2"],
	"\x{307b}" => ["\x{307b}","0"], # ho
	"\x{307c}" => ["\x{307b}","1"],
	"\x{307d}" => ["\x{307b}","2"],
	"\x{3064}" => ["\x{3064}","0"], #tsu
	"\x{3063}" => ["\x{3064}","1"], #small-tsu
	"\x{3065}" => ["\x{3064}","2"], #dzu
);

sub bydict {
	my ($a,$b) = ($main::a,$main::b);
	$a->{_sortkey} = _getkey($a->{SORTKEY}) unless $a->{_sortkey};
	my $akey = $a->{_sortkey};
	$b->{_sortkey} = _getkey($b->{SORTKEY}) unless $b->{_sortkey};
	my $bkey = $b->{_sortkey};
	return $akey->[2] <=> $bkey->[2] if $akey->[0] eq $bkey->[0]
		and $akey->[1] eq $bkey->[1];
	return $akey->[1] cmp $bkey->[1] if $akey->[0] eq $bkey->[0];
	return $akey->[0] cmp $bkey->[0];
}

sub _getkey {
	my ($aref) = @_;
	my ($w,$n) = @$aref;
	my $k = '';
	my $k2 = '';
	$w =~ s/^[^\t]*\p{InCJKUnifiedIdeographs}[^\t]*\t//;
	foreach (split(//,$w)) {
		last if /^\s/;
		next unless /^(\p{InHiragana}|\p{InKatakana})$/;
		if (/^\p{InKataKana}$/) {
			$_ = chr(ord($_)-96);
			$k2 .= "k";
		}else{
			$k2 .= "h";
		}
		if (ref $S{$_}) {
			$k .= $S{$_}->[0];
			$k2 .= $S{$_}->[1];
		}else{
			$k .= $_;
		}
	}
	return [$k,$k2,$n];
}

1;