#!/usr/bin/ruby -Ke
#
#cidbushu2.txt:
#1505,乾 L,常用          v,lr,十日十,乞 m,ur,十日十人,乙 ##≒十日十人
#↑v1参照 ↑v1での'l'
#
#会意文字の中に人が2個並んで入っていた時、それが从なのかどうかは解釈の問題
#中心となる分解データは意味向きから字形合成向きまで両方入るように
#
#意味向き
#部品が字だったらその場で分解しないように　必要なら篆書にもリンク張る
#合成オペレータの正確性は求めない
#
#字形合成向き
#勲や哀を一発で合成できるスペックは求めない 積極的にべた書き
#
#意味:m 意味分解困難:m! 字形:v 自動合成困難:v!
#パーツ範囲制限:l 異体字ルート:i m?:解字未調査 パーツ保留:v?
#mv: 普通に考えて正解 m!: パーツ提示しない m: 字形用としては不適、異体字との区別は考えない
#v: 意味用としては不適 mv!, v!: パーツは揃えたが自動合成困難
#哀 mv!,oi,衣,口 v,ud,亠,ud,口,衣の下
#
#bushu.rev用優先順位
#i l mv mv! v v! m
#
#CLWFK,CHISE用優先順位
#mv v mv! v! m
#
#1175,尉 L,常用          v.23,lr,oi,尸,示,寸 ##会意
#                         ↑bushu.rev用出力時には"尉示(2番目)寸(3番目)"となる
#
#lr,ud,oi,lu,ur,ld,dr,ps,+-,mm,xx
#
#ku,mo,ka,ha,ta,tu,ny   これらは内部ではoiとして扱われる
#F4,F5,F6,F7,F8,F9,FA(U+2F)
#
##旧フォーマット（v1）説明
#cidbushu.txt:
#1505,乾 l 常用          b l十日十     r乞         a u十日十人   r乙
#↑
#第1フィールド: 合成後どうなるか。コンマ区切りでいくらでも。
#               すぺて同一の字形を指し示すための表現である
# 4桁以上の生数字: Adobe-Japan1(-6)におけるCID
# U+〜〜: Unicode@<del>MSゴシック字形</del><ins>小塚明朝字形</ins>
# その他文字コード表現: （略）
# 生漢字1字{@90JIS}: その生漢字のバイト列が指すコードポイントの
#                    JIS X 0208:1990での例示字形
# 生漢字1字@78JIS: 前略JIS X 0208:1978後略
# 生漢字1字@04JIS: 前略JIS X 0213:2004後略
# その他（3桁以下の半角数字なども含む）: JIS X 0208:1990に入って
#                    いないグリフに適宜つけた通称
#
#それより後ろ: まずスペース区切り
#l: 次のフィールドと合わせて法律的情報
#a: 分けかたは合ってる&&字形をある程度表現できてる(括弧→字形を表現できてない主要因）
#b: 分けかたは間違ってる&&字形をある程度表現できてる
#c: 正しい分けかた&&意味記述（字形が表現できてない）または異体字情報
#!c: 分割不能な象形文字等 フィールドを消費しない
#その他半角英字や'?'など1文字: 作業中である。とりあえずbと同等に扱うこと
#
#cがなくてaがあるのは
#1. aが分けかた字形意味全てにおいて完璧
#2. aで括弧ついてないのは意味も合ってる（それで括弧つきのほうは一言じゃ説明できない）
#   （括弧つきのが正しい意味ならcに置けばいい）
#aと!c両方あるのは分割不能かどうか微妙
#c =?はその先が本字 たどっていくと意味が合ってる解字が出てくるはず
#c /?は自分が借字となってその字の意味を取り込んでしまったパターン
#c <?は自分がその字の省略形でありしかもその字の中に出てくる（）
#c >?はその字の俗字だったものが別の意味を獲得（のっとられ側の字がない借字?）
#c |?はどっちがえらいとも言い難い同音同義別解字（"異体""別体"もここに含むべき）
#（"通じる"は/?か|?か微妙）
#
#a-cの後ろに任意長さの分解定義が続く
#a (位置を表す英字1文字)(グリフを指し示す表現) ()() ... b ()() ...
#作業中なので位置指定子が記述されていない場合がある
#l&r: 偏と旁    u&d: 冠と脚    o&i: 構や垂や繞、あとむりやり間に挟める
#p&s: 品口3の口がp(arts)で3がs(tructure)  +&-: 引き算
#lまたはrとuまたはdの組み合わせは直線で切り分けられるのにしてはいけないパターン
#http://www.itscj.ipsj.or.jp/ipsj-ts/02-02/ips_charid/toc.htm
#↑のP4〜P7に近い感じ
#
#
#
#
#http://www2.odn.ne.jp/alt-quinon/files/ptex/x0213/jx2004tbl.pdf
#http://pc5.2ch.net/test/read.cgi/unix/1082032043/183
#http://www.taishukan.co.jp/kanji/archive/jinmei_minaoshi.html

def usage
  $stderr.print "ruby #{$0} (--format|--sort|--to-rev{-tc}|--to-thunder|--to-clwfk|--to-chise|--for-tex) cidbushu.txt\n"
  $stderr.print "\n"
  $stderr.print "convert from old(-200603) cidbushu.txt\n"
  $stderr.print "ruby #{$0} --from-v1-path1 cidbushu.txt | ruby #{$0} --from-v1-path2 > cidbushu2.txt\n"
end

      def treeize(pn)
        st = []
        pn.reverse.each {|pe|
          if pe =~ /^[a-z\+\-]+$/
            raise RuntimeError, 'few parts' if st.length < 2
            l = st.pop
            r = st.pop
            st.push([pe, l, r])
          else
            st.push(pe)
          end
        }
        raise RuntimeError, 'too many parts' if st.length != 1
        return st[0]
      end
      def to_oi(op)
        case op
        when 'ku','mo','ka','ha','ta','tu','ny'
          return 'oi'
        else
          return op
        end
      end


opt = ARGV.shift

if opt == '--from-v1-path1'

  while gets
    chop!
    if $_[0..0] == "#"
      puts $_
      next
    end
    $_ << '   '
    r = $_.split(/\s+/)
    if r.length > 2 && r[1] == 'l'
      print r[0], ' L,', r[2], ' ' * ([21-r[0].length-r[2].length, 1].max-1)
      r.shift
      r.shift
    else
      print r[0], ' ' * ([24 - r[0].length,1].max-1)
    end
    r.shift
    while r.length > 0 && r[0] !~ /^\#/
      if r[0] =~ /^([a-z\?]|\Sc)$/
        case r[0]
        when 'a'
          r[0] = 'mv'
        when 'b'
          r[0] = 'v'
        when 'c'
          r[0] = 'm'
        when 'i'
          r[0] = 'i'
        when 'd'
          r[0] = 'HOGEd'
        when '?'
          r[0] = 'HOGE?'
        when '!c'
          r[0] = 'm!'
        else
          r[0] = 'HOGE' + r[0]
          #p r
          #exit 0;
        end
        print ' ', r[0]
        r.shift
      else
	#r[0] = ' ' + r[0] if r[0] !~ /^[a-z\+\-\=\/\<\>\|]/
	print ',', r[0]
	r.shift
      end
    end
    print ' ', r.join(' ') if r.length > 0
    print "\n"
  end

elsif opt == '--from-v1-path2'

  while gets
    chop!
    if $_[0..0] == "#"
      puts $_
      next
    end
    $_ << '   '
    r = $_.split(/\s+/)
    if r.length > 1 && r[1][0..0] == 'L'
      print r[0], ' ', r[1], ' ' * ([23-r[0].length-r[1].length, 1].max-1)
      r.shift
    else
      print r[0], ' ' * ([24 - r[0].length,1].max-1)
    end
    r.shift
    while r.length > 0 && r[0] !~ /^\#/
      s = r[0].split(/,/)
      if s.length == 1
        print ' ', r[0]
        r.shift
        next
      end
      if s[1..-1].find {|i| i =~ /\(/ }
        s[0] += '?' if s[0] !~ /\?/
      end
      s1 = s[1..-1].collect {|i| i[0..0] }.join
      if s1.length == 1
        print ' ', s[0], ',', s1, ',', s[1][1..-1]
        r.shift
        next
      end
      h12 = {'u' => 'd', 'l' => 'r', 'o' => 'i'}
      h21 = {'d' => 'u', 'r' => 'l', 'i' => 'o'}
      case s1
      when 'lr', 'ud', 'oi', 'lu', 'ld', 'ur', 'dr', 'ps', '+-'
        #OK
      when 'ou','od','di','so'
        #OK
      when 'xx', 'mm'
        #OK
      when 'rl', 'du', 'io', 'ul', 'dl', 'ru', 'rd', 'sp','uo'
        s[0] += '.21' #if s[0] !~
        s[1], s[2] = s[2], s[1]
      when 'dmm'
        s[0] += '.31'
        s[1], s[2], s[3] = s[2], s[3], s[1]
      else
        if s1.length == 3
          if h12[s1[0..0]] == s1[1..1]
            if h12[s1[2..2]] #ex. 'udl'=>'lud'
              s[0] += '.23'
              s[1], s[2], s[3] = s[3], s[1], s[2]
            else #ex. 'udr'=>'udr'
              #OK
            end
          elsif h21[s1[0..0]] == s1[1..1]
            if h12[s1[2..2]] #ex. 'dul'=>'lud'
              s[0] += '.32'
              s[1], s[2], s[3] = s[3], s[2], s[1]
            else #ex. 'dur'=>'udr'
              s[0] += '.21'
              s[1], s[2], s[3] = s[2], s[1], s[3]
            end
          elsif h12[s1[0..0]] == s1[2..2]
            if h12[s1[1..1]] #ex. 'uld'=>'lud'
              s[0] += '.21'
              s[1], s[2], s[3] = s[2], s[1], s[3]
            else #ex. 'urd'=>'udr'
              s[0] += '.13'
              s[1], s[2], s[3] = s[1], s[3], s[2]
            end
          elsif h21[s1[0..0]] == s1[2..2]
            if h12[s1[1..1]] #ex. 'dlu'=>'lud'
              s[0] += '.31'
              s[1], s[2], s[3] = s[2], s[3], s[1]
            else #ex. 'dru'=>'udr'
              s[0] += '.23'
              s[1], s[2], s[3] = s[3], s[1], s[2]
            end
          elsif h12[s1[1..1]] == s1[2..2]
            if h12[s1[0..0]] #ex. 'lud'=>'lud'
              #OK
            else #ex. 'rud'=>'udr'
              s[0] += '.31'
              s[1], s[2], s[3] = s[2], s[3], s[1]
            end
          elsif h21[s1[1..1]] == s1[2..2]
            if h12[s1[0..0]] #ex. 'ldu'=>'lud'
              s[0] += '.13'
              s[1], s[2], s[3] = s[1], s[3], s[2]
            else #ex. 'rdu'=>'udr'
              s[0] += '.32'
              s[1], s[2], s[3] = s[3], s[2], s[1]
            end
          end
        end
      end
      s1 = s[1..-1].collect {|i| i[0..0] }.join
      case s1
      when 'lr', 'ud', 'oi', 'lu', 'ld', 'ur', 'dr', 'ps', '+-'
        #OK
      when 'ou','od','di','so'
        #OK
      when 'xx', 'mm'
        s[0].sub!(/v/, 'v!') if s[0] !~ /v\!/
      when 'mmd'
        s[0].sub!(/v/, 'v!') if s[0] !~ /v\!/
        s[1], s[2] = ['umm', s[1][1..-1], s[2][1..-1]].join(','), s[3]
        s.pop
        s1 = s[1..-1].collect {|i| i[0..0] }.join
      when /^lr[di]$/, /^ud[ri]$/, /^oi[rd]$/
        ss = h21[s1[2..2]] + s1[0..1]
        s[1], s[2] = [ss, s[1][1..-1], s[2][1..-1]].join(','), s[3]
        s.pop
        s1 = s[1..-1].collect {|i| i[0..0] }.join
      when /^[uo]lr$/, /^[lo]ud$/, /^[lu]oi$/
        ss = h12[s1[0..0]] + s1[1..-1]
        s[1], s[2] = s[1], [ss, s[2][1..-1], s[3][1..-1]].join(',')
        s.pop
        s1 = s[1..-1].collect {|i| i[0..0] }.join
      else
        p s
        exit
      end
      l = ([s1] + s[1..-1].collect {|i| i[1..-1] })
      l = treeize(l)
#
      def eval_oi(nd)
        return if nd.class == String
        if nd[0] == 'oi' && nd[1].class == String
          case nd[1]
          when /^(广|厂|やまいだれ|尸|在の外|虍|广廿|厂林|厂イ|雁たれ)$/
            nd[0].replace 'ta'
          when /^(囗四)$/
            nd[0].replace 'ku'
          when /^(門|冂|岡ひく山|風|微かまえ|戊|戌)$/
            nd[0].replace 'mo'
          when /^(勹|裁かまえ|武ひく止|)$/
            nd[0].replace 'tu'
          when /^(匚|匸)$/
            nd[0].replace 'ha'
          when /^(凵)$/
            nd[0].replace 'ka'
          when /^(之|え|廴|走|夂|九)$/
            nd[0].replace 'ny'
          when /^(行|衣)$/
          else
          end
        end
        eval_oi(nd[1])
        eval_oi(nd[2])
      end
#
      eval_oi(l)
      l = l.flatten
      print ' ', s[0], ',', l.join(',')
      r.shift
    end
    print ' ', r.join(' ') if r.length > 0
    print "\n"
  end



elsif opt == '--format'

  while gets
    chop!
    if $_[0..0] == "#"
      puts $_
      next
    end
    $_ << '   '
    r = $_.split(/\s+/)
    if r.length > 1 && r[1][0..0] == 'L'
      print r[0], ' ', r[1], ' ' * ([23-r[0].length-r[1].length, 1].max-1)
      r.shift
    else
      print r[0], ' ' * ([24 - r[0].length,1].max-1)
    end
    r.shift
    while r.length > 1 && r[0] !~ /^\#/
      print ' ', r[0]
      r.shift
    end
    print ' ', r.join(' ') if r.length > 0
    print "\n"
  end

elsif opt == '--sort' && nil #v1
  
  ali = Hash.new

  def ali.[](key)
    return key if ! self.has_key?(key)
    super
  end

  fp = open('cidbushu.alias')
  while fp.gets
    chop!
    next if $_ =~ /^\s*\#/
    gsub!(/\#.*$/, '')
    gsub!(/\s*$/, '')
    next if $_ =~ /^(delete|overwrite|protect)/
    r = $_.split(/\s+/)
    next if r.length <= 1
    r.shift if r[0] =~ /^(order-(strong|weak)|weak)/
    d = r.shift
    r.each {|i|
      ali[i] = d if i.length > 2
    }
  end
  fp.close

  d = Array.new
  while gets
    chop!
    r = $_.split(/\s+/)
    if r.length < 1 || r[0] =~ /^\#/
      puts $_
      next
    end
    r = r[0]
    r = r.split(/\,/)
    r.delete_if {|i| i =~ /^\d{4,}$/ }
    r = r[0]
    #p r
    d.push([r, $_])
    if $_ =~ /c \=(\S+)/ && $1.length > 2
      ali[$1] = r
      #p r, $1
    end
  end

  d.each {|i| i.unshift(ali[ali[ali[i[0]]]].gsub(/^(常用|旧|たて|よこ|たれ|にょう)(.)/) { $2 }) }
  d.sort! {|i, j|
    [i[0][0..1], i[1][0..2], j[1][2..-1]] <=> [j[0][0..1], j[1][0..2], i[1][2..-1]]
  }
  d.each {|i|
    #p i[0..1]
    puts i[2]
  }
  
elsif opt == '--to-expand'
elsif opt =~ /^--to-rev/ || opt == '--to-thunder'

  PREFER_REGULAR = false
  if opt == '--to-thunder'
    prefer = <<EOS
日月火初足
木金土ネ車
口宀  竹シ
サイ目才院
石言糸性山
禾食王示革
耳穴  雨冫
四彳舟独酉
羊大田虫攵
木广厂リ力
口貝  囗門
隹馬米之部
女巾立心十
欠病尸方弓
歹頁  匚行
見牛魚え鳥
EOS
    prefer = prefer.split(/\s*/).uniq
  end

  rev = {}  # [composed, class] => [[pos, part], ...]
  order_weak = []
  weak = {}
  ali = {}
  delp = []
  dels = []
  def ali.[](*key)
    if key.length == 1
      key = key[0]
      return key if ! self.has_key?(key)
      super
    elsif key.length == 2
      pos = key[1]
      key = key[0]
      if pos =~ /^[亜-腕]$/ && self.has_key?(key+'(inJIS1)')
        return super(key+'(inJIS1)')
      elsif pos =~ /^[弌-熙]$/ && self.has_key?(key+'(inJIS2)')
        return super(key+'(inJIS2)')
      end
      return key if ! self.has_key?(key)
      super(key)
    end
  end

  while gets
    chop!
    next if $_ =~ /^\s*\#/
    gsub!(/\#.*$/, '')
    gsub!(/\s*$/, '')
    r = $_.split(/\s+/)
    l = r[0]
    l = l.split(/\,/)
    l.delete_if {|i| i =~ /^\d{4,}$/ }
    l.each {|i|
      ali[i] = l[0].gsub(/\@90JIS$/, '')
    }
    composed = l[0].gsub(/\@90JIS$/, '')
    r.shift

    r.each_index {|i|
      parts = []
      if r[i][0..0] == 'L'
        next
      end
      l = r[i].split(/\,/)
      c = l[0]
      next if l.length <= 1
      if l[1] =~ /^[\=\/\<\>\|]$/
        next
      end
      l.shift
      l = l.collect {|j| j.gsub(/\((.*)\)$/) { $1 } }  #remove parenthesis
#
      def subparts(pn)
        #p pn
        return [[pn]] if pn.class == String
        pn0 = to_oi(pn[0])
        return subparts(pn[1]).collect {|br| [pn0[0..0]] + br } \
                + subparts(pn[2]).collect {|br| [pn0[1..1]] + br }
      end
#
      l = treeize(l)
      m = subparts(l)
      #p m
      m.each {|br|
	br = br[-2..-1]
	parts << br
      }
      if c =~ /\.\d\d/
        i1 = $&[1..1].to_i-1
        i2 = $&[2..2].to_i-1
        parts2 = [parts[i1], parts[i2]]
        (i1, i2) = [i1, i2].sort
        parts.delete_at(i2)
        parts.delete_at(i1)
        parts = parts2 + parts
        c.gsub!(/\.\d\d/, '')
      end
      c.gsub!(/^HOGE/, '')
      c.gsub!(/(.)\?/) { $1 }
      c = {'mv' => 'a', 'v' => 'b', 'm' => 'c', 'mv!' => 'e', 'v!' => 'f'}[c] || c
      if c.length != 1
        $stderr.puts [composed, c].inspect
        exit 1
      end
      if rev.has_key?([composed, c])
        $stderr.print "repeated composition definition:\n"
        $stderr.print "#{[composed, c].inspect} => #{rev[[composed, c]].inspect}\n#{$_}\n"
        next
      end
      rev[[composed, c]] = parts
      #p [composed, c, parts]
    }

  end

  fp = open('cidbushu.alias')
  prot = []
  while fp.gets
    chop!
    $stderr.print "#{$_}\n"
    next if $_[0..0] == '#'
    gsub!(/\#.*$/, '')
    gsub!(/\s*$/, '')
    r = $_.split(/\s+/)
    case r[0]
    when 'delete'
      for i in 1...r.length
        for j in 'abcdefijklmn?'.split(//)
          next unless rev.has_key?([r[i], j])
          parts = rev.delete([r[i], j])
        end
      end
    when 'overwrite'
      for j in 'abcdefijklmn?'.split(//)
        next unless rev.has_key?([r[1], j])
        parts = rev.delete([r[1], j])
      end
      parts = r[2..-1]
      parts.each_index {|i|
	l = parts[i]
	l = ' ' + l if l !~ /^[a-z\+\-]/
	l = [l[0..0], ali[ali[ali[ali[l[1..-1], r[1]], r[1]], r[1]], r[1]]]
	parts[i] = l
      }
      rev[[r[1], 'b']] = parts
    when 'protect'
      prot = r[1..-1]
#    when 'order-swap'
#      for i in 'abcdefijklmn?'.split(//)
#        next unless rev.has_key?([r[1], i])
#        parts = rev[[r[1], i]]
#        parts[0..1] = [parts[1], parts[0]]
#        rev[[r[1], i]] = parts
#      end
#      rev.each {|cc, parts|
#        next unless parts[0..1].collect {|i| i[1] }.include?(r[1])
#	parts[0..1] = [parts[1], parts[0]]
#	rev[cc] = parts
#      }
###
#    when 'try-making-malanalysis'
###
    when 'entry-prefix-will-be-deleted'
      delp += r[1..-1]
    when 'entry-suffix-will-be-deleted'
      dels += r[1..-1]
    else
      frec = nil
      if r[0] == 'order-strong'
        r.shift
        frec = :ostrong
      elsif r[0] == 'order-weak'
        r.shift
        frec = :oweak
      elsif r[0] == 'weak'
        r.shift
        frec = :weak
      end
      for j in 1...r.length  #後からoverwriteする場合の救済措置
        ali[r[j]] = r[0]
      end
      rev.each {|cc, parts|
        #next unless parts.collect {|i| i[1] }.include?(r[1])
        fcr = false
        for j in 1...r.length
          parts.each {|k|
            rj = r[j].dup
            fquit = false
            if r[j] =~ /\(inJIS1\)$/
              rj.gsub!(/\(inJIS1\)$/, '')
              #$stderr.puts cc.inspect
              next unless cc[0] =~ /^[亜-腕]$/
            elsif r[j] =~ /\(inJIS2\)$/
              rj.gsub!(/\(inJIS2\)$/, '')
              #$stderr.puts cc.inspect
              next unless cc[0] =~ /^[弌-熙]$/
            end
            if k[1] == rj
              weak[cc[0]] = [cc.collect {|l| l.dup}, parts.collect {|l| l.dup}] if frec == :weak  #後からaliasされるのには未対応
              k[1] = r[0]
              order_weak |= [cc[0]] if frec == :oweak
              fcr = true if r[0] == cc[0]
            elsif k[1] == r[0]
              order_weak |= [cc[0]] if frec == :ostrong
            end
          }
        end
        if fcr
          #$stderr.print "deleted a crunched definition: #{cc.inspect} => #{parts.inspect}\n"
          rev.delete(cc)
        end
      }
      for i in 'abcdefijklmn?'.split(//)
        for j in 1...r.length
          next if prot.include?(r[j])
          next unless rev.has_key?([r[j], i])
          parts = rev.delete([r[j], i])
          fcr = false
          parts.each {|k|
            fcr = true if k[1] == r[0]
          }
          if fcr
            $stderr.print "deleted a crunched definition: #{[r[0], i].inspect} => #{parts.inspect}\n"
          else
            if rev.has_key?([r[0], i])
              $stderr.print "overwriting a collided definition by: #{[r[0], i].inspect} => #{parts.inspect}\n"
            end
            rev[[r[0], i]] = parts
          end
        end
      end
      prot = []
    end
    
  end
  fp.close

  delp = Regexp.new('^(' + delp.collect {|i| Regexp.escape(i) }.join('|') + ')(.)$')
  dels = Regexp.new('^(.)(' + dels.collect {|i| Regexp.escape(i) }.join('|') + ')$')
  reva = rev.dup
  rev.each {|cc, parts|
    parts = parts.collect {|i| i[1] = i[1].gsub(delp) { $2 }; i }
    parts = parts.collect {|i| i[1] = i[1].gsub(dels) { $1 }; i }
    if parts.collect {|i| i[1] }.include?(cc[0])
      $stderr.print "deleted a crunched definition: #{cc.inspect} => #{parts.inspect}\n"
      rev.delete(cc)
      next
    end
    rev[cc] = parts
    if cc[0].length > 2 || parts.collect {|i| i[1].length }.max > 2
      $stderr.puts [cc, parts].inspect unless cc[0] =~ /@\d\dJIS$/
      rev.delete(cc)
      next
    end
    #parts.each {|i|
    #  if i[1] =~ /\?/
    #    rev.delete(cc)
    #    break
    #  end
    #}
  }
  rev2 = {}
  rev.each {|cc, parts|
    if rev2.has_key?(cc[0])
      p = (rev2[cc[0]]=~/^[a-c]$/ ? rev2[cc[0]] : 'b'+rev2[cc[0]])
      #p cc[1]
      n = (cc[1]=~/^[a-c]$/ ? cc[1] : 'b'+cc[1])
      if PREFER_REGULAR
      else
        p = p.gsub(/^([ab])/) { $1 == 'a' ? 'b' : 'a' }
        n = n.gsub(/^([ab])/) { $1 == 'a' ? 'b' : 'a' }
      end
      if p > n
        rev2[cc[0]] = cc[1]
      end
    else
      rev2[cc[0]] = cc[1]
    end
  }
  rev.each {|cc, parts|
    next if rev2[cc[0]] == cc[1]
    rev.delete(cc)
  }
  fow = {}
  rev2 = {}
  rev.each {|cc, parts|
    parts0 = parts
    parts = parts.collect {|i| i[1] }  #delete pos data
    if fow.has_key?(parts)
#    $stderr.puts [cc, parts].inspect, (order_weak.include?(fow[parts][0]) ? 2 : 0)+(order_weak.include?(cc[0]) ? 1 : 0) if cc[0] =~ /[壌壤]/
      case (weak.has_key?(fow[parts][0]) ? 2 : 0)+(weak.has_key?(cc[0]) ? 1 : 0)
      when 0
        #$stderr.print "undefined priority: #{fow[parts][0]}, #{cc[0]}\n"
      when 3
        #$stderr.print "can't determine priority: #{fow[parts][0]}, #{cc[0]}\n"
      when 1
        $stderr.print "canceling a replace: #{cc[0]}, #{parts}, #{weak[cc[0]][1]}\n"
        parts0 = weak[cc[0]][1].dup
        parts = parts0.collect {|i| i[1] }
      when 2
        cc2 = fow[parts]
        fow[parts] = cc
        cc = cc2
        parts02 = rev2.delete(cc)
        rev2[fow[parts]] = parts0
        parts0 = parts02
        $stderr.print "canceling a replace: #{cc[0]}, #{parts}, #{weak[cc[0]][1]}\n"
        parts0 = weak[cc[0]][1].dup
        parts = parts0.collect {|i| i[1] }
      end
    end
    if fow.has_key?(parts)
#    $stderr.puts [cc, parts].inspect, (order_weak.include?(fow[parts][0]) ? 2 : 0)+(order_weak.include?(cc[0]) ? 1 : 0) if cc[0] =~ /[壌壤]/
      case (order_weak.include?(fow[parts][0]) ? 2 : 0)+(order_weak.include?(cc[0]) ? 1 : 0)
      when 0
        $stderr.print "undefined order-priority: #{fow[parts][0]}, #{cc[0]}\n"
      when 3
        $stderr.print "can't determine order-priority: #{fow[parts][0]}, #{cc[0]}\n"
      when 1
        parts0 = parts0.dup
        parts0[0..1] = [parts0[1], parts0[0]]
        parts = parts0.collect {|i| i[1] }
      when 2
        cc2 = fow[parts]
        fow[parts] = cc
        cc = cc2
        parts02 = rev2.delete(cc)
        rev2[fow[parts]] = parts0
        parts0 = parts02
        parts0 = parts0.dup
        parts0[0..1] = [parts0[1], parts0[0]]
        parts = parts0.collect {|i| i[1] }
      end
    end
    if fow.has_key?(parts)
      $stderr.print "too crowded: #{fow[parts][0]}, #{cc[0]}\n"
    end
    fow[parts] = cc
    rev2[cc] = parts0
  }
  rev2 = rev2.collect {|cc, parts|
    [cc[0]] + parts.collect {|i| i[1] } + [cc[1]] + parts.collect {|i| i[0] }
  }
  rev2.sort! {|i, j| i[0] <=> j[0] }
  rev2.each {|i|
    if opt =~ /^--to-rev/
      puts i[0...i.length/2].to_s
    elsif opt == '--to-thunder'
      next if i[0] == '裁'
      puts i[0, 3].to_s + i[i.length/2, 3].to_s
    end
  }

elsif opt == '--to-clwfk' || opt == '--to-chise'

  $renh0 = {}
  $renh1 = {}
  fp = open('cidbushu.alias.by_pos')
  while fp.gets
    chop!
    next if $_[0..0] == '#'
    r = $_.split(/\s+/)
    break if r.length < 3
    case r[0]
    when '.'
      $renh0[r[1]] = r[2]
    when /\$$/
      $renh1[[r[0][0..-2], r[1]]] = r[2]
    else
    end
  end
  fp.close

  while gets
    chop!
    next if $_ =~ /^\s*\#/
    gsub!(/\#.*$/, '')
    gsub!(/\s*$/, '')
    r = $_.split(/\s+/)
    l = r[0]
    l = l.split(/\,/)
    l.delete_if {|i| i =~ /^\d{4,}$/ }
    composed = l[0].gsub(/\@90JIS$/, '')
    composed = $renh0[composed] if $renh0.has_key?(composed)
    r.shift

    r.each_index {|i|
      parts = []
      if r[i][0..0] == 'L'
        next
      end
      l = r[i].split(/\,/)
      c = l[0]
      next if l.length <= 1
      if l[1] =~ /^[\=\/\<\>\|]$/
        next
      end
      l.shift
      l = l.collect {|j| j.gsub(/\((.*)\)$/) { $1 } }  #remove parenthesis
#
      def renparts(nd, pt = [])
        if nd.class == String
          pt = pt.join
          if $renh1.has_key?([pt[-1..-1], nd])
            nd.replace($renh1[[pt[-1..-1], nd]])
          elsif $renh0.has_key?(nd)
            nd.replace($renh0[nd])
          end
          return
        end
        nd0 = to_oi(nd[0])
        renparts(nd[1], pt + [nd0[0..0]])
        renparts(nd[2], pt + [nd0[1..1]])
      end
#
      l = treeize(l)
      m = renparts(l)
      
      c.gsub!(/\.\d\d/, '')
      if opt == '--to-clwfk'
        def to_paren(nd)
          return nd if nd.class == String
          oh = {'lr' => 'yoko', 'ud' => 'tate', 'oi' => 'kamae',
                'ku'=>'kamae', 'mo'=>'kamae', 'ka'=>'kamae', 'ha'=>'kamae',
                'ta'=>'tare', 'tu'=>'kamae', 'ny'=>'nyou'}
          return "(#{(oh[nd[0]] || nd[0])} #{to_paren(nd[1])} #{to_paren(nd[2])})"
        end
        print "(setq #{composed} '#{to_paren(l)})\n"
      end
    }
  end
  
elsif opt == '--to-chise' && nil #v1

  while gets
    chop!
    next if $_ =~ /^\s*\#/
    gsub!(/\#.*$/, '')
    gsub!(/\s*$/, '')
    r = $_.split(/\s+/)
    l = r[0]
    l = l.split(/\,/)
    l.delete_if {|i| i =~ /^\d{4,}$/ }
    l[1..-1].each {|i|
      ali[i] = l[0].gsub(/\@90JIS$/, '')
    }
    composed = l[0].gsub(/\@90JIS$/, '')
    r.shift
    
    i = 0
    parts = []
    fparen = false
    while i < r.length
      if r[i] =~ /^([a-z\?]|\Sc)$/
	if r[i] == 'l'
	  i += 2
	  next
	elsif r[i] == '!c' || r[i] =~ /^\Sc$/
	  i += 1
	  next
	end
	c = r[i]
      else
	if r[i] =~ /^[\=\/\<\>\|]/
	  i += 1
	  next
	end
        l = r[i]
        fparen = true if l =~ /\((.*)\)$/
	l = ' ' + l if l !~ /^[a-z\+\-]/
	l = [l[0..0], l[1..-1]]
	parts << l
	if (i == r.length-1 || r[i+1] =~ /^([a-z\?]|\Sc)$/) && parts.length > 0
          if c !~ /^[ab]$/ || fparen
            parts = []
            fparen = false
            i += 1
            next
          end
	  if rev.has_key?([composed, c])
	    $stderr.print "repeated composition definition:\n"
	    $stderr.print "#{[composed, c].inspect} => #{rev[[composed, c]].inspect}\n#{$_}\n"
	    i += 1
	    next
	  end
	  rev[[composed, c]] = parts
	  parts = []
          fparen = false
	end
      end
      i += 1
    end

  end

  def xx(c1,c2); "U+2FFB(#{c1},#{c2})"; end
  def lr(c1,c2); "U+2FF0(#{c1},#{c2})"; end
  def ud(c1,c2); "U+2FF1(#{c1},#{c2})"; end
  def oi(c1,c2,c0)
    if c0 =~ /^()$/
    elsif c0 =~ /^(广|厂|やまいだれ|尸|在の外|虍|广廿|厂林|厂イ|雁たれ)$/
      "U+2FF8(#{c1},#{c2})"
    elsif c0 =~ /^(囗四)$/
      "U+2FF4(#{c1},#{c2})"
    elsif c0 =~ /^(門|冂|岡ひく山|風|微かまえ|戊|戌)$/
      "U+2FF5(#{c1},#{c2})"
    elsif c0 =~ /^(勹|裁かまえ|武ひく止|)$/
      "U+2FF9(#{c1},#{c2})"
    elsif c0 =~ /^(匚|匸)$/
      "U+2FF7(#{c1},#{c2})"
    elsif c0 =~ /^(凵)$/
      "U+2FF6(#{c1},#{c2})"
    elsif c0 =~ /^(之|え|廴|走|夂)$/
      "U+2FFA(#{c1},#{c2})"
    elsif c0 =~ /^(行|衣)$/
      #U+2FF3
      #o->u+d, cf.'udi'
      "U+2FFB(#{c1},#{c2})"
    end
  end
  
  rev2 = []
  rev.each {|cc, parts|
    parts.sort! {|i, j| %w(l r u d o i p s m x + -).index(i[0]) <=> %w(l r u d o i p s m x + -).index(j[0]) }
    case parts.collect {|i| i[0]}.to_s
    when '+-'
      next
    when 'xx'
      rev2 << [cc, xx(parts[0][1],parts[1][1])]
    when 'mm'
      rev2 << [cc, xx(parts[0][1],parts[1][1])]
    when 'lr'
      rev2 << [cc, lr(parts[0][1],parts[1][1])]
    when 'ud'
      rev2 << [cc, ud(parts[0][1],parts[1][1])]
    when 'oi'
      rev2 << [cc, oi(parts[0][1],parts[1][1], cc[0])]
    when 'lu'
      rev2 << [cc, "U+2FF9(#{parts[1][1]},#{parts[0][1]})"]
      #u->l2+r, lr(ud(l2,l),r) or
      #u->u2+d, ud(u2,lr(l,d))
    when 'ld'
      #
    when 'ru'
      rev2 << [cc, "U+2FF8(#{parts[1][1]},#{parts[0][1]})"]
    when 'rd'
      rev2 << [cc, "U+2FFA(#{parts[1][1]},#{parts[0][1]})"]
    when 'lri'
      rev2 << [cc, "U+2FF2(#{parts[0][1]},#{parts[2][1]},#{parts[1][1]})"]
    when 'udi'
      rev2 << [cc, "U+2FF3(#{parts[0][1]},#{parts[2][1]},#{parts[1][1]})"]
    when 'lru'
      rev2 << [cc, ud(parts[2][1], lr(parts[0][1], parts[1][1]))]
    when 'lrd'
      rev2 << [cc, ud(lr(parts[0][1], parts[1][1]), parts[2][1])]
    when 'udo'
      rev2 << [cc, oi(parts[2][1], ud(parts[0][1], parts[1][1]))]
    when 'uo'
      #o->o2+i, oi(o2,ud(u,i))
      raise notImplementedError, "#{parts.collect {|i| i[0]}.to_s}"
    when 'do'
      raise notImplementedError, "#{parts.collect {|i| i[0]}.to_s}"
    else
      raise notImplementedError, "#{parts.collect {|i| i[0]}.to_s}"
    end
  }

  
elsif opt == '--for-tex'

#http://psitau.at.infoseek.co.jp/otf.html
#http://oku.edu.mie-u.ac.jp/~okumura/texwiki/?OTF

  puts '\documentclass{jarticle}'
  puts '\usepackage{utf}'
  puts '\begin{document}'
  while gets
    chop!
    gsub!(/\s\#$/, '')
    gsub!(/  /, '　')
    gsub!(/[\#$%&_{}^~<>\\|]/) {|i| i =~ /[<>\\]/ ? '\verb|'+i+'|' : i == '|' ? '\textbar ' : '\\'+i }
    gsub!(/^\d{4,}/) { '{\small '+$&+'}\CID{'+$&+'}' }
    gsub!(/U\+([0-9A-F]{4})/) { $&+'\UTF{'+$1+'}' }
    puts $_
    puts
  end
  puts '\end{document}'

elsif false

  itaiji = {}
  fp = open('./emacs/tcode/itaiji.maz')
  while fp.gets
    chop!
    i = split(/\s/)
    itaiji[i[0]] = i[1]
  end
  fp.close

  while gets
    chop!
    if $_ =~ /^\s*\#/
      puts $_
      next
    end
    r = $_.split(/\s+/)
    l = r[0]
    l = l.split(/\,/)
    l.delete_if {|i| i =~ /^\d{4,}$/ }
    c = l[0].gsub(/\@90JIS$/, '')
    if itaiji[c]
      $_ << " #c =#{itaiji[$1]} "
    end
    puts $_
  end

elsif false

#ftp://ftp.oreilly.com/pub/examples/nutshell/cjkv/adobe/

  while gets
    chop!
    next if $_ =~ /^\#/
    next if $_ =~ /^CID/
    codes = $_.split(/\t/)
    cid = codes[0].to_i
    c90jis = codes[1].hex
    c78jis = codes[13].hex
    next if c78jis == 0 && c90jis == 0
    w = '  '
    w[0] = (c90jis>>8)+0x80
    w[1] = (c90jis&0xff)+0x80
    c90jis = w
    w = '  '
    w[0] = (c78jis>>8)+0x80
    w[1] = (c78jis&0xff)+0x80
    c78jis = w
    if c90jis == c78jis
      printf "%-8d\#%s\n", cid, c90jis
    else
      l = []
      l << c90jis+"@90JIS" if c90jis.length > 0
      l << c78jis+"@78JIS" if c78jis.length > 0
      printf "%-8d#%s\n", cid, l.join(",")
    end
  end


else

  usage()


end    ####