i

[chise/ruby.git] / chise / parser.rb

diff --git a/chise/parser.rb b/chise/parser.rb
index 812b77c..f790468 100755 (executable)
--- a/chise/parser.rb
+++ b/chise/parser.rb
@@ -1,39 +1,16 @@
 # Copyright (C) 2002-2004 Kouichirou Eto, All rights reserved.
 
-module CHISE
-  class CharacterParser
-    def parse(c) # parse a value and return a number (MCS)
-      raise "c is nil" if c.nil?
-
-      if c.kind_of?(String)
-       if /\A\?/ =~ c
-         c = c.sub(/\A\?/, "") # remove "?" in the head
-         u4 = c.u8tou32 # translate from UTF-8 to UTF-32
-         return u4.u32to_i # translate UTF-32 to UCS number
-       end
-
-       return parse_er(c) if is_er?(c) # ER?
-
-       return c.to_i if /^\d+$/ =~ c # only numbers?
-
-       raise "unknown format"
-      end
-
-      if c.kind_of?(Numeric)
-       c = 0x80000000 + c if c < 0 # negative value
-       return c.to_i
-      end
-      
-      raise "unknown object"
-    end
+require "chise/chisedb"
 
+module CHISE
+  module EntityReference
     PART    = "&([-+0-9A-Za-z#]+);"
-    ALL     = "\\A#{PART}\\Z"
+    ALL     = '\A'+PART+'\Z'
     PART_RE = Regexp.new(PART)
     ALL_RE  = Regexp.new(ALL)
 
     def contain_er?(s) (PART_RE =~ s) != nil;  end
-    def is_er?(s)      (ALL_RE =~ s) != nil;   end
+    def is_er?(s)      (ALL_RE =~ s)  != nil;  end
 
     # the order is important.  The primary charset should be selectable.
     CODESYS_TABLE = [
@@ -61,6 +38,35 @@ module CHISE
       %w( =gt-k                        GT-K 5 d),
     ]
     PRIVATE_USE_AREA = 0xe000
+  end
+
+  class CharacterParser
+    include EntityReference
+
+    def parse(c) # parse a value and return a number (MCS)
+      raise "c is nil" if c.nil?
+
+      if c.kind_of?(String)
+       if /\A\?/ =~ c
+         c = c.sub(/\A\?/, "") # remove "?" in the head
+         u4 = c.u8tou32 # translate from UTF-8 to UTF-32
+         return u4.u32to_i # translate UTF-32 to UCS number
+       end
+
+       return parse_er(c) if is_er?(c) # ER?
+
+       return c.to_i if /^\d+$/ =~ c # only numbers?
+
+       raise "unknown format"
+      end
+
+      if c.kind_of?(Numeric)
+       c = 0x80000000 + c if c < 0 # negative value
+       return c.to_i
+      end
+      
+      raise "unknown object"
+    end
 
     def parse_er(s) # parse a Entity Reference and return a number (MCS)
       raise "wrong ER." unless ALL_RE =~ s # don't use is_er? for getting $1.
@@ -101,10 +107,10 @@ module CHISE
          code = codestr.hex
        end
 
-       u8 = CodesysDB.instance.get(codesys, code)
+       u8 = get_ccs(codesys, code)
        next if u8.nil?
 
-       num = parse("?"+u8)
+       num = parse(u8)
        next if num.nil?
 
        return num
@@ -113,5 +119,47 @@ module CHISE
       raise "unknown Entity Reference"
     end
 
+    private
+    def get_ccs(ccs, code_point)
+      cd = ChiseDB.instance
+      cd.decode_char(ccs, code_point)
+    end
+  end
+
+  class EntityReferenceParser
+    include EntityReference
+
+    def de_er(s) # replace EntityReference with corresponding character.
+      return s unless PART_RE =~ s # don't use contain_er? to get $1
+
+      er = "&"+$1+";"
+      char = Character.get(er)
+      ss = s.sub(Regexp.new(Regexp.escape(er)), char.utf8_mcs)
+
+      return de_er(ss) if contain_er?(ss) # recursive
+      ss
+    end
+  end
+
+  class EntityReferenceEncoder
+    include EntityReference
+
+    def to_er(char)
+      cid = char.char_id
+      return "&#x%04x;" % cid if cid <=  0xffff
+      return "&#x%05x;" % cid if cid <= 0xfffff
+
+      CODESYS_TABLE.each {|codesys, er_prefix, keta, numtype|
+       code = char[codesys]
+       next if code.nil?
+       return "&#{er_prefix}%0#{keta}#{numtype};" % code
+      }
+
+      "&MCS-%08X;" % cid # the last answer
+    end
+
+    def to_er_by_ccs(cid, codesys) # not yet
+    end
+
   end
 end