text_alignment 0.2.9 → 0.3.9

data/lib/text_alignment/find_divisions.rb

@@ -7,211 +7,209 @@ module TextAlignment; end unless defined? TextAlignment
 # to work on the hash representation of denotations
 # to assume that there is no bag representation to this method
 
-module TextAlignment
-  TextAlignment::SIMILARITY_THRESHOLD = 0.7
-end
+TextAlignment::SIMILARITY_THRESHOLD = 0.7 unless defined? TextAlignment::SIMILARITY_THRESHOLD
 
 class << TextAlignment
 
-  # It finds, among the sources, the right divisions for the taraget text to fit in.
-  def find_divisions(target, sources, mappings = [])
-    raise ArgumentError, "nil target"           if target == nil
-    raise ArgumentError, "nil or empty sources" if sources == nil || sources.empty?
-    raise ArgumentError, "nil mappings"         if mappings == nil
-
-    character_mappings = mappings.select{|m| m[0].length == 1 && m[1].length == 1}
-    mappings.delete_if{|m| m[0].length == 1 && m[1].length == 1}
-    characters_from = character_mappings.collect{|m| m[0]}.join
-    characters_to   = character_mappings.collect{|m| m[1]}.join
-    characters_to.gsub!(/-/, '\-')
-
-    target.tr!(characters_from, characters_to)
-    sources.each{|source| source[:text].tr!(characters_from, characters_to)}
-
-    # to process smaller ones first
-    sources.sort!{|s1, s2| s1[:text].size <=> s2[:text].size}
-
-    TextAlignment._find_divisions(target, sources)
-  end
-
-  def _find_divisions(_target, _sources)
-    indice = []
-    history = []
-    cache = {}
-    target = _target.dup
-    sources = _sources.dup
-    until target.strip.empty? || sources.empty?
-      mode, cmp = nil, nil
-      candidates = []
-      sources.each_with_index do |source, i|
-        if target.size < source[:text].size
-          mode = :t_in_s
-          str1 = target
-          str2 = source[:text]
-        else
-          mode = :s_in_t
-          str1 = source[:text]
-          str2 = target
-        end
-
-        len1 = str1.length
-        len2 = str2.length
-
-        offset_begin, offset_end = if (len2 - len1) > len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD)
-          approximate_fit(str1, str2)
-        else
-          # the whole target
-          [0, -1]
-        end
-
-        unless offset_begin.nil?
-          key = str1 + ' _:_ ' + str2[offset_begin .. offset_end]
-          cmp = if cache.has_key? key
-            cache[key]
-          else
-            cmp = TextAlignment::LCSComparison.new(str1, str2[offset_begin .. offset_end])
-          end
-          cache[key] = cmp
-
-          if (cmp.similarity > TextAlignment::SIMILARITY_THRESHOLD) && ((len1 - (cmp.str1_match_final - cmp.str1_match_initial + 1)) < len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD))
-            candidates << {idx:i, offset:offset_begin, mode:mode, cmp:cmp}
-          end
-        end
-      end
-
-      # return remaining target and sources if m.nil?
-      break if candidates.empty?
-
-      choice = candidates.max{|a, b| a[:cmp].similarity <=> a[:cmp].similarity}
-      m = choice[:idx]
-      mode = choice[:mode]
-
-      index = if mode == :t_in_s
-        {divid:sources[m][:divid], region:[0, target.size]}
-      else # :s_in_t
-        cmp = choice[:cmp]
-        offset = choice[:offset]
-        {divid:sources[m][:divid], region:[cmp.str2_match_initial + offset, cmp.str2_match_final + offset + 1]}
-      end
-
-      target = target[0 ... index[:region][0]] + target[index[:region][1] .. -1]
-      history << index[:region].dup
-
-      before_begin = index[:region][0]
-      before_end = index[:region][1]
-
-      rhistory = history.reverse
-      rhistory.shift
-      rhistory.each do |h|
-        gap = h[1] - h[0]
-        index[:region][0] += gap if index[:region][0] >= h[0]
-        index[:region][1] += gap if index[:region][1] >  h[0]
-      end
-
-      indice << index
-
-      sources.delete_at(m)
-    end
-
-    unless target.strip.empty? && sources.empty?
-      index = {divid:nil}
-      index[:remaining_target] = target unless target.strip.empty?
-      index[:remaining_sources] = sources.collect{|s| s[:divid]} unless sources.empty?
-      indice << index
-    end
-
-    indice
-  end
-
-  def _find_divisions_old(target, sources)
-    mode, m, c, offset_begin = nil, nil, nil, nil
-
-    sources.each_with_index do |source, i|
-      if target.size < source[:text].size
-        mode = :t_in_s
-        str1 = target
-        str2 = source[:text]
-      else
-        mode = :s_in_t
-        str1 = source[:text]
-        str2 = target
-      end
-
-      len1 = str1.length
-      len2 = str2.length
-
-      offset_begin, offset_end = 0, -1
-      offset_begin, offset_end = approximate_fit(str1, str2) if (len2 - len1) > len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD)
-
-      unless offset_begin.nil?
-        c = TextAlignment::LCSComparison.new(str1, str2[offset_begin .. offset_end])
-        if (c.similarity > TextAlignment::SIMILARITY_THRESHOLD) && ((len1 - (c.str1_match_final - c.str1_match_initial + 1)) < len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD))
-          m = i
-          break
-        end
-      end
-    end
-
-    # return remaining target and sources if m.nil?
-    return [[-1, [target, sources.collect{|s| s[:divid]}]]] if m.nil?
-
-    index = if mode == :t_in_s
-      [sources[m][:divid], [0, target.size]]
-    else # :s_in_t
-      [sources[m][:divid], [c.str2_match_initial + offset_begin, c.str2_match_final + offset_begin + 1]]
-    end
-
-    next_target = target[0 ... index[1][0]] + target[index[1][1] .. -1]
-    sources.delete_at(m)
-
-    if next_target.strip.empty? || sources.empty?
-      return [index]
-    else
-      more_index = _find_divisions(next_target, sources)
-      gap = index[1][1] - index[1][0]
-      more_index.each do |i|
-        if (i[0] > -1)
-          i[1][0] += gap if i[1][0] >= index[1][0]
-          i[1][1] += gap if i[1][1] >  index[1][0]
-        end
-      end
-      return [index] + more_index
-    end
-  end
+	# It finds, among the targets, the right divisions for the taraget text to fit in.
+	def find_divisions(source, targets, mappings = [])
+		raise ArgumentError, "nil source"           if source == nil
+		raise ArgumentError, "nil or empty targets" if targets == nil || targets.empty?
+		raise ArgumentError, "nil mappings"         if mappings == nil
+
+		character_mappings = mappings.select{|m| m[0].length == 1 && m[1].length == 1}
+		mappings.delete_if{|m| m[0].length == 1 && m[1].length == 1}
+		characters_from = character_mappings.collect{|m| m[0]}.join
+		characters_to   = character_mappings.collect{|m| m[1]}.join
+		characters_to.gsub!(/-/, '\-')
+
+		source.tr!(characters_from, characters_to)
+		targets.each{|target| target[:text].tr!(characters_from, characters_to)}
+
+		# to process smaller ones first
+		targets.sort!{|s1, s2| s1[:text].size <=> s2[:text].size}
+
+		TextAlignment._find_divisions(source, targets)
+	end
+
+	def _find_divisions(_source, _targets)
+		indice = []
+		history = []
+		cache = {}
+		source = _source.dup
+		targets = _targets.dup
+		until source.strip.empty? || targets.empty?
+			mode, cmp = nil, nil
+			candidates = []
+			targets.each_with_index do |target, i|
+				if source.size < target[:text].size
+					mode = :t_in_s
+					str1 = source
+					str2 = target[:text]
+				else
+					mode = :s_in_t
+					str1 = target[:text]
+					str2 = source
+				end
+
+				len1 = str1.length
+				len2 = str2.length
+
+				offset_begin, offset_end = if (len2 - len1) > len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD)
+					approximate_fit(str1, str2)
+				else
+					# the whole source
+					[0, -1]
+				end
+
+				unless offset_begin.nil?
+					key = str1 + ' _:_ ' + str2[offset_begin .. offset_end]
+					cmp = if cache.has_key? key
+						cache[key]
+					else
+						cmp = TextAlignment::LCSComparison.new(str1, str2[offset_begin .. offset_end])
+					end
+					cache[key] = cmp
+
+					if (cmp.similarity > TextAlignment::SIMILARITY_THRESHOLD) && ((len1 - (cmp.str1_match_final - cmp.str1_match_initial + 1)) < len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD))
+						candidates << {idx:i, offset:offset_begin, mode:mode, cmp:cmp}
+					end
+				end
+			end
+
+			# return remaining source and targets if m.nil?
+			break if candidates.empty?
+
+			choice = candidates.max{|a, b| a[:cmp].similarity <=> a[:cmp].similarity}
+			m = choice[:idx]
+			mode = choice[:mode]
+
+			index = if mode == :t_in_s
+				{divid:targets[m][:divid], region:[0, source.size]}
+			else # :s_in_t
+				cmp = choice[:cmp]
+				offset = choice[:offset]
+				{divid:targets[m][:divid], region:[cmp.str2_match_initial + offset, cmp.str2_match_final + offset + 1]}
+			end
+
+			source = source[0 ... index[:region][0]] + source[index[:region][1] .. -1]
+			history << index[:region].dup
+
+			before_begin = index[:region][0]
+			before_end = index[:region][1]
+
+			rhistory = history.reverse
+			rhistory.shift
+			rhistory.each do |h|
+				gap = h[1] - h[0]
+				index[:region][0] += gap if index[:region][0] >= h[0]
+				index[:region][1] += gap if index[:region][1] >  h[0]
+			end
+
+			indice << index
+
+			targets.delete_at(m)
+		end
+
+		unless source.strip.empty? && targets.empty?
+			index = {divid:nil}
+			index[:remaining_source] = source unless source.strip.empty?
+			index[:remaining_targets] = targets.collect{|s| s[:divid]} unless targets.empty?
+			indice << index
+		end
+
+		indice
+	end
+
+	def _find_divisions_old(source, targets)
+		mode, m, c, offset_begin = nil, nil, nil, nil
+
+		targets.each_with_index do |target, i|
+			if source.size < target[:text].size
+				mode = :t_in_s
+				str1 = source
+				str2 = target[:text]
+			else
+				mode = :s_in_t
+				str1 = target[:text]
+				str2 = source
+			end
+
+			len1 = str1.length
+			len2 = str2.length
+
+			offset_begin, offset_end = 0, -1
+			offset_begin, offset_end = approximate_fit(str1, str2) if (len2 - len1) > len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD)
+
+			unless offset_begin.nil?
+				c = TextAlignment::LCSComparison.new(str1, str2[offset_begin .. offset_end])
+				if (c.similarity > TextAlignment::SIMILARITY_THRESHOLD) && ((len1 - (c.str1_match_final - c.str1_match_initial + 1)) < len1 * (1 - TextAlignment::SIMILARITY_THRESHOLD))
+					m = i
+					break
+				end
+			end
+		end
+
+		# return remaining source and targets if m.nil?
+		return [[-1, [source, targets.collect{|s| s[:divid]}]]] if m.nil?
+
+		index = if mode == :t_in_s
+			[targets[m][:divid], [0, source.size]]
+		else # :s_in_t
+			[targets[m][:divid], [c.str2_match_initial + offset_begin, c.str2_match_final + offset_begin + 1]]
+		end
+
+		next_source = source[0 ... index[1][0]] + source[index[1][1] .. -1]
+		targets.delete_at(m)
+
+		if next_source.strip.empty? || targets.empty?
+			return [index]
+		else
+			more_index = _find_divisions(next_source, targets)
+			gap = index[1][1] - index[1][0]
+			more_index.each do |i|
+				if (i[0] > -1)
+					i[1][0] += gap if i[1][0] >= index[1][0]
+					i[1][1] += gap if i[1][1] >  index[1][0]
+				end
+			end
+			return [index] + more_index
+		end
+	end
 
 end
 
 if __FILE__ == $0
-  require 'json'
-  if ARGV.length == 2
-    target  = JSON.parse File.read(ARGV[0]), :symbolize_names => true
-    target_text = target[:text].strip
-
-    sources = JSON.parse File.read(ARGV[1]), :symbolize_names => true
-    div_index = TextAlignment::find_divisions(target_text, sources)
-    pp div_index
-
-    # str1 = File.read(ARGV[0]).strip
-    # str2 = File.read(ARGV[1]).strip
-    # div_index = TextAlignment::find_divisions(str1, [str2])
-
-    # puts "target length: #{target_text.length}"
-    # div_index.each do |i|
-    #   unless i[:divid].nil?
-    #     puts "[Div: #{i[:divid]}] (#{i[:region][0]}, #{i[:region][1]})"
-    #     puts target_text[i[:region][0] ... i[:region][1]]
-    #     puts "=========="
-    #   else
-    #     p i
-    #   end
-
-    #   # if i[0] >= 0
-    #   #   puts "[Div: #{i[0]}] (#{i[1][0]}, #{i[1][1]})"
-    #   #   puts target_text[i[1][0] ... i[1][1]]
-    #   #   puts "=========="
-    #   # else
-    #   #   p i
-    #   # end
-    # end
-  end
+	require 'json'
+	if ARGV.length == 2
+		source  = JSON.parse File.read(ARGV[0]), :symbolize_names => true
+		source_text = source[:text].strip
+
+		targets = JSON.parse File.read(ARGV[1]), :symbolize_names => true
+		div_index = TextAlignment::find_divisions(source_text, targets)
+		pp div_index
+
+		# str1 = File.read(ARGV[0]).strip
+		# str2 = File.read(ARGV[1]).strip
+		# div_index = TextAlignment::find_divisions(str1, [str2])
+
+		# puts "source length: #{source_text.length}"
+		# div_index.each do |i|
+		#   unless i[:divid].nil?
+		#     puts "[Div: #{i[:divid]}] (#{i[:region][0]}, #{i[:region][1]})"
+		#     puts source_text[i[:region][0] ... i[:region][1]]
+		#     puts "=========="
+		#   else
+		#     p i
+		#   end
+
+		#   # if i[0] >= 0
+		#   #   puts "[Div: #{i[0]}] (#{i[1][0]}, #{i[1][1]})"
+		#   #   puts source_text[i[1][0] ... i[1][1]]
+		#   #   puts "=========="
+		#   # else
+		#   #   p i
+		#   # end
+		# end
+	end
 end

data/lib/text_alignment/glcs_alignment.rb

@@ -6,306 +6,306 @@ module TextAlignment; end unless defined? TextAlignment
 # An instance of this class holds the results of generalized LCS computation for the two strings str1 and str2.
 # an optional dictionary is used for generalized suffix comparision.
 class TextAlignment::GLCSAlignment
-  # The mapping function from str1 to str2
-  attr_reader :position_map_begin, :position_map_end
-
-  # The position initial and final position of matching on str1 and str2
-  attr_reader :str1_match_begin, :str1_match_end, :str2_match_begin, :str2_match_end
-
-  # The length of GLCS
-  attr_reader :length
-
-  # the elements that are common in the two strings, str1 and str2
-  attr_reader :common_elements
-
-  # the elements that are mapped to each other in the two strings, str1 and str2
-  attr_reader :mapped_elements
-
-  # the string of non-mapped characters
-  attr_reader :diff_strings
-
-  attr_reader :similarity
-
-  # It initializes the GLCS table for the given two strings, str1 and str2.
-  # When the array, mappings, is given, general suffix comparision is performed based on the mappings.
-  # Exception is raised when nil given passed to either str1, str2 or dictionary
-  def initialize(str1, str2, mappings = [])
-    raise ArgumentError, "nil string"     if str1 == nil || str2 == nil
-    raise ArgumentError, "nil dictionary" if mappings == nil
-
-    # index the mappings in hash.
-    @dic = (mappings + mappings.map{|e| e.reverse}).to_h
-
-    # prefix dictionary
-    @pdic = Dictionary.new(mappings.flatten)
-
-    @len1 = str1.length
-    @len2 = str2.length
-
-    # add a final marker to the end of the strings
-    @str1 = str1 + '_'
-    @str2 = str2 + '_'
-
-    # compute the GLCS table
-    @glcs = _compute_glcs_table
-    @length = @glcs[0][0]
-
-    _trace_glcs_table
-  end
-
-  # Prints the GLCS table
-  def show_glcs
-    puts "\t\t" + @str2.split(//).join("\t")
-    @glcs.each_with_index do |row, i|
-      h = (@str1[i].nil?)? '' : @str1[i]
-      puts i.to_s + "\t" + h + "\t" + row.join("\t")
-    end
-  end
-
-  # Returns the character-by-character difference
-  def cdiff
-    cdiff1, cdiff2 = '', ''
-    p1, p2 = 0, 0
-    begin
-      s1, s2 = _prefix_eq(@str1[p1...@len1], @str2[p2...@len2])
-      if s1 != nil
-        l1, l2 = s1.length, s2.length
-
-        cdiff1 += s1; cdiff2 += s2
-        if l1 > l2 then cdiff2 += ' ' * (l1 - l2) else cdiff1 += ' ' * (l2 - l1) end
-        p1 += s1.length;  p2 += s2.length
-      elsif p2 < @len2 && (p1 == @len1 or @glcs[p1][p2 + 1] > @glcs[p1 + 1][p2])
-        cdiff1 += ' '
-        cdiff2 += @str2[p2]
-        p2 += 1
-      elsif p1 < @len1 && (p2 == @len2 or @glcs[p1][p2 + 1] <= @glcs[p1 + 1][p2])
-        cdiff1 += @str1[p1]
-        cdiff2 += ' '
-        p1 += 1
-      end
-    end until p1 == @len1 && p2 == @len2
-
-    return [cdiff1, cdiff2]
-  end
-
-
-  # Computes the similarity of the two strings
-  def similarity(cut = false)
-    c = @length
-
-    l1 = c + @diff_strings[0].length
-    l2 = c + @diff_strings[1].length
-
-    if cut
-      l1 -= front_overflow if front_overflow > 0
-      l1 -= rear_overflow  if rear_overflow  > 0
-      l1 += front_overflow if front_overflow < 0
-      l1 += rear_overflow  if rear_overflow  < 0
-    end
-
-    similarity = 2 * c / (l1 + l2).to_f
-  end
-
-  def transform_a_span(span)
-    {:begin=>@position_map_begin[span[:begin]], :end=>@position_map_end[span[:end]]}
-  end
-
-  def transform_spans(spans)
-    spans.map{|span| transform_a_span(span)}
-  end
-
-
-  private
-
-  # Computes the GLCS table for the two strings, @str1 and @str2.
-  # Unlike normal LCS algorithms, the computation is performed from the end to the beginning of the strings.
-  def _compute_glcs_table
-    glcs = Array.new(@len1 + 1) { Array.new(@len2 + 1) }
-
-    # initialize the final row and the final column
-    (0..@len1).each {|p| glcs[p][@len2] = 0}
-    (0..@len2).each {|p| glcs[@len1][p] = 0}
-
-    # compute the GLCS table
-    str1_reverse_iteration = (0...@len1).to_a.reverse
-    str2_reverse_iteration = (0...@len2).to_a.reverse
-
-    str1_reverse_iteration.each do |p1|
-      str2_reverse_iteration.each do |p2|
-        s1, s2 = _prefix_eq(@str1[p1...@len1], @str2[p2...@len2])
-        unless s1 == nil
-          glcs[p1][p2] = glcs[p1 + s1.length][p2 + s2.length] + 1
-        else
-          glcs[p1][p2] = (glcs[p1][p2 + 1] > glcs[p1 + 1][p2])? glcs[p1][p2 + 1] : glcs[p1 + 1][p2]
-        end
-      end
-    end
-
-    glcs
-  end
-
-  # Backtrace the GLCS table, computing the mapping function from str1 to str2
-  # As its side effect, it updates four global variables
-  # * front_overflow: the length of the front part of str1 that cannot fit in str2.
-  # * rear_overflow: the length of the rear part of str1 that cannot fit in str2.
-  # * common_elements: an array which stores the common elements in the two strings.
-  # * mapped_elements: an array which stores the mapped elements in the two strings.
-  def _trace_glcs_table
-    @front_overflow, @rear_overflow  = 0, 0
-    @common_elements, @mapped_elements = [], []
-    diff_string1, diff_string2 = '', ''
-
-    @position_map_begin, @position_map_end = {}, {}
-    addition, deletion = [], []
-    p1, p2 = 0, 0
-
-    while p1 <= @len1 && p2 <= @len2
-      s1, s2 = _prefix_eq(@str1[p1..@len1], @str2[p2..@len2])
-      if s1 != nil
-        l1, l2 = s1.length, s2.length
-
-        @position_map_begin[p1], @position_map_end[p1] = p2, p2
-        (p1 + 1 ... p1 + l1).each{|i| @position_map_begin[i], @position_map_end[i] = nil, nil}
-
-        @common_elements << [s1, s2]
-
-        if !addition.empty? && deletion.empty?
-          # If an addition is found in the front or the rear, it is a case of underflow
-          @str2_match_begin = addition.length if p1 == 0
-          @str2_match_end = l2 - addition.length if p1 == @len1
-
-          if p1 == 0
-            # leave as it is
-          elsif p1 == @len1
-            # retract from the end
-            @position_map_begin[p1] = p2 - addition.length
-            @position_map_end[p1] = @position_map_begin[p1]
-          else
-            # correct the position for end
-            @position_map_end[p1] = p2 - addition.length
-          end
-        elsif addition.empty? && !deletion.empty?
-          # If a deletion is found in the front or the rear, it is a case of overflow
-          @str1_match_begin = deletion.length if p1 == deletion.length
-          @str1_match_end = l1 - deletion.length if p1 == @len1
-
-          deletion.each{|p| @position_map_begin[p], @position_map_end[p] = p2, p2}
-        elsif !addition.empty? && !deletion.empty?
-          # If an addition and a deletion are both found in the front or the rear,
-          # the overflow/underflow is approximated to the difference.
-          al, dl = addition.length, deletion.length
-          @front_overflow = dl - al if p1 == dl
-          @rear_overflow  = dl - al if p1 == @len1
-
-          @mapped_elements << [@str1[deletion[0], dl], @str2[addition[0], al]]
-
-          @position_map_begin[deletion[0]], @position_map_end[deletion[0]] = addition[0], addition[0]
-          deletion[1..-1].each{|p| @position_map_begin[p], @position_map_end[p] = nil, nil}
-        end
-
-        addition.clear; deletion.clear
-        p1 += l1; p2 += l2
-
-      elsif p2 < @len2 && (p1 == @len1 || @glcs[p1][p2 + 1] >  @glcs[p1 + 1][p2])
-        diff_string2 += @str2[p2]
-
-        addition << p2
-        p2 += 1
-      elsif p1 < @len1 && (p2 == @len2 || @glcs[p1][p2 + 1] <= @glcs[p1 + 1][p2])
-        diff_string1 += @str1[p1]
-
-        deletion << p1
-        p1 += 1
-      end
-    end
-
-    @common_elements.pop
-    @diff_strings = [diff_string1, diff_string2]
-  end
-
-  # General prefix comparison is performed based on the dictionary.
-  # The pair of matched suffixes are returned when found.
-  # Otherwise, the pair of nil values are returned.
-  def _prefix_eq(str1, str2)
-    return nil, nil if str1.empty? || str2.empty?
-    prefixes1 = @pdic.prefixes(str1)
-    prefixes1.each {|p1| p2 = @dic[p1]; return p1, p2 if str2.start_with?(p2)}
-    return str1[0], str2[0] if (str1[0] == str2[0])
-    return nil, nil
-  end
+	# The mapping function from str1 to str2
+	attr_reader :position_map_begin, :position_map_end
+
+	# The position initial and final position of matching on str1 and str2
+	attr_reader :str1_match_begin, :str1_match_end, :str2_match_begin, :str2_match_end
+
+	# The length of GLCS
+	attr_reader :length
+
+	# the elements that are common in the two strings, str1 and str2
+	attr_reader :common_elements
+
+	# the elements that are mapped to each other in the two strings, str1 and str2
+	attr_reader :mapped_elements
+
+	# the string of non-mapped characters
+	attr_reader :diff_strings
+
+	attr_reader :similarity
+
+	# It initializes the GLCS table for the given two strings, str1 and str2.
+	# When the array, mappings, is given, general suffix comparision is performed based on the mappings.
+	# Exception is raised when nil given passed to either str1, str2 or dictionary
+	def initialize(str1, str2, mappings = [])
+		raise ArgumentError, "nil string"     if str1 == nil || str2 == nil
+		raise ArgumentError, "nil dictionary" if mappings == nil
+
+		# index the mappings in hash.
+		@dic = (mappings + mappings.map{|e| e.reverse}).to_h
+
+		# prefix dictionary
+		@pdic = Dictionary.new(mappings.flatten)
+
+		@len1 = str1.length
+		@len2 = str2.length
+
+		# add a final marker to the end of the strings
+		@str1 = str1 + '_'
+		@str2 = str2 + '_'
+
+		# compute the GLCS table
+		@glcs = _compute_glcs_table
+		@length = @glcs[0][0]
+
+		_trace_glcs_table
+	end
+
+	# Prints the GLCS table
+	def show_glcs
+		puts "\t\t" + @str2.split(//).join("\t")
+		@glcs.each_with_index do |row, i|
+			h = (@str1[i].nil?)? '' : @str1[i]
+			puts i.to_s + "\t" + h + "\t" + row.join("\t")
+		end
+	end
+
+	# Returns the character-by-character difference
+	def cdiff
+		cdiff1, cdiff2 = '', ''
+		p1, p2 = 0, 0
+		begin
+			s1, s2 = _prefix_eq(@str1[p1...@len1], @str2[p2...@len2])
+			if s1 != nil
+				l1, l2 = s1.length, s2.length
+
+				cdiff1 += s1; cdiff2 += s2
+				if l1 > l2 then cdiff2 += ' ' * (l1 - l2) else cdiff1 += ' ' * (l2 - l1) end
+				p1 += s1.length;  p2 += s2.length
+			elsif p2 < @len2 && (p1 == @len1 or @glcs[p1][p2 + 1] > @glcs[p1 + 1][p2])
+				cdiff1 += ' '
+				cdiff2 += @str2[p2]
+				p2 += 1
+			elsif p1 < @len1 && (p2 == @len2 or @glcs[p1][p2 + 1] <= @glcs[p1 + 1][p2])
+				cdiff1 += @str1[p1]
+				cdiff2 += ' '
+				p1 += 1
+			end
+		end until p1 == @len1 && p2 == @len2
+
+		return [cdiff1, cdiff2]
+	end
+
+
+	# Computes the similarity of the two strings
+	def similarity(cut = false)
+		c = @length
+
+		l1 = c + @diff_strings[0].length
+		l2 = c + @diff_strings[1].length
+
+		if cut
+			l1 -= front_overflow if front_overflow > 0
+			l1 -= rear_overflow  if rear_overflow  > 0
+			l1 += front_overflow if front_overflow < 0
+			l1 += rear_overflow  if rear_overflow  < 0
+		end
+
+		similarity = 2 * c / (l1 + l2).to_f
+	end
+
+	def transform_a_span(span)
+		{:begin=>@position_map_begin[span[:begin]], :end=>@position_map_end[span[:end]]}
+	end
+
+	def transform_spans(spans)
+		spans.map{|span| transform_a_span(span)}
+	end
+
+
+	private
+
+	# Computes the GLCS table for the two strings, @str1 and @str2.
+	# Unlike normal LCS algorithms, the computation is performed from the end to the beginning of the strings.
+	def _compute_glcs_table
+		glcs = Array.new(@len1 + 1) { Array.new(@len2 + 1) }
+
+		# initialize the final row and the final column
+		(0..@len1).each {|p| glcs[p][@len2] = 0}
+		(0..@len2).each {|p| glcs[@len1][p] = 0}
+
+		# compute the GLCS table
+		str1_reverse_iteration = (0...@len1).to_a.reverse
+		str2_reverse_iteration = (0...@len2).to_a.reverse
+
+		str1_reverse_iteration.each do |p1|
+			str2_reverse_iteration.each do |p2|
+				s1, s2 = _prefix_eq(@str1[p1...@len1], @str2[p2...@len2])
+				unless s1 == nil
+					glcs[p1][p2] = glcs[p1 + s1.length][p2 + s2.length] + 1
+				else
+					glcs[p1][p2] = (glcs[p1][p2 + 1] > glcs[p1 + 1][p2])? glcs[p1][p2 + 1] : glcs[p1 + 1][p2]
+				end
+			end
+		end
+
+		glcs
+	end
+
+	# Backtrace the GLCS table, computing the mapping function from str1 to str2
+	# As its side effect, it updates four global variables
+	# * front_overflow: the length of the front part of str1 that cannot fit in str2.
+	# * rear_overflow: the length of the rear part of str1 that cannot fit in str2.
+	# * common_elements: an array which stores the common elements in the two strings.
+	# * mapped_elements: an array which stores the mapped elements in the two strings.
+	def _trace_glcs_table
+		@front_overflow, @rear_overflow  = 0, 0
+		@common_elements, @mapped_elements = [], []
+		diff_string1, diff_string2 = '', ''
+
+		@position_map_begin, @position_map_end = {}, {}
+		addition, deletion = [], []
+		p1, p2 = 0, 0
+
+		while p1 <= @len1 && p2 <= @len2
+			s1, s2 = _prefix_eq(@str1[p1..@len1], @str2[p2..@len2])
+			if s1 != nil
+				l1, l2 = s1.length, s2.length
+
+				@position_map_begin[p1], @position_map_end[p1] = p2, p2
+				(p1 + 1 ... p1 + l1).each{|i| @position_map_begin[i], @position_map_end[i] = nil, nil}
+
+				@common_elements << [s1, s2]
+
+				if !addition.empty? && deletion.empty?
+					# If an addition is found in the front or the rear, it is a case of underflow
+					@str2_match_begin = addition.length if p1 == 0
+					@str2_match_end = l2 - addition.length if p1 == @len1
+
+					if p1 == 0
+						# leave as it is
+					elsif p1 == @len1
+						# retract from the end
+						@position_map_begin[p1] = p2 - addition.length
+						@position_map_end[p1] = @position_map_begin[p1]
+					else
+						# correct the position for end
+						@position_map_end[p1] = p2 - addition.length
+					end
+				elsif addition.empty? && !deletion.empty?
+					# If a deletion is found in the front or the rear, it is a case of overflow
+					@str1_match_begin = deletion.length if p1 == deletion.length
+					@str1_match_end = l1 - deletion.length if p1 == @len1
+
+					deletion.each{|p| @position_map_begin[p], @position_map_end[p] = p2, p2}
+				elsif !addition.empty? && !deletion.empty?
+					# If an addition and a deletion are both found in the front or the rear,
+					# the overflow/underflow is approximated to the difference.
+					al, dl = addition.length, deletion.length
+					@front_overflow = dl - al if p1 == dl
+					@rear_overflow  = dl - al if p1 == @len1
+
+					@mapped_elements << [@str1[deletion[0], dl], @str2[addition[0], al]]
+
+					@position_map_begin[deletion[0]], @position_map_end[deletion[0]] = addition[0], addition[0]
+					deletion[1..-1].each{|p| @position_map_begin[p], @position_map_end[p] = nil, nil}
+				end
+
+				addition.clear; deletion.clear
+				p1 += l1; p2 += l2
+
+			elsif p2 < @len2 && (p1 == @len1 || @glcs[p1][p2 + 1] >  @glcs[p1 + 1][p2])
+				diff_string2 += @str2[p2]
+
+				addition << p2
+				p2 += 1
+			elsif p1 < @len1 && (p2 == @len2 || @glcs[p1][p2 + 1] <= @glcs[p1 + 1][p2])
+				diff_string1 += @str1[p1]
+
+				deletion << p1
+				p1 += 1
+			end
+		end
+
+		@common_elements.pop
+		@diff_strings = [diff_string1, diff_string2]
+	end
+
+	# General prefix comparison is performed based on the dictionary.
+	# The pair of matched suffixes are returned when found.
+	# Otherwise, the pair of nil values are returned.
+	def _prefix_eq(str1, str2)
+		return nil, nil if str1.empty? || str2.empty?
+		prefixes1 = @pdic.prefixes(str1)
+		prefixes1.each {|p1| p2 = @dic[p1]; return p1, p2 if str2.start_with?(p2)}
+		return str1[0], str2[0] if (str1[0] == str2[0])
+		return nil, nil
+	end
 
 end
 
 if __FILE__ == $0
 
-  dictionary = [
-                ["×", "x"],       #U+00D7 (multiplication sign)
-                ["•", "*"],       #U+2022 (bullet)
-                ["Δ", "delta"],   #U+0394 (greek capital letter delta)
-                ["Φ", "phi"],     #U+03A6 (greek capital letter phi)
-                ["α", "alpha"],   #U+03B1 (greek small letter alpha)
-                ["β", "beta"],    #U+03B2 (greek small letter beta)
-                ["γ", "gamma"],   #U+03B3 (greek small letter gamma)
-                ["δ", "delta"],   #U+03B4 (greek small letter delta)
-                ["ε", "epsilon"], #U+03B5 (greek small letter epsilon)
-                ["κ", "kappa"],   #U+03BA (greek small letter kappa)
-                ["λ", "lambda"],  #U+03BB (greek small letter lambda)
-                ["μ", "mu"],      #U+03BC (greek small letter mu)
-                ["χ", "chi"],     #U+03C7 (greek small letter chi)
-                ["ϕ", "phi"],     #U+03D5 (greek phi symbol)
-                [" ", " "],       #U+2009 (thin space)
-                [" ", " "],       #U+200A (hair space)
-                [" ", " "],       #U+00A0 (no-break space)
-                ["　", " "],       #U+3000 (ideographic space)
-                ["−", "-"],       #U+2212 (minus sign)
-                ["–", "-"],       #U+2013 (en dash)
-                ["′", "'"],       #U+2032 (prime)
-                ["‘", "'"],       #U+2018 (left single quotation mark)
-                ["’", "'"],       #U+2019 (right single quotation mark)
-                ["“", '"'],       #U+201C (left double quotation mark)
-                ["”", '"']        #U+201D (right double quotation mark)
-               ]
-
-  # str1 = "-betakappaxyz-"
-  # str2 = "-ijkβκ-"
-
-  # str1 = "-βκ-β-z-xy"
-  # str2 = "abc-betakappa-beta-z"
-
-  # str1 = "-βκ-z-xy"
-  # str2 = "abc-betakappa-z"
-
-  # str1 = "abc-βκ-β-z"
-  # str2 = "-betakappa-beta-z-xyz"
-
-  # str1 = "-β-"
-  # str2 = "-beta-"
-
-  # str1 = "-κ-"
-  # str2 = "-kappa-"
-
-  # str1 = File.read(ARGV[0]).strip
-  # str2 = File.read(ARGV[1]).strip
-
-  str1 = "beta"
-  str2 = "β***"
-  
-  # puts "str1: #{str1}"
-  # puts "str2: #{str2}"
-  sa = TextAlignment::GLCSAlignment.new(str1, str2, dictionary)
-  sa.position_map_begin.each {|h| p h}
-  puts '-----'
-  sa.position_map_end.each {|h| p h}
-  puts '-----'
-  puts "common_elements: #{sa.common_elements}"
-  puts '-----'
-  puts "mapped_elements: #{sa.mapped_elements}"
-  puts '-----'
-  # puts "diff_string1: #{sa.diff_strings[0]}"
-  # puts "diff_string2: #{sa.diff_strings[1]}"
-  puts "front_overflow: #{sa.front_overflow}"
-  puts "rear_overflow : #{sa.rear_overflow}"
-  puts '-----'
-  puts "similarity     : #{sa.similarity}"
-  puts "similarity(cut): #{sa.similarity(true)}"
+	dictionary = [
+								["×", "x"],       #U+00D7 (multiplication sign)
+								["•", "*"],       #U+2022 (bullet)
+								["Δ", "delta"],   #U+0394 (greek capital letter delta)
+								["Φ", "phi"],     #U+03A6 (greek capital letter phi)
+								["α", "alpha"],   #U+03B1 (greek small letter alpha)
+								["β", "beta"],    #U+03B2 (greek small letter beta)
+								["γ", "gamma"],   #U+03B3 (greek small letter gamma)
+								["δ", "delta"],   #U+03B4 (greek small letter delta)
+								["ε", "epsilon"], #U+03B5 (greek small letter epsilon)
+								["κ", "kappa"],   #U+03BA (greek small letter kappa)
+								["λ", "lambda"],  #U+03BB (greek small letter lambda)
+								["μ", "mu"],      #U+03BC (greek small letter mu)
+								["χ", "chi"],     #U+03C7 (greek small letter chi)
+								["ϕ", "phi"],     #U+03D5 (greek phi symbol)
+								[" ", " "],       #U+2009 (thin space)
+								[" ", " "],       #U+200A (hair space)
+								[" ", " "],       #U+00A0 (no-break space)
+								["　", " "],       #U+3000 (ideographic space)
+								["−", "-"],       #U+2212 (minus sign)
+								["–", "-"],       #U+2013 (en dash)
+								["′", "'"],       #U+2032 (prime)
+								["‘", "'"],       #U+2018 (left single quotation mark)
+								["’", "'"],       #U+2019 (right single quotation mark)
+								["“", '"'],       #U+201C (left double quotation mark)
+								["”", '"']        #U+201D (right double quotation mark)
+							 ]
+
+	# str1 = "-betakappaxyz-"
+	# str2 = "-ijkβκ-"
+
+	# str1 = "-βκ-β-z-xy"
+	# str2 = "abc-betakappa-beta-z"
+
+	# str1 = "-βκ-z-xy"
+	# str2 = "abc-betakappa-z"
+
+	# str1 = "abc-βκ-β-z"
+	# str2 = "-betakappa-beta-z-xyz"
+
+	# str1 = "-β-"
+	# str2 = "-beta-"
+
+	# str1 = "-κ-"
+	# str2 = "-kappa-"
+
+	# str1 = File.read(ARGV[0]).strip
+	# str2 = File.read(ARGV[1]).strip
+
+	str1 = "beta"
+	str2 = "β***"
+
+	# puts "str1: #{str1}"
+	# puts "str2: #{str2}"
+	sa = TextAlignment::GLCSAlignment.new(str1, str2, dictionary)
+	sa.position_map_begin.each {|h| p h}
+	puts '-----'
+	sa.position_map_end.each {|h| p h}
+	puts '-----'
+	puts "common_elements: #{sa.common_elements}"
+	puts '-----'
+	puts "mapped_elements: #{sa.mapped_elements}"
+	puts '-----'
+	# puts "diff_string1: #{sa.diff_strings[0]}"
+	# puts "diff_string2: #{sa.diff_strings[1]}"
+	puts "front_overflow: #{sa.front_overflow}"
+	puts "rear_overflow : #{sa.rear_overflow}"
+	puts '-----'
+	puts "similarity     : #{sa.similarity}"
+	puts "similarity(cut): #{sa.similarity(true)}"
 end