text_alignment 0.4.2 → 0.4.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d56b44ee3cc3451b78ddb87134fefd6c043b4730f1170b07069a417bb162bad1
-  data.tar.gz: 67318d70bee5bb82127c98e79e0b45e5e337832a2b8d23469474215168579f20
+  metadata.gz: 52bc948955e2df858e397b14eabb4411f73b3ff1e4d879ff4b7015d3b5e03308
+  data.tar.gz: fd20caec51c95bdc475e0698a52bb7fdebc9e22c43bb47267a883bcc75862268
 SHA512:
-  metadata.gz: 60ae2292556d15d860e0c7377be9c1fd5aec6741b26f562fee8c0c2a9a123d0b8c656b2f7f34b8f74f83ea9ac4878b53f17454e3b25d27d00cd98ecf8175a387
-  data.tar.gz: fe8fe545101cddb638c00ffb6f6785f8b2f5ddb109b4010739e399a2c50fe41164d791fe4a5599a76fca5553c494f78fd49630368d48e406a8211fc0516916f5
+  metadata.gz: dbcb7ab70a64d4a398a5c5761cc5b2f5de6835ccc0e2d0854556f03ef91d0c0294986cc2ff1273788e6b7b0c73dfdf86fd16ee1ef8ce35ecc11d61f8eaab9521
+  data.tar.gz: 01d21cdcc0ab81d61e08ff1f52360ba35973756fd5060ce866391ff622d4cf87da945dba43a622a0581d63ee96c8723e1cb28991bfa02f4e1e803896bdc64d7f

data/lib/text_alignment/constants.rb

@@ -4,4 +4,4 @@ TextAlignment::SIZE_NGRAM = 8 unless defined? TextAlignment::SIZE_NGRAM
 TextAlignment::SIZE_WINDOW = 60 unless defined? TextAlignment::SIZE_WINDOW
 TextAlignment::BUFFER_RATE = 0.1 unless defined? TextAlignment::BUFFER_RATE
 TextAlignment::BUFFER_MIN = 20 unless defined? TextAlignment::BUFFER_MIN
-TextAlignment::TEXT_SIMILARITY_THRESHOLD = 0.87 unless defined? TextAlignment::TEXT_SIMILARITY_THRESHOLD
+TextAlignment::TEXT_SIMILARITY_THRESHOLD = 0.9 unless defined? TextAlignment::TEXT_SIMILARITY_THRESHOLD

data/lib/text_alignment/mixed_alignment.rb

@@ -10,8 +10,6 @@ require 'text_alignment/mappings'
 
 module TextAlignment; end unless defined? TextAlignment
 
-TextAlignment::NOMATCH_CHARS = "@^|#$%&_" unless defined? TextAlignment::NOMATCH_CHARS
-
 class TextAlignment::MixedAlignment
 	attr_reader :sdiff
 	attr_reader :position_map_begin, :position_map_end
@@ -21,58 +19,7 @@ class TextAlignment::MixedAlignment
 
 	def initialize(str1, str2, mappings = [])
 		raise ArgumentError, "nil string" if str1.nil? || str2.nil?
-		raise ArgumentError, "nil mappings" if mappings.nil?
-
-		## preprocessing
-		str1 = str1.dup
-		str2 = str2.dup
-		mappings = mappings.dup
-
-		## find the first nomatch character
-		TextAlignment::NOMATCH_CHARS.each_char do |c|
-			if str2.index(c).nil?
-				@nomatch_char1 = c
-				break
-			end
-		end
-		raise RuntimeError, "Cannot find nomatch character" if @nomatch_char1.nil?
-
-		## find the first nomatch character
-		TextAlignment::NOMATCH_CHARS.each_char do |c|
-			if c != @nomatch_char1 && str1.index(c).nil?
-				@nomatch_char2 = c
-				break
-			end
-		end
-		raise RuntimeError, "Cannot find nomatch character" if @nomatch_char2.nil?
-
-		# single character mappings
-		character_mappings = mappings.select{|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!(/-/, '\-')
-
-		str1.tr!(characters_from, characters_to)
-		str2.tr!(characters_from, characters_to)
-
-		mappings.delete_if{|m| m[0].length == 1 && m[1].length == 1}
-
-		# ASCII foldings
-		ascii_foldings = mappings.select{|m| m[0].length == 1 && m[1].length > 1}
-		ascii_foldings.each do |f|
-			from = f[1]
-
-			if str2.index(f[0])
-				to   = f[0] + (@nomatch_char1 * (f[1].length - 1))
-				str1.gsub!(from, to)
-			end
-
-			if str1.index(f[0])
-				to   = f[0] + (@nomatch_char2 * (f[1].length - 1))
-				str2.gsub!(from, to)
-			end
-		end
-		mappings.delete_if{|m| m[0].length == 1 && m[1].length > 1}
+		mappings ||= []
 
 		_compute_mixed_alignment(str1, str2, mappings)
 	end

data/lib/text_alignment/text_alignment.rb

@@ -5,19 +5,17 @@ require 'text_alignment/mixed_alignment'
 
 module TextAlignment; end unless defined? TextAlignment
 
+TextAlignment::PADDING_LETTERS = ['@', '^', '|', '#', '$', '%', '&', '_'] unless defined? TextAlignment::PADDING_LETTERS
+
 class TextAlignment::TextAlignment
 	attr_reader :block_alignments
 	attr_reader :similarity
 	attr_reader :lost_annotations
 
-	def initialize(str1, str2, _size_ngram = nil, _size_window = nil, _text_similiarity_threshold = nil)
-		raise ArgumentError, "nil string" if str1.nil? || str2.nil?
-
-		size_ngram  = _size_ngram  || TextAlignment::SIZE_NGRAM
-		size_window = _size_window || TextAlignment::SIZE_WINDOW
-		sim_threshold = _text_similiarity_threshold || TextAlignment::TEXT_SIMILARITY_THRESHOLD
+	def initialize(_str1, _str2, _size_ngram = nil, _size_window = nil, _text_similiarity_threshold = nil)
+		raise ArgumentError, "nil string" if _str1.nil? || _str2.nil?
 
-		mappings ||= TextAlignment::MAPPINGS
+		str1, str2, mappings = string_preprocessing(_str1, _str2)
 
 		# try exact match
 		block_begin = str2.index(str1)
@@ -26,7 +24,7 @@ class TextAlignment::TextAlignment
 			return @block_alignments
 		end
 
-		anchor_finder = TextAlignment::AnchorFinder.new(str1, str2, size_ngram, size_window, sim_threshold)
+		anchor_finder = TextAlignment::AnchorFinder.new(str1, str2, _size_ngram, _size_window, _text_similiarity_threshold)
 
 		# To collect matched blocks
 		mblocks = []
@@ -232,4 +230,60 @@ class TextAlignment::TextAlignment
 		r
 	end
 
+	private
+
+
+	def string_preprocessing(_str1, _str2)
+		str1 = _str1.dup
+		str2 = _str2.dup
+		mappings = TextAlignment::MAPPINGS.dup
+
+		## single character mappings
+		character_mappings = mappings.select{|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!(/-/, '\-')
+
+		str1.tr!(characters_from, characters_to)
+		str2.tr!(characters_from, characters_to)
+
+		mappings.delete_if{|m| m[0].length == 1 && m[1].length == 1}
+
+		## long to one character mappings
+		pletters = TextAlignment::PADDING_LETTERS
+
+		# find the padding letter for str1
+		padding_letter1 = begin
+			i = pletters.index{|l| str2.index(l).nil?}
+			raise RuntimeError, "Could not find a padding letter for str1" if i.nil?
+			TextAlignment::PADDING_LETTERS[i]
+		end
+
+		# find the padding letter for str2
+		padding_letter2 = begin
+			i = pletters.index{|l| l != padding_letter1 && str1.index(l).nil?}
+			raise RuntimeError, "Could not find a padding letter for str2" if i.nil?
+			TextAlignment::PADDING_LETTERS[i]
+		end
+
+		# ASCII foldings
+		ascii_foldings = mappings.select{|m| m[0].length == 1 && m[1].length > 1}
+		ascii_foldings.each do |f|
+			from = f[1]
+
+			if str2.index(f[0])
+				to   = f[0] + (padding_letter1 * (f[1].length - 1))
+				str1.gsub!(from, to)
+			end
+
+			if str1.index(f[0])
+				to   = f[0] + (padding_letter2 * (f[1].length - 1))
+				str2.gsub!(from, to)
+			end
+		end
+		mappings.delete_if{|m| m[0].length == 1 && m[1].length > 1}
+
+		[str1, str2, mappings]
+	end
+
 end

data/lib/text_alignment/version.rb

@@ -1,3 +1,3 @@
 class TextAlignment
-	VERSION = '0.4.2'
+	VERSION = '0.4.3'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: text_alignment
 version: !ruby/object:Gem::Version
-  version: 0.4.2
+  version: 0.4.3
 platform: ruby
 authors:
 - Jin-Dong Kim