text_alignment 0.12.9 → 0.12.11

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7da208b2cd252983fd0c7f8378130f7b13bba3df1c698f55e2133d54d9dab61d
-  data.tar.gz: 998405eb7c03b065faae083368a285cf526a193d7b59d22aafce643887b4150e
+  metadata.gz: 77b16002a92b4aedc56351a568698fa4032f750b49cbdf1ee4927548a278fe2a
+  data.tar.gz: 40b69153c41b6af4182f15769b5dc0eef2b988d4c838d16c3f3a803a6d03cad3
 SHA512:
-  metadata.gz: 246e0796040cedd989e12f7ef51f33a3209e0f82a396813ad9062323459e731c8d7b651a2461b732d90119c2e95586e1dbc296db91a99aeca574f4337aca6c8a
-  data.tar.gz: 06dedaa086dd878e41d26d18e427cf1d7c564583b34d047758a57e4508506cba6714abe9d8e04fa63c474ec2cc3fb9a585fef00312efe409c22b73528bdbde97
+  metadata.gz: 71d13c5a54db05fba9fd32b6eaa1cd7a714a9bef0b69c83dc9fae240e302a7aa7c06591b26749a6d4206bf25ba83ddf1411d37ee74dffd26b93a782a99926983
+  data.tar.gz: fb67eb1ecc933228ede81ecc2690a577853a9a16d12e17e0b5be696dcc661f1aa39399ff877ad567179a0a1d3d3a07134388d87dfd7a72510b56943c87339fdb

data/lib/text_alignment/anchor_finder.rb

@@ -28,6 +28,9 @@ class TextAlignment::AnchorFinder
 		# positions of last match
 		@pos_s1_last_match = 0
 		@pos_s2_last_match = 0
+
+		# Performance: cache for character classification
+		@half_ngram = @size_ngram / 2
 	end
 
 	def get_next_anchor
@@ -36,11 +39,13 @@ class TextAlignment::AnchorFinder
 		beg_s2 = for beg_s1 in @pos_s1_last_match .. @pos_s1_final_possible_begin
 			iterations += 1
 
+			char = @s1[beg_s1]
+
 			# To skip whitespace letters
-			next if [' ', "\n", "\t"].include? @s1[beg_s1]
+			next if char == ' ' || char == "\n" || char == "\t"
 
 			# Skip positions that start with punctuation or numbers (likely poor anchors)
-			next if @s1[beg_s1] =~ /[^a-zA-Z]/
+			next if char < 'A' || (char > 'Z' && char < 'a') || char > 'z'
 
 			# Performance optimization: skip if we've had too many failed attempts recently
 			if iterations > 50 && @recent_failures && @recent_failures > 10
@@ -93,11 +98,11 @@ class TextAlignment::AnchorFinder
 
 		# Quick frequency check: skip very short or very common ngrams
 		return nil if anchor.length < @size_ngram
-		return nil if anchor =~ /^(.)\1+$/ # Skip repeating character patterns like "aaaaaaaa"
+		return nil if anchor.chars.uniq.length == 1 # Skip repeating character patterns like "aaaaaaaa"
 
 		# Skip ngrams that are mostly whitespace or punctuation
 		non_alnum_count = anchor.count("^a-zA-Z0-9")
-		return nil if non_alnum_count > @size_ngram / 2
+		return nil if non_alnum_count > @half_ngram
 
 		search_position = @to_ignore_text_order ? 0 : @pos_s2_last_match
 
@@ -126,45 +131,26 @@ class TextAlignment::AnchorFinder
 	end
 
 	def find_valid_beg_s2(beg_s1, beg_s2_candidates)
-		valid_beg_s2 = nil
-
-		(10 .. 30).step(10).each do |size_window|
-			valid_beg_s2 = nil
-
-			r = beg_s2_candidates.each do |beg_s2|
-				# if both the begining points are sufficiantly close to the end points of the last match
-				# break if beg_s1 > 0 && beg_s2 > 0 && (beg_s1 - @pos_s1_last_match < 5) && (beg_s2 >= @pos_s2_last_match) && (beg_s2 - @pos_s2_last_match < 5)
+		[10, 20, 30].each do |size_window|
+			beg_s2_candidates.each do |beg_s2|
+				# if both the beginning points are sufficiently close to the end points of the last match
 				if beg_s1 > 0 && beg_s2 > 0 && (beg_s1 - @pos_s1_last_match < 5) && (beg_s2 - @pos_s2_last_match < 5)
-					break unless valid_beg_s2.nil?
-					valid_beg_s2 = beg_s2
-					next
+					return beg_s2
 				end
 
 				left_window_s1, left_window_s2 = @method_get_left_windows.call(beg_s1, beg_s2, size_window)
 				if left_window_s1 && (text_similarity(left_window_s1, left_window_s2) > @sim_threshold)
-					break unless valid_beg_s2.nil?
-					valid_beg_s2 = beg_s2
-					next
+					return beg_s2
 				end
 
 				right_window_s1, right_window_s2 = @method_get_right_windows.call(beg_s1, beg_s2, size_window)
 				if right_window_s2 && (text_similarity(right_window_s1, right_window_s2) > @sim_threshold)
-					break unless valid_beg_s2.nil?
-					valid_beg_s2 = beg_s2
-					next
+					return beg_s2
 				end
 			end
-
-			# r == nil means that the inner loop was broken (multiple candidates had passed the tests)
-			# r != nil means that the inner loop was completed (with or w/o a valid beg_s2 found)
-			if r.nil?
-				valid_beg_s2 = nil
-			else
-				break
-			end
 		end
 
-		valid_beg_s2
+		nil
 	end
 
 	def get_left_windows(beg_s1, beg_s2, size_window = nil)
@@ -173,29 +159,31 @@ class TextAlignment::AnchorFinder
 		# comment out below with the assumption that the beginning of a document gives a significant locational information
 		# return if beg_s1 < size_window || beg_s2 < size_window
 
-		window_s1 = ''
+		chars1 = []
 		loc = beg_s1 - 1
 		count = 0
 		while count < size_window && loc >= 0
-			if @s1[loc] =~ /[0-9a-zA-Z]/
-				window_s1 += @s1[loc]
+			char = @s1[loc]
+			if alnum_char?(char)
+				chars1 << char
 				count += 1
 			end
 			loc -= 1
 		end
 
-		window_s2 = ''
+		chars2 = []
 		loc = beg_s2 - 1
 		count = 0
 		while count < size_window && loc >= 0
-			if @s2[loc] =~ /[0-9a-zA-Z]/
-				window_s2 += @s2[loc]
+			char = @s2[loc]
+			if alnum_char?(char)
+				chars2 << char
 				count += 1
 			end
 			loc -= 1
 		end
 
-		[window_s1, window_s2]
+		[chars1.join, chars2.join]
 	end
 
 	def get_right_windows(beg_s1, beg_s2, size_window = nil)
@@ -204,31 +192,33 @@ class TextAlignment::AnchorFinder
 		# commend below with the assumption that the end of a document gives a significant locational
 		# return if (beg_s1 + @size_ngram > (@s1.length - size_window)) || (beg_s2 + @size_ngram > (@s2.length - size_window))
 
-		window_s1 = ''
+		chars1 = []
 		loc = beg_s1 + @size_ngram
 		len_s1 = @s1.length
 		count = 0
 		while count < size_window && loc < len_s1
-			if @s1[loc] =~ /[0-9a-zA-Z]/
-				window_s1 += @s1[loc]
+			char = @s1[loc]
+			if alnum_char?(char)
+				chars1 << char
 				count += 1
 			end
 			loc += 1
 		end
 
-		window_s2 = ''
+		chars2 = []
 		loc = beg_s2 + @size_ngram
 		len_s2 = @s2.length
 		count = 0
 		while count < size_window && loc < len_s2
-			if @s2[loc] =~ /[0-9a-zA-Z]/
-				window_s2 += @s2[loc]
+			char = @s2[loc]
+			if alnum_char?(char)
+				chars2 << char
 				count += 1
 			end
 			loc += 1
 		end
 
-		[window_s1, window_s2]
+		[chars1.join, chars2.join]
 	end
 
 	def get_left_windows_no_squeeze_ws(beg_s1, beg_s2, size_window = nil)
@@ -273,4 +263,9 @@ class TextAlignment::AnchorFinder
 		return 0 if str1.nil? || str2.nil?
 		String::Similarity.cosine(str1, str2, ngram:ngram_order)
 	end
+
+	# Fast alphanumeric character check without regex
+	def alnum_char?(char)
+		(char >= 'a' && char <= 'z') || (char >= 'A' && char <= 'Z') || (char >= '0' && char <= '9')
+	end
 end

data/lib/text_alignment/char_mapping.rb

@@ -151,29 +151,31 @@ class TextAlignment::CharMapping
 
 	def enmap_text(_text, char_mapping, no_ws = false)
 		text = _text.dup
-
-		# To perform the single letter mapping replacement
-		char_mapping.each do |one, long|
-			# text.gsub!(one, long) if long.length == 1
-			text.tr!(one, long) if long.length == 1
-		end
-
-		# To get the replacement positions, (position, old_length, new_length), for char mappings
 		rpositions = []
-		char_mapping.each do |one, long|
-			next if long.length == 1
 
-			init_next = 0
-			while loc = text.index(long, init_next)
-				# Huristics to check if the surrounding letters are sufficiently distinguished.
-				if long.length > 3 || ((text[loc - 1, 2] !~ /[a-z][a-z]/) && (text[loc + long.length - 1, 2] !~ /[a-z][a-z]/))
-				# if true
-					rpositions << [loc, long.length, 1]
+		# Skip character mapping if text is pure ASCII (performance optimization)
+		unless text.ascii_only?
+			# To perform the single letter mapping replacement
+			char_mapping.each do |one, long|
+				text.gsub!(one, long) if long.length == 1
+			end
 
-					# a workaround to avoid messing-up due to embedding
-					text[loc, long.length] = one * long.length
+			# To get the replacement positions, (position, old_length, new_length), for char mappings
+			char_mapping.each do |one, long|
+				next if long.length == 1
+
+				init_next = 0
+				while loc = text.index(long, init_next)
+					# Huristics to check if the surrounding letters are sufficiently distinguished.
+					if long.length > 3 || ((text[loc - 1, 2] !~ /[a-z][a-z]/) && (text[loc + long.length - 1, 2] !~ /[a-z][a-z]/))
+					# if true
+						rpositions << [loc, long.length, 1]
+
+						# a workaround to avoid messing-up due to embedding
+						text[loc, long.length] = one * long.length
+					end
+					init_next = loc + long.length
 				end
-				init_next = loc + long.length
 			end
 		end
 

data/lib/text_alignment/version.rb

@@ -1,3 +1,3 @@
 class TextAlignment
-	VERSION = '0.12.9'
+	VERSION = '0.12.11'
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: text_alignment
 version: !ruby/object:Gem::Version
-  version: 0.12.9
+  version: 0.12.11
 platform: ruby
 authors:
 - Jin-Dong Kim
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2025-09-30 00:00:00.000000000 Z
+date: 2025-10-06 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ruby-dictionary