canon 0.3.35 → 0.3.36

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 746a0b0b72cc9111986fb3aab9b8c5de994d71b30a011983665b5a24130b092c
-  data.tar.gz: e9e5a262b99c0e44c2f11385e911faaf5216b52f32bc027d1a0ae82f0bae985e
+  metadata.gz: 53129bc040b25503b792d4c1afa96498715cd05363d462afb87aa004dbe37004
+  data.tar.gz: de9d2c032c3b20e73f6737bcd0d092130dd9622851cca6c9353d6c10ea7da0f8
 SHA512:
-  metadata.gz: 9edbbe4c9eb6adb6218b2d282f1ac030fa4642ece72cedae34b54ea6cbac987acfd46bebbf1277da017b4bfcf511be98f41c64951fee9d7034d1410f8414a38b
-  data.tar.gz: 23d088e8fb394d5f36553b64f9e05955f788ca85615147878d6116552ed2ac3b297ebdc739604de009c754a5a8600a57cad97668c6c325a018635d6c6010df04
+  metadata.gz: 2471eebf546ec921fa8d307a02363b9d1112ef5111fd0f0edd57d1894a3e04e7039965fede0a4b4f8e02fa05540dad05324e6dc1e052676512407a7965c0ea83
+  data.tar.gz: a58130908f35634360d689eba9e27d48845f955c9cd44eae6eab771e0bfb0b84af35cd97db05137eb1563baa3449a1591237fc3eac01a74ef839df5c3811e386

data/lib/canon/diff/diff_node_enricher.rb

@@ -613,13 +613,17 @@ module Canon
         # occurrence. Shorter strings match too many places in the document.
         min_locate_length = 3
 
-        # Element added (only in text2)
+        # Element added (only in text2). Index-aware first — see the
+        # removed branch note (identical-sibling anchor bug).
         if before.nil?
-          loc = if after.length < min_locate_length && path
-                  locate_via_parent_element(path, @text2, @line_map2)
-                else
-                  SourceLocator.locate(after, @text2, @line_map2)
+          loc = if path
+                  locate_element_at_index(after, @text2, @line_map2, path)
                 end
+          loc ||= if after.length < min_locate_length && path
+                    locate_via_parent_element(path, @text2, @line_map2)
+                  else
+                    SourceLocator.locate(after, @text2, @line_map2)
+                  end
 
           if loc
             end_line = find_end_line(loc[:line_number], @line_map2, after)
@@ -643,13 +647,20 @@ module Canon
           return
         end
 
-        # Element removed (only in text1)
+        # Element removed (only in text1). The path-aware locate runs
+        # FIRST: an identical sibling earlier in the document would win
+        # a plain first-occurrence search, anchoring the deletion to
+        # the wrong element and letting the line builder absorb it
+        # into a formatting gap (issue #85/#86).
         if after.nil?
-          loc = if before.length < min_locate_length && path
-                  locate_via_parent_element(path, @text1, @line_map1)
-                else
-                  SourceLocator.locate(before, @text1, @line_map1)
+          loc = if path
+                  locate_element_at_index(before, @text1, @line_map1, path)
                 end
+          loc ||= if before.length < min_locate_length && path
+                    locate_via_parent_element(path, @text1, @line_map1)
+                  else
+                    SourceLocator.locate(before, @text1, @line_map1)
+                  end
 
           if loc
             end_line = find_end_line(loc[:line_number], @line_map1, before)
@@ -982,6 +993,39 @@ module Canon
         SourceLocator.locate(value, text, line_map)
       end
 
+      # Index-aware location for element_structure changes: the path's
+      # last bracketed segment names the element itself (unlike
+      # text_content paths, whose last segment is the text node).
+      # Picks the occurrence whose element index matches the path, so
+      # identical siblings anchor at the right one; falls back to the
+      # first occurrence when no indexed match exists.
+      def locate_element_at_index(value, text, line_map, path)
+        return nil if value.nil? || value.length < 3
+
+        segments = path.split("/").reject(&:empty?)
+        element_segment = segments.reverse.find { |seg| seg.include?("[") }
+        return nil unless element_segment
+
+        element_match = element_segment.match(/([a-zA-Z0-9_:-]+)\[(\d+)\]/)
+        return nil unless element_match
+
+        element_name = element_match[1]
+        target_index = element_match[2].to_i
+
+        # Occurrence offsets point AT the element's own opening tag,
+        # so the prefix scan counts exactly the preceding siblings —
+        # no "inside the element" correction (count_elements_before_
+        # position subtracts one for text-node offsets).
+        opener = /<#{element_name}[>\s]/
+        occurrences = SourceLocator.locate_all(value, text, line_map)
+        occurrences.each do |occ|
+          count = text[0...occ[:char_offset]].scan(opener).length
+          return occ if count == target_index
+        end
+
+        SourceLocator.locate(value, text, line_map)
+      end
+
       # Fallback location strategy for text_content when locate_at_element_index
       # fails (e.g., the text value is too short to locate reliably).
       # Walks the full element hierarchy from the path to locate the correct

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.35"
+  VERSION = "0.3.36"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.35
+  version: 0.3.36
 platform: ruby
 authors:
 - Ribose Inc.