canon 0.3.1 → 0.3.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 8180c4d10ad157f5fd7d401d7c4723a70d53daa90894a43c33016253d0355093
-  data.tar.gz: 513538a92c8f209b2eeec7ecf2adedf9e2a86de8e2c6c0e807b45522afffa9b3
+  metadata.gz: 52174949c77e8f33286b2875ac5e4052e6f5671b64e53e48a3b0b3128bb598a1
+  data.tar.gz: 0b2efb1c4fa7e060f6d54f1c149bd775d45fab66e09ac7bbf306f0a6142360fa
 SHA512:
-  metadata.gz: 15bbb37c41f2fe0809be3a1ae9e6e3bf438d7d509eb478b92eb1ace044d511017a257409024c0c7e042bf5dc146af0dbbfc9397c6ffea13a5d410fb9019c716d
-  data.tar.gz: 81c275d67d47bfeb87a1c5969e90643da41e8816e882a40fa5affa8337fbd3c5bd0ee82428ef63b36bc5dff8338776e8df4da53394d3319663f3167433664785
+  metadata.gz: 4e78d8618af59c255979bf3987a1be3e7419e009354b73ca64df75600ce8c2f1cc90aa515eb14be97149fe5426243efacda4ff42eab63b5fc8f291407baa046e
+  data.tar.gz: 2596286ac94921e6021590fe6f1ece71d827145743cc993f703cfca079fb91472845c67f25a3e9ad03aecd471c6cc482cd78854a5e9ce85d16bd8f0ea5643636

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.1"
+  VERSION = "0.3.3"
 end

data/lib/canon/xml/data_model.rb

@@ -154,11 +154,12 @@ module Canon
       end
 
       def self.build_node_from_nokogiri(nokogiri_node,
-preserve_whitespace: false)
+preserve_whitespace: false, inherited_namespaces: nil)
         case nokogiri_node
         when Nokogiri::XML::Element
           build_element_node(nokogiri_node,
-                             preserve_whitespace: preserve_whitespace)
+                             preserve_whitespace: preserve_whitespace,
+                             inherited_namespaces: inherited_namespaces)
         when Nokogiri::XML::Text, Nokogiri::XML::CDATA
           build_text_node(nokogiri_node,
                           preserve_whitespace: preserve_whitespace)
@@ -169,56 +170,44 @@ preserve_whitespace: false)
         end
       end
 
-      def self.build_element_node(nokogiri_element, preserve_whitespace: false)
+      def self.build_element_node(nokogiri_element, preserve_whitespace: false,
+inherited_namespaces: nil)
         element = Nodes::ElementNode.new(
           name: nokogiri_element.name,
           namespace_uri: nokogiri_element.namespace&.href,
           prefix: nokogiri_element.namespace&.prefix,
         )
 
-        build_namespace_nodes(nokogiri_element, element)
+        scope = nokogiri_namespace_scope(nokogiri_element, inherited_namespaces)
+        scope.each do |prefix, uri|
+          element.add_namespace(Nodes::NamespaceNode.new(
+                                  prefix: prefix,
+                                  uri: uri,
+                                ))
+        end
+
         build_attribute_nodes(nokogiri_element, element)
 
         nokogiri_element.children.each do |child|
           node = build_node_from_nokogiri(child,
-                                          preserve_whitespace: preserve_whitespace)
+                                          preserve_whitespace: preserve_whitespace,
+                                          inherited_namespaces: scope)
           element.add_child(node) if node
         end
 
         element
       end
 
-      def self.build_namespace_nodes(nokogiri_element, element)
-        namespaces = collect_in_scope_namespaces(nokogiri_element)
-
-        namespaces.each do |prefix, uri|
-          ns_node = Nodes::NamespaceNode.new(
-            prefix: prefix,
-            uri: uri,
-          )
-          element.add_namespace(ns_node)
-        end
-      end
-
-      def self.collect_in_scope_namespaces(nokogiri_element)
-        namespaces = {}
-
-        current = nokogiri_element
-        while current && !current.is_a?(Nokogiri::XML::Document)
-          if current.is_a?(Nokogiri::XML::Element)
-            current.namespace_definitions.each do |ns|
-              prefix = ns.prefix || ""
-              unless namespaces.key?(prefix)
-                namespaces[prefix] = ns.href
-              end
-            end
-          end
-          current = current.parent
+      # In-scope namespace bindings: the element's own declarations
+      # shadow inherited ones, xml is prebound. The scope is passed down
+      # the recursion — one definition fetch per element instead of an
+      # ancestor walk per element (O(n) total, not O(n * depth)).
+      def self.nokogiri_namespace_scope(nokogiri_element, inherited)
+        scope = inherited ? inherited.dup : { "xml" => "http://www.w3.org/XML/1998/namespace" }
+        nokogiri_element.namespace_definitions.each do |ns|
+          scope[ns.prefix || ""] = ns.href
         end
-
-        namespaces["xml"] ||= "http://www.w3.org/XML/1998/namespace"
-
-        namespaces
+        scope
       end
 
       def self.build_attribute_nodes(nokogiri_element, element)
@@ -239,7 +228,9 @@ preserve_whitespace: false)
       def self.build_text_node(nokogiri_text, preserve_whitespace: false)
         content = nokogiri_text.content
 
-        if !preserve_whitespace && content.strip.empty? && nokogiri_text.parent.is_a?(Nokogiri::XML::Element)
+        unless WhitespacePolicy.keep_dom_text?(content,
+                                               preserve_whitespace: preserve_whitespace,
+                                               element_parent: nokogiri_text.parent.is_a?(Nokogiri::XML::Element))
           return nil
         end
 
@@ -405,7 +396,7 @@ preserve_whitespace: false)
       # whitespace-only skip rule matches the wrapper path exactly.
       def self.build_moxml_text_from_record(record, preserve_whitespace)
         content = record[:text].to_s
-        return nil if !preserve_whitespace && content.strip.empty?
+        return nil unless WhitespacePolicy.keep_dom_text?(content, preserve_whitespace: preserve_whitespace)
 
         Nodes::TextNode.new(value: content, original: content)
       end
@@ -513,7 +504,9 @@ inherited_namespaces: nil)
       def self.build_moxml_text_node(moxml_text, preserve_whitespace: false)
         content = moxml_text.content
 
-        if !preserve_whitespace && content.strip.empty? && moxml_text.parent.is_a?(Moxml::Element)
+        unless WhitespacePolicy.keep_dom_text?(content,
+                                               preserve_whitespace: preserve_whitespace,
+                                               element_parent: moxml_text.parent.is_a?(Moxml::Element))
           return nil
         end
 

data/lib/canon/xml/sax.rb

@@ -10,13 +10,16 @@ module Canon
     # API to that protocol. OCP: a new engine is a new driver plus one line
     # here; the builder never changes.
     #
-    # Engine choice: leptris SAX outperforms Nokogiri SAX (1.10x min-of-N
-    # on canon's builder, 53KB doc) through moxml 0.5.12's
-    # interest-proportional callbacks — BUT the batch-attribute transport
-    # reads uninitialized memory at certain buffer offsets, corrupting the
-    # first N attributes of an element (leptris-ruby#95). Until that is
-    # fixed, CRuby keeps the Nokogiri driver; flipping is the one-line
-    # change below.
+    # Engine choice: the leptris SAX flip is armed but blocked twice over.
+    # The batch-attribute corruption (leptris-ruby#95) is answered by a
+    # correctness-first path in leptris 1.9.36 — attributes are correct,
+    # but that path DROPS namespace declarations from SAX events
+    # entirely (raw `["<root xmlns:a=.../>", []]` — xmlns pairs vanish),
+    # blinding namespace comparison. Until xmlns reporting is restored
+    # (and then the fast path returns with the C fix), CRuby keeps the
+    # Nokogiri driver; the flip is the one-line change below. With
+    # leptris absent (moxml resolving nokogiri) the driver calls Nokogiri
+    # directly — the wrapper layer only adds overhead there.
     module Sax
       autoload :NokogiriDriver, "canon/xml/sax/nokogiri_driver"
       autoload :MoxmlDriver, "canon/xml/sax/moxml_driver"

data/lib/canon/xml/sax_builder.rb

@@ -198,22 +198,12 @@ strip_doctype: false)
           return
         end
 
-        # Skip whitespace-only text nodes unless:
-        # 1. preserve_whitespace is true, OR
-        # 2. The content contains CR (from 
 entities) which must be preserved for C14N, OR
-        # 3. The content contains non-ASCII whitespace (NBSP U+00A0, ideographic
-        #    space U+3000, etc.) — those are semantically meaningful content,
-        #    not pretty-print indentation, and must survive parsing so the
-        #    comparator can detect Unicode whitespace-type differences.
-        #
-        # Strip only when the node is pure ASCII whitespace (space, tab, CR, LF).
-        # This lets pretty-printed fixtures work (indent nodes stripped) while
-        # preserving NBSP-only text nodes.
-        if !@preserve_whitespace && decoded_string.gsub(/[ \t\r\n]/,
-                                                        "").empty? && parent.node_type == :element && !decoded_string.include?("\r")
-          # Only skip if parent is an element (not root)
-          return
-        end
+        # Skip whitespace-only text nodes unless preserving, per the SAX
+        # policy (CR-bearing content always survives for C14N;
+        # non-ASCII whitespace is meaningful content).
+        return unless WhitespacePolicy.keep_sax_text?(decoded_string,
+                                                      preserve_whitespace: @preserve_whitespace,
+                                                      element_parent: parent.node_type == :element)
 
         text = Nodes::TextNode.new(value: decoded_string, original: raw_string)
         parent.add_child(text)

data/lib/canon/xml/whitespace_policy.rb

@@ -0,0 +1,37 @@
+# frozen_string_literal: true
+
+module Canon
+  module Xml
+    # Parse-time whitespace policy: whether a character-data node
+    # survives conversion. One home for the keep/strip rules so the
+    # DOM/SAX differences are visible here instead of implied by
+    # copy-paste across conversion sites.
+    module WhitespacePolicy
+      module_function
+
+      # DOM conversion rule: whitespace-only text is dropped unless
+      # preserving. Non-ASCII whitespace (NBSP, U+3000) survives —
+      # String#strip only removes ASCII whitespace.
+      #
+      # NOTE: CR-only nodes are dropped on this path; the SAX rule keeps
+      # them (character references must survive for C14N).
+      def keep_dom_text?(content, preserve_whitespace:, element_parent: true)
+        return true if preserve_whitespace
+        return true unless element_parent
+
+        !content.strip.empty?
+      end
+
+      # SAX rule: same shape, plus CR-bearing content is always kept
+      # (
 must survive parsing for C14N) — only runs of pure ASCII
+      # whitespace (space, tab, CR, LF) are dropped when not preserving.
+      def keep_sax_text?(content, preserve_whitespace:, element_parent: true)
+        return true if preserve_whitespace
+        return true unless element_parent
+        return true if content.include?("\r")
+
+        !content.gsub(/[ \t\r\n]/, "").empty?
+      end
+    end
+  end
+end

data/lib/canon/xml.rb

@@ -28,6 +28,7 @@ module Canon
     autoload :Sax, "canon/xml/sax"
     autoload :SaxBuilder, "canon/xml/sax_builder"
     autoload :WhitespaceNormalizer, "canon/xml/whitespace_normalizer"
+    autoload :WhitespacePolicy, "canon/xml/whitespace_policy"
     autoload :XmlBaseHandler, "canon/xml/xml_base_handler"
     autoload :XPathEngine, "canon/xml/xpath_engine"
   end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.1
+  version: 0.3.3
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-08-28 00:00:00.000000000 Z
+date: 2026-08-29 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: diff-lcs
@@ -401,6 +401,7 @@ files:
 - lib/canon/xml/sax/nokogiri_driver.rb
 - lib/canon/xml/sax_builder.rb
 - lib/canon/xml/whitespace_normalizer.rb
+- lib/canon/xml/whitespace_policy.rb
 - lib/canon/xml/xml_base_handler.rb
 - lib/canon/xml/xpath_engine.rb
 - lib/canon/xml_backend.rb