canon 0.3.2 → 0.3.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b000e9e89684d66d4f037a950a2121a88717dd6c72d975648742f32b5a83fb9f
-  data.tar.gz: c5e8b5ecfdcb8435559be3e6c940e8e8bec99202054b90fa857f9e3afd3097e2
+  metadata.gz: 53a9382cf5977e8e1b65a847cd11f9f2c61a162c3913a64588e4d48c71cfbf94
+  data.tar.gz: 851f0227e56cffdd225fce23ec32b78b3b44f43e08137a39572a2db237d94aaf
 SHA512:
-  metadata.gz: '0804e7a2a07ff2687e363ee2bd0a1e2d00a5fe826e45ae5bd69fe26749bf2e2b1517a6fb2fcd3518e41d7d9bf6bf0b76fd8e704c288293724d27cbc3613978f4'
-  data.tar.gz: 6d9dca5aa3c29bbde5d067ca02e2a22c4d7bee098635e0fd8b48df16fbf08102fa713608f1965d198f3ace1c064743916b57d34d7a2cc2a6b845902e1553a045
+  metadata.gz: 67710ade1d18f99c23e4477540303f45867d145c7f92df59f2f1277cf442c576eca2a3d8562cb92972f56d80dbe111e6104d28353cb7c5091d690490605772ff
+  data.tar.gz: e8d2783baa8fba878cb3cbb7a602dff3563f23dfb15df95ca5c24c075475d53c8319f2579bf3ce3c533e1989b1c75b91b72913c57c30279278a59de68202aef0

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.2"
+  VERSION = "0.3.4"
 end

data/lib/canon/xml/data_model.rb

@@ -126,28 +126,49 @@ module Canon
         end
       end
 
+      # --- Shared construction kernels (identical semantics per engine) ---
+
+      # Attach document-level children (prolog/epilog PIs, comments,
+      # document-level text) to the canon root, in document order,
+      # skipping the document element and the doctype. `converter`
+      # converts one engine child to a canon node (or nil).
+      def self.add_document_children(root, children, document_element,
+                                      skip_types, converter)
+        children.each do |child|
+          next if child.equal?(document_element)
+          next if skip_types.any? { |type| child.is_a?(type) }
+
+          node = converter.call(child)
+          root.add_child(node) if node
+        end
+      end
+
+      # In-scope namespace bindings: the element's own declarations
+      # shadow inherited ones; xml is prebound at the base. Declaration
+      # pairs are [prefix-or-nil, uri]; nil and "" both mean the default
+      # namespace.
+      def self.merge_namespace_scope(inherited, declaration_pairs)
+        scope = inherited ? inherited.dup : { "xml" => "http://www.w3.org/XML/1998/namespace" }
+        declaration_pairs.each do |prefix, uri|
+          scope[prefix || ""] = uri
+        end
+        scope
+      end
+
       def self.build_from_nokogiri(nokogiri_doc, preserve_whitespace: false)
         root = Nodes::RootNode.new
 
+        converter = ->(child) { build_node_from_nokogiri(child, preserve_whitespace: preserve_whitespace) }
+        skip_types = defined?(Nokogiri) ? [Nokogiri::XML::DTD] : []
+
         if nokogiri_doc.is_a?(Nokogiri::XML::Document) && nokogiri_doc.root
           root.add_child(build_element_node(nokogiri_doc.root,
                                             preserve_whitespace: preserve_whitespace))
-          nokogiri_doc.children.each do |child|
-            next if child == nokogiri_doc.root
-            next if child.is_a?(Nokogiri::XML::DTD)
-
-            node = build_node_from_nokogiri(child,
-                                            preserve_whitespace: preserve_whitespace)
-            root.add_child(node) if node
-          end
+          add_document_children(root, nokogiri_doc.children, nokogiri_doc.root,
+                                skip_types, converter)
         else
-          nokogiri_doc.children.each do |child|
-            next if child.is_a?(Nokogiri::XML::DTD)
-
-            node = build_node_from_nokogiri(child,
-                                            preserve_whitespace: preserve_whitespace)
-            root.add_child(node) if node
-          end
+          add_document_children(root, nokogiri_doc.children, nil,
+                                skip_types, converter)
         end
 
         root
@@ -198,16 +219,14 @@ inherited_namespaces: nil)
         element
       end
 
-      # In-scope namespace bindings: the element's own declarations
-      # shadow inherited ones, xml is prebound. The scope is passed down
+      # In-scope bindings via the shared kernel; 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
-        scope
+        merge_namespace_scope(inherited,
+                              nokogiri_element.namespace_definitions.map do |ns|
+                                [ns.prefix, ns.href]
+                              end)
       end
 
       def self.build_attribute_nodes(nokogiri_element, element)
@@ -228,7 +247,9 @@ inherited_namespaces: nil)
       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
 
@@ -281,26 +302,18 @@ inherited_namespaces: nil)
       def self.build_from_moxml(moxml_doc, preserve_whitespace: false)
         root = Nodes::RootNode.new
 
+        converter = ->(child) { build_moxml_node(child, preserve_whitespace: preserve_whitespace) }
+        skip_types = [Moxml::Doctype]
+
         if moxml_doc.is_a?(Moxml::Document) && moxml_doc.root
           element = build_moxml_subtree_from_records(moxml_doc.root,
                                                      preserve_whitespace: preserve_whitespace)
           root.add_child(element) if element
-          moxml_doc.children.each do |child|
-            next if child.equal?(moxml_doc.root)
-            next if child.is_a?(Moxml::Doctype)
-
-            node = build_moxml_node(child,
-                                    preserve_whitespace: preserve_whitespace)
-            root.add_child(node) if node
-          end
+          add_document_children(root, moxml_doc.children, moxml_doc.root,
+                                skip_types, converter)
         else
-          moxml_doc.children.each do |child|
-            next if child.is_a?(Moxml::Doctype)
-
-            node = build_moxml_node(child,
-                                    preserve_whitespace: preserve_whitespace)
-            root.add_child(node) if node
-          end
+          add_document_children(root, moxml_doc.children, nil,
+                                skip_types, converter)
         end
 
         root
@@ -394,19 +407,16 @@ 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
 
-      # Expand each element's own declarations to in-scope bindings:
-      # the element's declarations shadow inherited ones; xml is
-      # prebound. Scope flows down the canon tree — no engine calls.
+      # Expand each element's own declarations to in-scope bindings via
+      # the shared kernel. Scope flows down the canon tree — no engine
+      # calls.
       def self.assign_moxml_namespace_scopes(element, inherited, own_namespaces)
-        scope = inherited.dup
-        own_namespaces[element].to_a.each do |prefix, uri|
-          scope[prefix || ""] = uri
-        end
+        scope = merge_namespace_scope(inherited, own_namespaces[element].to_a)
 
         scope.each do |prefix, uri|
           element.add_namespace(Nodes::NamespaceNode.new(
@@ -472,11 +482,10 @@ inherited_namespaces: nil)
       # of an ancestor walk through the adapter layer (the Nokogiri
       # collector walks ancestors because those calls are cheap there).
       def self.moxml_namespace_scope(moxml_element, inherited)
-        scope = inherited ? inherited.dup : { "xml" => "http://www.w3.org/XML/1998/namespace" }
-        moxml_element.namespace_definitions.each do |decl|
-          scope[decl.prefix || ""] = decl.uri
-        end
-        scope
+        merge_namespace_scope(inherited,
+                              moxml_element.namespace_definitions.map do |decl|
+                                [decl.prefix, decl.uri]
+                              end)
       end
 
       def self.build_moxml_attribute_nodes(moxml_element, element)
@@ -502,7 +511,9 @@ 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: 0.3.2
+  version: 0.3.4
 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