canon 0.3.14 → 0.3.15

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d9954ac6056439abc3c5b211ecf17b9af2391ad0b809eb3c239fef55cbdc723a
-  data.tar.gz: dbe215e817f160c840769b5bbd0969fe83d315d2b4429cec125de43312ec821a
+  metadata.gz: d48fa5b4488a85c95b613bcab72881a5e286fdb9859da76aafc78553ff506419
+  data.tar.gz: f0bd3b76b81a0181dc4ae801bed2288828f8d8bc85a6414a98c993d10d770b7a
 SHA512:
-  metadata.gz: b1552e6f9a7735903f10b8f73902c6896cfab41b61da29b9bf725c6d5a1132483566eec58eeeb5ce2a59714a5e228ba8cb2979fa79f853cf5bd361f5ffcd849f
-  data.tar.gz: c2b93d2248033a3680164769ab57163f44935314a0929155fa98e2a488f1ea278bff8dba5bbdc38b19d313db13e9ddc6ffeb8b56cdf7537d1ba5c08098ec9bb2
+  metadata.gz: 1b32e0477c8932748179074f076478747d3eb1bce3892e6d7fd8ecae3e1dc7e686fa51907cdbd6444ad87267eb9521206b4caaa6f2d85a93b36f5cc2b3758e4a
+  data.tar.gz: 11c89d97b5144a557746e7d84e9128db443e8951fc40381486da826935f012ccc15249de122232a0ca95b32b5c32ebffe64a2415b6db8f8b9790af2e1eda1b10

data/lib/canon/comparison/match_options.rb

@@ -70,6 +70,10 @@ module Canon
           when :strict
             text1 == text2
           when :normalize
+            # Identical strings normalize identically — skip the two
+            # gsub+strip chains.
+            return true if text1 == text2
+
             if whitespace_type == :normalize
               normalize_text(text1) == normalize_text(text2)
             else
@@ -92,22 +96,34 @@ module Canon
             .strip # Remove leading/trailing whitespace
         end
 
-        # Normalize text preserving Unicode whitespace type distinctions.
         # Fast form of `normalize_text(text).empty?`, called per text
         # node by node_excluded?. Pure-ASCII whitespace (the common
         # case — pretty-print indentation) matches a plain class with
         # no intermediate strings and no \p{} property (Opal's JS
-        # regexes do not honor \p{Space}); anything else falls back to
-        # normalize_text's exact semantics. NUL is included because
-        # String#strip strips nulls too.
+        # regexes do not honor \p{Space}); a pure-ASCII miss is
+        # conclusively non-whitespace; only non-ASCII text falls back
+        # to normalize_text's exact Unicode semantics. NUL is included
+        # because String#strip strips nulls too.
         ASCII_WHITESPACE_ONLY = /\A[ \t\r\n\v\f\x00]*\z/
+        ASCII_ONLY = /\A[\x00-\x7f]*\z/
 
         def whitespace_only?(text)
           text = text.to_s
-          text.empty? || text.match?(ASCII_WHITESPACE_ONLY) ||
-            normalize_text(text).empty?
+          return true if text.empty?
+          return true if text.match?(ASCII_WHITESPACE_ONLY)
+          # A pure-ASCII string that failed the class contains an ASCII
+          # non-whitespace character, which survives both the collapse
+          # and the strip — conclusively not whitespace-only, with no
+          # intermediate strings. Only non-ASCII text (NBSP, U+3000,
+          # ...) needs normalize_text's exact Unicode semantics — and
+          # no \p{} classes appear in the fast paths, which Opal's JS
+          # regexes would not honor anyway.
+          return false if text.match?(ASCII_ONLY)
+
+          normalize_text(text).empty?
         end
 
+        # Normalize text preserving Unicode whitespace type distinctions.
         def normalize_text_preserving_type(text)
           return "" if text.nil?
 

data/lib/canon/comparison/xml_comparator/attribute_comparator.rb

@@ -26,6 +26,14 @@ module Canon
           match_opts = opts[:match_opts]
           attribute_order_behavior = match_opts[:attribute_order] || :strict
 
+          # FAST PATH: identical filtered attributes in identical key
+          # order — exact equality implies behavioral equality for every
+          # value behavior, and no order difference exists. Most element
+          # pairs hit this and skip the sort/hash/value work entirely.
+          if attrs1 == attrs2 && attrs1.keys == attrs2.keys
+            return Comparison::EQUIVALENT
+          end
+
           # Check attribute order if not ignored
           keys1 = attrs1.keys.map(&:to_s)
           keys2 = attrs2.keys.map(&:to_s)

data/lib/canon/comparison/xml_comparator/namespace_comparator.rb

@@ -17,6 +17,9 @@ module Canon
           ns_decls1 = extract_declarations(node1)
           ns_decls2 = extract_declarations(node2)
 
+          # Most element pairs declare nothing — skip the set algebra.
+          return Comparison::EQUIVALENT if ns_decls1.empty? && ns_decls2.empty?
+
           # Find missing, extra, and changed namespace declarations
           missing = ns_decls1.keys - ns_decls2.keys  # In node1 but not node2
           extra = ns_decls2.keys - ns_decls1.keys    # In node2 but not node1

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.14"
+  VERSION = "0.3.15"
 end

data/lib/canon/xml/node.rb

@@ -4,6 +4,13 @@ module Canon
   module Xml
     # Base class for all XPath data model nodes
     class Node
+      # Shared by every childless node: leaves (text, attributes,
+      # namespaces, comments, PIs) and empty elements read `children`
+      # without materializing a private array. Mutating the returned
+      # array is a bug — writers go through add_child/children=, which
+      # install a private array first.
+      EMPTY_CHILDREN = [].freeze
+
       attr_reader :parent
 
       def initialize
@@ -12,12 +19,8 @@ module Canon
         @in_node_set = true
       end
 
-      # Leaf nodes (text, attributes, namespaces, comments, PIs) never
-      # gain children; allocating the array eagerly cost one Array per
-      # node — the single largest retained-allocation source in a built
-      # tree — so it materializes on first add or read.
       def children
-        @children ||= []
+        @children || EMPTY_CHILDREN
       end
 
       def children=(new_children)
@@ -26,7 +29,7 @@ module Canon
 
       def add_child(child)
         child.parent = self
-        children << child
+        (@children ||= []) << child
       end
 
       def in_node_set?

data/lib/canon/xml/nodes/element_node.rb

@@ -15,12 +15,16 @@ module Canon
           @attribute_nodes = nil
         end
 
-        # Lazy: most elements carry no attributes, and inherited
-        # namespace nodes arrive as one shared frozen array (see
-        # TreeBuilder#attach_namespace_scope) — eager per-element arrays
-        # were the largest retained-allocation source in a built tree.
+        # Lazy: attribute-free elements (common in real documents) share
+        # one frozen empty array on read instead of materializing one
+        # per element; writers install a private array first. Inherited
+        # namespace nodes arrive as one shared frozen array from
+        # TreeBuilder#attach_namespace_scope.
+        EMPTY_ATTRIBUTE_NODES = [].freeze
+        EMPTY_NAMESPACE_NODES = [].freeze
+
         def namespace_nodes
-          @namespace_nodes ||= []
+          @namespace_nodes || EMPTY_NAMESPACE_NODES
         end
 
         def namespace_nodes=(nodes)
@@ -28,7 +32,7 @@ module Canon
         end
 
         def attribute_nodes
-          @attribute_nodes ||= []
+          @attribute_nodes || EMPTY_ATTRIBUTE_NODES
         end
 
         def node_type
@@ -41,12 +45,12 @@ module Canon
 
         def add_namespace(namespace_node)
           namespace_node.parent = self
-          namespace_nodes << namespace_node
+          (@namespace_nodes ||= []) << namespace_node
         end
 
         def add_attribute(attribute_node)
           attribute_node.parent = self
-          attribute_nodes << attribute_node
+          (@attribute_nodes ||= []) << attribute_node
         end
 
         # Get namespace nodes in sorted order (lexicographically by local name)

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.14
+  version: 0.3.15
 platform: ruby
 authors:
 - Ribose Inc.