canon 0.3.19 → 0.3.21

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b04c8b44b470bb85f2ec003f434a0fa8ede8eec74f687b388241814e60841415
-  data.tar.gz: a944aa18b2dd0993362d15a92cf70488a0618841445873c617c206a5e6fe4f72
+  metadata.gz: fd899c17d1512c815d189076fe08721dbe66b7d964f8bac69abde35f207b8c76
+  data.tar.gz: 205ad0b365288bf883f2ce57f68b612217151e30ae90fb0f959b29f252063826
 SHA512:
-  metadata.gz: dfc4366fc597d9a8c33201cf7ed4857f3577e791d0fc80160c0e6dbeda359f5c7ea0d8df49f4dcec22ffe7c596e1a694ec114b8005bce7275f9e6572b3643f77
-  data.tar.gz: 3b00f6d66f7c75e3f4dccb224347b3092ea4713aa44906c1bc623b3fee6086f940d9c13b5af10c8c60f063d8e16e82a99adc412e662cb4fe092bb23f8129c1b4
+  metadata.gz: 5aa4d552592cda5a55417415efb225151eb0d39d899fe0471eb26f10ceeca64ae56de786d8e807e0085264ba800cfd5937a6ec15e182e71957381414f3036105
+  data.tar.gz: b52ceb55e3ca6c6affcca3c7091a0e3f6e22d1c077b65732e7073915c3f6914cb86a61e675b11636baeb1ea09e94c223c933e961c0b366fb3e621c592a50170b

data/lib/canon/cache.rb

@@ -44,7 +44,8 @@ module Canon
           cache.delete(oldest_key) if oldest_key
         end
 
-        cache[key] = { value: value, accessed: Time.now }
+        @clock = (@clock || 0) + 1
+        cache[key] = { value: value, accessed: @clock }
         value
       end
 
@@ -77,12 +78,6 @@ module Canon
         "doc:#{format}:#{preprocessing}:#{content_hash(content)}"
       end
 
-      # Generate cache key for format detection
-      def key_for_format_detection(content)
-        preview = content[0..100].b
-        "fmt:#{content_hash(preview + content.length.to_s)}"
-      end
-
       # Generate cache key for XML canonicalization
       def key_for_c14n(content, with_comments)
         "c14n:#{with_comments}:#{content_hash(content)}"

data/lib/canon/comparison/format_detector.rb

@@ -53,15 +53,16 @@ module Canon
           raise Canon::Error, "Unknown format for object: #{obj.class}"
         end
 
-        # Detect the format of a string with caching
+        # Detect the format of a string
         #
         # @param str [String] String to detect format of
         # @return [Symbol] Format type
         def detect_string(str)
-          # Use cache for format detection
-          Cache.fetch(:format_detect, Cache.key_for_format_detection(str)) do # rubocop:disable Lint/UselessDefaultValueArgument
-            detect_string_uncached(str)
-          end
+          # Detection is a handful of prefix checks and one anchored
+          # regex — building the SHA256 cache key (plus LRU clock
+          # bookkeeping) cost several times the detection itself, so
+          # the answer is computed directly.
+          detect_string_uncached(str)
         end
 
         # Detect the format of a string without caching

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.19"
+  VERSION = "0.3.21"
 end

data/lib/canon/xml/character_encoder.rb

@@ -5,10 +5,18 @@ module Canon
     # Character encoder for C14N 1.1
     # Handles UTF-8 encoding and character reference encoding per spec
     class CharacterEncoder
+      # Most text and attribute values contain nothing to escape —
+      # the zero-allocation guard keeps those from paying the gsub
+      # copy on every render.
+      TEXT_ESCAPABLE = /[&<>\r]/
+      ATTRIBUTE_ESCAPABLE = /[&<"\t\n\r]/
+
       # Encode text node content
       # Replace: & → &amp;, < → &lt;, > → &gt;, #xD → &#xD;
       def encode_text(text)
-        text.gsub(/[&<>\r]/) do |char|
+        return text unless text.match?(TEXT_ESCAPABLE)
+
+        text.gsub(TEXT_ESCAPABLE) do |char|
           case char
           when "&" then "&amp;"
           when "<" then "&lt;"
@@ -22,7 +30,9 @@ module Canon
       # Replace: & → &amp;, < → &lt;, " → &quot;,
       #          #x9 → &#x9;, #xA → &#xA;, #xD → &#xD;
       def encode_attribute(value)
-        value.gsub(/[&<"\t\n\r]/) do |char|
+        return value unless value.match?(ATTRIBUTE_ESCAPABLE)
+
+        value.gsub(ATTRIBUTE_ESCAPABLE) do |char|
           case char
           when "&" then "&amp;"
           when "<" then "&lt;"

data/lib/canon/xml/namespace_handler.rb

@@ -14,7 +14,7 @@ module Canon
       def process_namespaces(element, output, parent_element = nil)
         return unless element.in_node_set?
 
-        namespaces = element.sorted_namespace_nodes.select(&:in_node_set?)
+        namespaces = renderable_namespaces(element)
 
         # Check if we need to emit xmlns="" for empty default namespace
         if should_emit_empty_default_namespace?(element, namespaces,
@@ -61,26 +61,45 @@ module Canon
       end
 
       # Check if a namespace node should be skipped
-      def should_skip_namespace?(ns, _element, parent_element)
+      def should_skip_namespace?(ns, element, parent_element)
         # Skip xml namespace with standard URI
         return true if ns.xml_namespace?
 
         # Skip if an ancestor already declared this namespace
-        return true if namespace_declared_by_ancestor?(ns, parent_element)
+        return true if namespace_declared_by_ancestor?(ns, element,
+                                                       parent_element)
 
         false
       end
 
       # Check if a namespace is already declared by an ancestor
-      def namespace_declared_by_ancestor?(ns, parent_element)
+      def namespace_declared_by_ancestor?(ns, element, parent_element)
         return false unless parent_element
 
-        parent_ns = parent_element.namespace_nodes.find do |parent_ns|
+        parent_nodes = parent_element.namespace_nodes
+        # Elements that declared nothing share their parent's scope
+        # object (TreeBuilder scope sharing), so identical arrays mean
+        # identical bindings — every namespace of this element is
+        # already declared above and will be skipped without a scan.
+        return true if parent_nodes.equal?(element.namespace_nodes)
+
+        parent_ns = parent_nodes.find do |parent_ns|
           parent_ns.prefix == ns.prefix && parent_ns.in_node_set?
         end
 
         parent_ns && parent_ns.uri == ns.uri
       end
+
+      # Sorted, in-node-set namespaces for one element. Undeclaring
+      # elements share one frozen array object, and node-set marking is
+      # fixed before processing runs, so the rendered form is cached
+      # per array identity for the lifetime of this handler (one
+      # canonicalization run).
+      def renderable_namespaces(element)
+        cache = (@renderable_namespaces ||= {}.compare_by_identity)
+        nodes = element.namespace_nodes
+        cache[nodes] ||= nodes.sort_by(&:local_name).select(&:in_node_set?)
+      end
     end
   end
 end

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

@@ -53,15 +53,26 @@ module Canon
           (@attribute_nodes ||= []) << attribute_node
         end
 
-        # Get namespace nodes in sorted order (lexicographically by local name)
+        # Get namespace nodes in sorted order (lexicographically by local name).
         def sorted_namespace_nodes
-          namespace_nodes.sort_by(&:local_name)
+          nodes = namespace_nodes
+          return nodes if nodes.empty?
+
+          nodes.sort_by(&:local_name)
         end
 
-        # Get attribute nodes in sorted order (by namespace URI then local name)
+        # Get attribute nodes in sorted order (by namespace URI then local
+        # name). A comparator block instead of sort_by keys allocates no
+        # per-attribute key arrays; (uri, name) pairs are unique per
+        # element (duplicates are resolved at build), so the order is
+        # identical.
         def sorted_attribute_nodes
-          attribute_nodes.sort_by do |attr|
-            [attr.namespace_uri.to_s, attr.local_name]
+          attrs = attribute_nodes
+          return attrs if attrs.empty?
+
+          attrs.sort do |a, b|
+            (a.namespace_uri.to_s <=> b.namespace_uri.to_s).nonzero? ||
+              (a.local_name <=> b.local_name)
           end
         end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.19
+  version: 0.3.21
 platform: ruby
 authors:
 - Ribose Inc.