moxml 0.5.1 → 0.5.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ca6fefbf114afcaf13d9c241b4df5081eef1e004019618ca12d31a655d89dd53
-  data.tar.gz: d361373f16dafff997d4ca53b5dbc23c332b09331e8f49ef5abed09f526bdea7
+  metadata.gz: 6cadb7739071fa9a823f6a08fe5daa33c2313d22e4c4454aa491e975f62da5ee
+  data.tar.gz: b6682343586e2bf873c9067fccc68f03f2c67c8b10eafc418fcad6663320f2f2
 SHA512:
-  metadata.gz: 0a2e61a613511970e4c6945c074a578aefcb12591def797a726bf0959796c43dbf2e71b77509271b1dad15b9f0f1e81f5ec3b8b0af0e384bccad35745bc2dbb6
-  data.tar.gz: 183964ef083c888b3b7d58b896a547acf25c0414d997040c10b805b8379088825c2ec996b149aa9e8fe51b6419135d1341d8ebe9c88ddb0656101e25d3291557
+  metadata.gz: 67da3161b095e47aafe6d7e7327c2125a4458a9ef557c41c5f9d6465fccb01374d29c409178910b39426cf0cf7a42140d6858d4177c3f63b529690f397303796
+  data.tar.gz: c88f55e02c026e18f4c1a25578d9e27094835288e6ccc0dccff40633ff31029975125d9fcfb51ae5353f00ad5fa54fd8bef135d9e44c64da794b45513f728903

data/lib/moxml/adapter/leptris.rb

@@ -546,6 +546,10 @@ module Moxml
         # returns those without name accessors). Both engines resolve
         # expression prefixes against document-declared namespaces.
         NATIVE_XPATH_CACHE = XPath::Cache.new(100)
+        # The gate decision is expression-intrinsic; re-walking the
+        # cached AST three times per xpath call cost ~23% of repeated
+        # selective queries, so the boolean caches beside it.
+        NATIVE_GATE_CACHE = XPath::Cache.new(100)
 
         def xpath(node, expression, namespaces = {})
           native = native_xpath(node, expression, namespaces)
@@ -623,11 +627,13 @@ module Moxml
         # attribute-node results, or the nokogiri-compat xmlns:
         # reserved prefix convention.
         def native_expression?(expression)
-          ast = XPath::Parser.parse_with_cache(expression)
-          return false if ast_contains_type?(ast, :variable)
-          return false if uses_xmlns_prefix?(ast)
+          NATIVE_GATE_CACHE.get_or_set(expression) do
+            ast = XPath::Parser.parse_with_cache(expression)
+            next false if ast_contains_type?(ast, :variable)
+            next false if uses_xmlns_prefix?(ast)
 
-          !selects_attribute_results?(ast)
+            !selects_attribute_results?(ast)
+          end
         rescue XPath::SyntaxError
           false
         end

data/lib/moxml/adapter/libxml.rb

@@ -7,6 +7,11 @@ require "libxml"
 module Moxml
   module Adapter
     class Libxml < Base
+      # DOCTYPE prolog sniffer; see parse for the index-then-match
+      # strategy that keeps this off the hot path.
+      DOCTYPE_RE = /<!DOCTYPE\s+([^\s>]+)(?:\s+PUBLIC\s+"([^"]+)"\s+"([^"]+)"|\s+SYSTEM\s+"([^"]+)")?\s*>/i
+      private_constant :DOCTYPE_RE
+
       autoload :EntityRefRegistry, "moxml/adapter/libxml/entity_ref_registry"
 
       # Wrapper class to store DOCTYPE information
@@ -89,8 +94,15 @@ module Moxml
           # parameter or XML declaration for byte interpretation.
           xml_string = preprocess_entities(xml_string)
 
-          # Extract DOCTYPE before parsing
-          doctype_match = xml_string.match(/<!DOCTYPE\s+([^\s>]+)(?:\s+PUBLIC\s+"([^"]+)"\s+"([^"]+)"|\s+SYSTEM\s+"([^"]+)")?\s*>/i)
+          # Extract DOCTYPE before parsing. Locate the literal first:
+          # a plain memscan is ~2x cheaper than the regex's literal
+          # prefix scan on doctype-less documents (the common case),
+          # and matching from the found offset keeps captures
+          # identical.
+          doctype_idx = xml_string.index("<!DOCTYPE")
+          doctype_match = if doctype_idx
+                            xml_string.match(DOCTYPE_RE, doctype_idx)
+                          end
 
           native_doc = begin
             # Handle both string and file inputs

data/lib/moxml/attribute_resolver.rb

@@ -39,11 +39,15 @@ module Moxml
         return nil if uri.nil?
 
         element.attributes.find do |attr|
-          local_name(attr.name) == local && attribute_uri(element, attr) == uri
+          # attr.name allocates on several adapters (native string,
+          # plus leptris' force_encoding dup) — read it once per probe
+          attr_name = attr.name
+          local_name(attr_name) == local && attribute_uri(element, attr, attr_name) == uri
         end
       else
         element.attributes.find do |attr|
-          local_name(attr.name) == name && attribute_uri(element, attr).nil?
+          attr_name = attr.name
+          local_name(attr_name) == name && attribute_uri(element, attr, attr_name).nil?
         end
       end
     end
@@ -170,11 +174,11 @@ module Moxml
     # adapters that report a prefix without resolving it (rexml/ox for
     # the prebound xml prefix) and adapters that expose neither
     # namespace nor separate prefix (leptris qualified names).
-    def attribute_uri(element, attr)
+    def attribute_uri(element, attr, attr_name = nil)
       ns = attr.namespace
       return ns.uri unless ns.nil? || ns.uri.to_s.empty?
 
-      prefix = ns&.prefix || prefix_part(attr.name)
+      prefix = ns&.prefix || prefix_part(attr_name || attr.name)
       return nil if prefix.nil? || prefix.empty?
       return XML_NAMESPACE_URI if prefix == "xml"
 

data/lib/moxml/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Moxml
-  VERSION = "0.5.1"
+  VERSION = "0.5.3"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: moxml
 version: !ruby/object:Gem::Version
-  version: 0.5.1
+  version: 0.5.3
 platform: ruby
 authors:
 - Ribose Inc.