moxml 0.5.1 → 0.5.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ca6fefbf114afcaf13d9c241b4df5081eef1e004019618ca12d31a655d89dd53
-  data.tar.gz: d361373f16dafff997d4ca53b5dbc23c332b09331e8f49ef5abed09f526bdea7
+  metadata.gz: 566f12c7bcda29e079d0fd5b496a6becff643b9c15afd328d5f99fb7ef70d35c
+  data.tar.gz: a4d7858218db7d38300b963b11d5e40b5aec176d2072a177c4e5448f839b8671
 SHA512:
-  metadata.gz: 0a2e61a613511970e4c6945c074a578aefcb12591def797a726bf0959796c43dbf2e71b77509271b1dad15b9f0f1e81f5ec3b8b0af0e384bccad35745bc2dbb6
-  data.tar.gz: 183964ef083c888b3b7d58b896a547acf25c0414d997040c10b805b8379088825c2ec996b149aa9e8fe51b6419135d1341d8ebe9c88ddb0656101e25d3291557
+  metadata.gz: 45cf69a5c753718d3bcef44580ef2e8e0dcfa536033537602952e8cbeb73e8bb59e68ad365fdc72c6b7422a998c50e623a1f7b894ed5e7da3cd8e129ff40c703
+  data.tar.gz: 0fd18d4891bd29ce56ab1174aa494f8dddc251d298067a0290bf84df41f6854d61bdeedf6e9fd813c32669a9a550419865a637633ddd6e558e84ff0d9f6eca2e

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/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Moxml
-  VERSION = "0.5.1"
+  VERSION = "0.5.2"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: moxml
 version: !ruby/object:Gem::Version
-  version: 0.5.1
+  version: 0.5.2
 platform: ruby
 authors:
 - Ribose Inc.