Merge pull request #2221 from sparklemotion/1940-parse-xslt-styleshee…

…t-using-preferred-options

parse XSLT stylesheets using libxslt-preferred options

---

**What problem is this PR intended to solve?**

See #1940 where some XSL transformations ended up producing slightly different results from `xsltproc`

**Have you included adequate test coverage?**

Yes!

**Does this change affect the behavior of either the C or the Java implementations?**

This behavior appears to be libxslt-specific, and so the test only runs for CRuby; though the document ends up being parsed with the new options on both implementations.

CHANGELOG.md

@@ -4,6 +4,14 @@ Nokogiri follows [Semantic Versioning](https://semver.org/), please see the [REA
 
 ---
 
+## next / unreleased
+
+### Changed
+
+* Introduce `Nokogiri::XML::ParseOptions::DEFAULT_XSLT` which adds the libxslt-preferred options of `NOENT | DTDLOAD | DTDATTR | NOCDATA` to `ParseOptions::DEFAULT_XML`.
+* `Nokogiri.XSLT` parses the stylesheet using `ParseOptions::DEFAULT_XSLT`, which should make some edge-case XSL transformations match libxslt's default behavior. [[#1940](https://github.com/sparklemotion/nokogiri/issues/1940)]
+
+
 ## 1.11.3 / 2021-04-07
 
 ### Fixed

lib/nokogiri/xml/parse_options.rb

@@ -71,6 +71,8 @@ class ParseOptions
 
       # the default options used for parsing XML documents
       DEFAULT_XML  = RECOVER | NONET
+      # the default options used for parsing XSLT stylesheets
+      DEFAULT_XSLT = RECOVER | NONET | NOENT | DTDLOAD | DTDATTR | NOCDATA
       # the default options used for parsing HTML documents
       DEFAULT_HTML = RECOVER | NOERROR | NOWARNING | NONET
       # the default options used for parsing XML schemas

lib/nokogiri/xslt.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
-require 'nokogiri/xslt/stylesheet'
+require "nokogiri/xslt/stylesheet"
 
 module Nokogiri
   class << self
@@ -22,32 +22,32 @@ module XSLT
     class << self
       ###
       # Parse the stylesheet in +string+, register any +modules+
-      def parse string, modules = {}
+      def parse(string, modules = {})
         modules.each do |url, klass|
-          XSLT.register url, klass
+          XSLT.register(url, klass)
         end
 
+        doc = XML::Document.parse(string, nil, nil, XML::ParseOptions::DEFAULT_XSLT)
         if Nokogiri.jruby?
-          Stylesheet.parse_stylesheet_doc(XML.parse(string), string)
+          Stylesheet.parse_stylesheet_doc(doc, string)
         else
-          Stylesheet.parse_stylesheet_doc(XML.parse(string))
+          Stylesheet.parse_stylesheet_doc(doc)
         end
       end
 
       ###
       # Quote parameters in +params+ for stylesheet safety
-      def quote_params params
+      def quote_params(params)
         parray = (params.instance_of?(Hash) ? params.to_a.flatten : params).dup
-        parray.each_with_index do |v,i|
-          if i % 2 > 0
-            parray[i]=
-              if v =~ /'/
-                "concat('#{ v.gsub(/'/, %q{', "'", '}) }')"
-              else
-                "'#{v}'";
-              end
+        parray.each_with_index do |v, i|
+          parray[i] = if i % 2 > 0
+            if v =~ /'/
+              "concat('#{v.gsub(/'/, %q{', "'", '})}')"
+            else
+              "'#{v}'"
+            end
           else
-            parray[i] = v.to_s
+            v.to_s
           end
         end
         parray.flatten

lib/nokogiri/xslt/stylesheet.rb

@@ -18,7 +18,7 @@ class Stylesheet
       # Apply an XSLT stylesheet to an XML::Document.
       # +params+ is an array of strings used as XSLT parameters.
       # returns serialized document
-      def apply_to document, params = []
+      def apply_to(document, params = [])
         serialize(transform(document, params))
       end
     end