Support HTML5 parsing using Nokogiri >= 1.12.0

.travis.yml

@@ -33,8 +33,9 @@ cache:
 
 matrix:
   include:
-    - gemfile: gemfiles/Gemfile.gumbo
+    - gemfile: Gemfile
       rvm: 3.0
+      env: HTML5_PARSING=true
       script: bundle exec rake rack_smoke
     - gemfile: Gemfile
       rvm: 2.7

History.md

@@ -8,6 +8,7 @@ Release date: unreleased
 ### Added
 
 * `allow_label_click` accepts click options to be used when clicking an associated label
+* Deprecated `allow_gumbo=` in favor of `use_html5_parsing=` to enable use of Nokogiri::HTL5 when available
 
 ### Fixed
 

gemfiles/Gemfile.gumbo

@@ -1,7 +0,0 @@
-source 'https://rubygems.org'
-
-gem 'bundler', '< 3.0'
-gemspec path: '..'
-
-gem 'xpath', github: 'teamcapybara/xpath'
-gem 'nokogumbo'

lib/capybara.rb

@@ -39,7 +39,7 @@ module Capybara
     #   See {Capybara.configure}
     # @!method javascript_driver
     #   See {Capybara.configure}
-    # @!method allow_gumbo
+    # @!method use_html5_parsing
     #   See {Capybara.configure}
     Config::OPTIONS.each do |method|
       def_delegators :config, method, "#{method}="
@@ -69,7 +69,7 @@ module Capybara
     #
     # #### Configurable options
     #
-    # - **allow_gumbo** (Boolean = `false`) - When `nokogumbo` is available, whether it will be used to parse HTML strings.
+    # - **use_html5_parsing** (Boolean = `false`) - When Nokogiri >= 1.12.0 or `nokogumbo` is installed, whether HTML5 parsing will be used for HTML strings.
     # - **always_include_port** (Boolean = `false`) - Whether the Rack server's port should automatically be inserted into every visited URL
     #   unless another port is explicitly specified.
     # - **app_host** (String, `nil`) - The default host to use when giving a relative URL to visit, must be a valid URL e.g. `http://www.example.com`.
@@ -385,26 +385,21 @@ module Capybara
     # @return [Nokogiri::HTML::Document]      HTML document
     #
     def HTML(html) # rubocop:disable Naming/MethodName
-      if Nokogiri.respond_to?(:HTML5) && Capybara.allow_gumbo # Nokogumbo installed and allowed for use
-        Nokogiri::HTML5(html).tap do |document|
-          document.xpath('//template').each do |template|
-            # template elements content is not part of the document
-            template.inner_html = ''
-          end
-          document.xpath('//textarea').each do |textarea|
-            # The Nokogumbo HTML5 parser already returns spec compliant contents
-            textarea['_capybara_raw_value'] = textarea.content
-          end
-        end
+      # Nokogiri >= 1.12.0 or Nokogumbo installed and allowed for use
+      html_parser, using_html5 = if defined?(Nokogiri::HTML5) && Capybara.use_html5_parsing
+        [Nokogiri::HTML5, true]
       else
-        Nokogiri::HTML(html).tap do |document|
-          document.xpath('//template').each do |template|
-            # template elements content is not part of the document
-            template.inner_html = ''
-          end
-          document.xpath('//textarea').each do |textarea|
-            textarea['_capybara_raw_value'] = textarea.content.delete_prefix("\n")
-          end
+        [defined?(Nokogiri::HTML4) ? Nokogiri::HTML4 : Nokogiri::HTML, false]
+      end
+
+      html_parser.parse(html).tap do |document|
+        document.xpath('//template').each do |template|
+          # template elements content is not part of the document
+          template.inner_html = ''
+        end
+        document.xpath('//textarea').each do |textarea|
+          # The Nokogiri HTML5 parser already returns spec compliant contents
+          textarea['_capybara_raw_value'] = using_html5 ? textarea.content : textarea.content.delete_prefix("\n")
         end
       end
     end
@@ -516,6 +511,6 @@ Capybara.configure do |config|
   config.test_id = nil
   config.predicates_wait = true
   config.default_normalize_ws = false
-  config.allow_gumbo = false
+  config.use_html5_parsing = false
   config.w3c_click_offset = false
 end

lib/capybara/config.rb

@@ -7,9 +7,11 @@ module Capybara
   class Config
     extend Forwardable
 
-    OPTIONS = %i[app reuse_server threadsafe server default_driver javascript_driver allow_gumbo].freeze
+    OPTIONS = %i[
+      app reuse_server threadsafe server default_driver javascript_driver use_html5_parsing allow_gumbo
+    ].freeze
 
-    attr_accessor :app, :allow_gumbo
+    attr_accessor :app, :use_html5_parsing
     attr_reader :reuse_server, :threadsafe, :session_options # rubocop:disable Style/BisectedAttrAccessor
     attr_writer :default_driver, :javascript_driver
 
@@ -88,5 +90,15 @@ module Capybara
       end
       @deprecation_notified[method] = true
     end
+
+    def allow_gumbo=(val)
+      deprecate('allow_gumbo=', 'use_html5_parsing=')
+      self.use_html5_parsing = val
+    end
+
+    def allow_gumbo
+      deprecate('allow_gumbo', 'use_html5_parsing')
+      use_html5_parsing
+    end
   end
 end

lib/capybara/node/simple.rb

@@ -110,7 +110,7 @@ module Capybara
           # No need for an xpath if only checking the current element
           !(native.key?('hidden') ||
             /display:\s?none/.match?(native[:style] || '') ||
-            %w[script head].include?(tag_name))
+            %w[script head style].include?(tag_name))
         end
       end
 

lib/capybara/spec/spec_helper.rb

@@ -36,7 +36,7 @@ module Capybara
         Capybara.test_id = nil
         Capybara.predicates_wait = true
         Capybara.default_normalize_ws = false
-        Capybara.allow_gumbo = true
+        Capybara.use_html5_parsing = !ENV['HTML5_PARSING'].nil?
         Capybara.w3c_click_offset = false
         reset_threadsafe
       end

spec/basic_node_spec.rb

@@ -125,12 +125,13 @@ RSpec.describe Capybara do
 
     it 'drops illegal fragments when using gumbo' do
       skip 'libxml is less strict than Gumbo' unless Nokogiri.respond_to?(:HTML5)
+      described_class.use_html5_parsing = true
       expect(described_class.string('<td>1</td>')).not_to have_css('td')
     end
 
-    it 'can disable use of gumbo' do
-      skip "Test doesn't make sense unlesss nokogumbo is loaded" unless Nokogiri.respond_to?(:HTML5)
-      described_class.allow_gumbo = false
+    it 'can disable use of HTML5 parsing' do
+      skip "Test doesn't make sense unlesss HTML5 parsing is loaded (Nokogumbo or Nokogiri >= 1.12.0)" unless Nokogiri.respond_to?(:HTML5)
+      described_class.use_html5_parsing = false
       expect(described_class.string('<td>1</td>')).to have_css('td')
     end
 

spec/rack_test_spec.rb

@@ -1,12 +1,6 @@
 # frozen_string_literal: true
 
 require 'spec_helper'
-nokogumbo_required = begin
-  require 'nokogumbo'
-  true
-rescue LoadError
-  false
-end
 
 module TestSessions
   RackTest = Capybara::Session.new(:rack_test, TestApp)
@@ -256,11 +250,12 @@ RSpec.describe Capybara::RackTest::Driver do
 end
 
 RSpec.describe 'Capybara::String' do
-  it 'should use gumbo' do
-    skip 'Only  valid if gumbo is included' unless nokogumbo_required
-    allow(Nokogiri).to receive(:HTML5).and_call_original
+  it 'should use HTML5 parsing' do
+    skip 'Only valid if Nokogiri >= 1.12.0 or gumbo is included' unless defined? Nokogiri::HTML5
+    Capybara.use_html5_parsing = true
+    allow(Nokogiri::HTML5).to receive(:parse).and_call_original
     Capybara.string('<div id=test_div></div>')
-    expect(Nokogiri).to have_received(:HTML5)
+    expect(Nokogiri::HTML5).to have_received(:parse)
   end
 end