Protect included code snippets from "recursive" highlighting

[woopsn]

Description

This PR aims to fix an issue with how yard handles syntax highlighted snippets from {include:file:...} source.

When an included file contains code blocks, they'll be fed through #parse_codeblocks more than once, as the file and then the parent are processed depth first. After the first highlighting pass, the code block contains additional <span>s and escapes (< etc), and later passes are unable to tell that these weren't part of the original source.

yard's built-in ruby highlighter does work in these cases -- its tokenizer tends to raise an exception when the source looks like HTML instead of Ruby, and by inspecting the source it can tell that the block was already highlighted:

# lib/yard/templates/helpers/html_syntax_highlight_helper.rb#L41-L42
def html_syntax_highlight_ruby_ripper(source)
  resolver = Parser::Ruby::TokenResolver.new(source, object)
  ...
rescue Parser::ParserSyntaxError
  source =~ /^<span\s+class=/ ? source : h(source)
end

Plugins, on the other hand, are unaware that the code snippets they're given in some cases are already marked up. This leads to the final document showing a mess of escaped html within the <pre>.[1]

In this branch, data-highlighted=true is added to code snippets when they're processed through {include:...}, allowing them to be skipped during subsequent passes through #parse_codeblocks. The attribute is then removed by the outermost #htmlify call, so it's only an internal annotation and shouldn't make it into the final output (unless a theme calls #insert_include directly).

[1] This can be seen in the output of

$ cat setup.rb
module YARD::Templates::Helpers::HtmlHelper
  def html_syntax_highlight_html(source)
    %(<strong>#{CGI.escapeHTML(source)}</strong>)
  end
end

$ cat example.rdoc
  !!!html
  <h1>

$ cat README.rdoc
{include:file:example.rdoc}

$ yard doc -e setup.rb

Completed Tasks

• I have read the Contributing Guide.
• The pull request is complete (implemented / written).
• Git commits have been cleaned up (squash WIP / revert commits).
• I wrote tests and ran bundle exec rake locally (if code is attached to PR).

[lsegal]

Thanks for contributing! I added a few inline comments; in short, I'm all for this fix but there are a few blockers in the current approach in the form of a breaking change and regression.

Those would need to be resolved before getting this through.

[lsegal]

nit pick here but this should be styled as internal = false with spaces.

[lsegal]

Unfortunately using keyword args limits us to Ruby 3.x and would be a breaking change that could not be merged.

The good news is this could easily be changed to a positional arg.

[lsegal]

This is an odd syntax. Did you simply mean pre_attrs = %(class="code #{language}")?

[lsegal]

This is a subtle but problematic change that I think will cause a regression. This change was explicitly added to permit pre-formatted HTML extra-files to be added via: yard - OTHER.html where OTHER.html might include:

<pre class="code ruby"><code class="ruby"><span class='kw'>return</span></code></pre>

The output of this section would generate double encoded output if you ran it through this block:

If removing this HTML detection causes a failure in your implementation, you may have to rethink part of your approach.

[woopsn]

Thanks for taking a look! I fixed the syntax and style issues you mentioned, and restored the bypass behavior for pre-formatted blocks. I'd misinterpreted what that check was for and thought it was to account for the same recursive/nested case that this branch considers.

I brought the check back, but lifted it out of the rescue block and into and into HtmlHelper#html_syntax_highlight as a deliberate special case, and added this to the spec:

yard/lib/yard/templates/helpers/html_helper.rb

Lines 206 to 208 in a9babdb

	if type.to_s == "ruby" && source =~ /<span\s+class=/
	return source
	end

yard/spec/templates/helpers/html_helper_spec.rb

Lines 675 to 679 in a9babdb

	it "wraps but doesn't alter ruby snippets that already contain highlighting tags" do
	expect(subject.htmlify('<pre lang="ruby"><span class="kw">for</span></pre>', :html)).to eq(
	'<pre class="code ruby"><code class="ruby"><span class="kw">for</span></code></pre>'
	)
	end

Does this look alright? It seems slightly better to not feed non-ruby input to the parser and catch the exception, but the rest of my changes should be compatible either way.