Auto-linkify markdown text

[pablobm]

In #6518 (and further in #6597) we implemented "linkification" in plain text: strings like node/123456 will automatically turn into links to features/etc in user comments.

Since this was only implemented for plain text, it only works in notes, note comments, changeset descriptions, and changeset comments, if I'm not missing anything. Text written in Markdown (diary entries, diary comments, direct messages, user profiles, etc) is not linkified.

So let's make this work in Markdown!

The problem is that we have to be careful with which specific text to linkify. For example we don't want to linkify inside code blocks (text rendered within <code> and <pre> elements), links (rendered with <a>)... and I'm not sure if others?

[pablobm]

Idea: perhaps we can use a stream HTML parser, avoiding the linkification when we are inside an inappropriate element. Something like this:

UNLINKABLE_ELEMENTS = ["pre", "code", "a"]

unlinkable_elements_stack = []

parser = StreamingParser.new(html) do |events|
  events.on_opening_element do |element|
    unlinkable_elements_stag.push(element.tag_name) if UNLINKABLE_ELEMENTS.include?(element.tag_name)
  end
  
  events.on_closing_element do |element|
    element.text = linkify(text) if inappropriate_elements_stack.empty?
    
    unlinkable_elements_stag.pop if UNLINKABLE_ELEMENTS.include?(element.tag_name)
  end
end

parser.run

I'm totally making up the API above. Nokogiri has something like it, but I don't remember the detail now. LibXML probably has it too.

[1ec5]

For example we don't want to linkify inside code blocks (text rendered within <code> and <pre> elements), links (rendered with <a>)... and I'm not sure if others?

Off the top of my head, we also support <samp> which is similar. If we don’t already, we probably should have a list of elements that we don’t want to do transforms inside of.