Use XPath to group siblings from an HTML/XML docum

I want to transform an HTML or XML document by grouping previously ungrouped sibling nodes.

For example, I want to take the following fragment:

<h2>Header</h2>
<p>First paragraph</p>
<p>Second paragraph</p>

<h2>Second header</h2>
<p>Third paragraph</p>
<p>Fourth paragraph</p>

Into this:

<section>
  <h2>Header</h2>
  <p>First paragraph</p>
  <p>Second paragraph</p>
</section>

<section>
  <h2>Second header</h2>
  <p>Third paragraph</p>
  <p>Fourth paragraph</p>
</section>

Is this possible using simple Xpath selectors and an XML parser like Nokogiri? Or do I need to implement a SAX parser for this task?

回答1:

Updated Answer

Here's a general solution that creates a hierarchy of <section> elements based on header levels and their following siblings:

class Nokogiri::XML::Node
  # Create a hierarchy on a document based on heading levels
  #   wrap   : e.g. "<section>" or "<div class='section'>"
  #   stops  : array of tag names that stop all sections; use nil for none
  #   levels : array of tag names that control nesting, in order
  def auto_section(wrap='<section>', stops=%w[hr], levels=%w[h1 h2 h3 h4 h5 h6])
    levels = Hash[ levels.zip(0...levels.length) ]
    stops  = stops && Hash[ stops.product([true]) ]
    stack = []
    children.each do |node|
      unless level = levels[node.name]
        level = stops && stops[node.name] && -1
      end
      stack.pop while (top=stack.last) && top[:level]>=level if level
      stack.last[:section].add_child(node) if stack.last
      if level && level >=0
        section = Nokogiri::XML.fragment(wrap).children[0]
        node.replace(section); section << node
        stack << { :section=>section, :level=>level }
      end
    end
  end
end

Here is this code in use, and the result it gives.

The original HTML

<body>
<h1>Main Section 1</h1>
<p>Intro</p>
<h2>Subhead 1.1</h2>
<p>Meat</p><p>MOAR MEAT</p>
<h2>Subhead 1.2</h2>
<p>Meat</p>
<h3>Caveats</h3>
<p>FYI</p>
<h4>ProTip</h4>
<p>Get it done</p>
<h2>Subhead 1.3</h2>
<p>Meat</p>

<h1>Main Section 2</h1>
<h3>Jumpin' in it!</h3>
<p>Level skip!</p>
<h2>Subhead 2.1</h2>
<p>Back up...</p>
<h4>Dive! Dive!</h4>
<p>...and down</p>

<hr /><p id="footer">Copyright &copy; All Done</p>
</body>

The conversion code

# Use XML only so that we can pretty-print the results; HTML works fine, too
doc = Nokogiri::XML(html,&:noblanks) # stripping whitespace allows indentation
doc.at('body').auto_section          # make the magic happen
puts doc.to_xhtml                    # show the result with indentation

The result

<body>
  <section>
    <h1>Main Section 1</h1>
    <p>Intro</p>
    <section>
      <h2>Subhead 1.1</h2>
      <p>Meat</p>
      <p>MOAR MEAT</p>
    </section>
    <section>
      <h2>Subhead 1.2</h2>
      <p>Meat</p>
      <section>
        <h3>Caveats</h3>
        <p>FYI</p>
        <section>
          <h4>ProTip</h4>
          <p>Get it done</p>
        </section>
      </section>
    </section>
    <section>
      <h2>Subhead 1.3</h2>
      <p>Meat</p>
    </section>
  </section>
  <section>
    <h1>Main Section 2</h1>
    <section>
      <h3>Jumpin' in it!</h3>
      <p>Level skip!</p>
    </section>
    <section>
      <h2>Subhead 2.1</h2>
      <p>Back up...</p>
      <section>
        <h4>Dive! Dive!</h4>
        <p>...and down</p>
      </section>
    </section>
  </section>
  <hr />
  <p id="footer">Copyright  All Done</p>
</body>

Original Answer

Here's an answer using no XPath, but Nokogiri. I've taken the liberty of making the solution somewhat flexible, handling arbitrary start/stops (but not nested sections).

html = "<h2>Header</h2>
<p>First paragraph</p>
<p>Second paragraph</p>

<h2>Second header</h2>
<p>Third paragraph</p>
<p>Fourth paragraph</p>

<hr>
<p id='footer'>All done!</p>"

require 'nokogiri'
class Nokogiri::XML::Node
  # Provide a block that returns:
  #  true  - for nodes that should start a new section
  #  false - for nodes that should not start a new section
  #  :stop - for nodes that should stop any current section but not start a new one
  def group_under(name="section")
    group = nil
    element_children.each do |child|
      case yield(child)
        when false, nil
          group << child if group
        when :stop
          group = nil 
        else
          group = document.create_element(name)
          child.replace(group)
          group << child
      end
    end
  end
end

doc = Nokogiri::HTML(html)
doc.at('body').group_under do |node|
  if node.name == 'hr'
    :stop
  else
    %w[h1 h2 h3 h4 h5 h6].include?(node.name)
  end
end

puts doc
#=> <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd">
#=> <html><body>
#=> <section><h2>Header</h2>
#=> <p>First paragraph</p>
#=> <p>Second paragraph</p></section>
#=> 
#=> <section><h2>Second header</h2>
#=> <p>Third paragraph</p>
#=> <p>Fourth paragraph</p></section>
#=> 
#=> <hr>
#=> <p id="footer">All done!</p>
#=> </body></html>

For XPath, see XPath : select all following siblings until another sibling

回答2:

One way using xpath is to select all the p elements that follow your h2 and from them subtract the p elements that also follow the next h2:

doc = Nokogiri::HTML.fragment(html)
doc.css('h2').each do |h2|
    nodeset = h2.xpath('./following-sibling::p')
    next_h2 = h2.at('./following-sibling::h2')
    nodeset -= next_h2.xpath('./following-sibling::p') if next_h2
    section_tag = h2.add_previous_sibling Nokogiri::XML::Node.new('section',doc)
    h2.parent = section_tag
    nodeset.each {|n| n.parent = section_tag}
end

回答3:

XPath can only select things from your input document, it can't transform it into a new document. For that you need XSLT or some other transformation language. I guess if you're into Nokogiri then the previous answers will be useful, but for completeness, here's what it looks like in XSLT 2.0:

<xsl:for-each-group select="*" group-starting-with="h2">
  <section>
    <xsl:copy-of select="current-group()"/>
  </section>
</xsl:for-each-group>