使用Nokogiri清理HTML

Question

I'm trying to clean up some CMS entered HTML that has extraneous paragraph tags and br tags everywhere. The Sanitize gem has proved very useful to do this but I am stuck with a particular issue.

The problem is when there is a br tag directly after/before a paragraph tag eg

<p>
  <br />
  Some text here
  <br />
  Some more text
  <br />
</p>

I would like to strip out the extraneous first and last br tags, but not the middle one.

I'm very much hoping I can use a sanitize transformer to do this but can't seem to find the right matcher to achieve this.

Any help would be much appreciated.

Answer 1

Here's how to locate the particular <br> nodes that are contained by <p> :

require 'nokogiri'

doc = Nokogiri::HTML::DocumentFragment.parse(<<EOT)
<p>
  <br />
  Some text here
  <br />
  Some more text
  <br />
</p>
EOT

doc.search('p > br').map(&:to_html)
# => ["<br>", "<br>", "<br>"]

Once we know we can find them, it's easy to remove specific ones:

br_nodes = doc.search('p > br')
br_nodes.first.remove
br_nodes.last.remove
doc.to_html
# => "<p>\n  \n  Some text here\n  <br>\n  Some more text\n  \n</p>\n"

Notice that Nokogiri removed them, but their associated Text nodes that are their immediate siblings, containing their "\\n" are left behind. A browser will gobble those up and not display the line-ends, but you might be feeling OCD, so here's how to remove those also:

br_nodes = doc.search('p > br')
[br_nodes.first, br_nodes.last].each do |br|
  br.next_sibling.remove
  br.remove
end
doc.to_html
# => "<p>\n  <br>\n  Some more text\n  </p>\n"

Answer 2

initial_linebreak_transformer = lambda {|options|
  node = options[:node]
  if node.present? && node.element? && node.name.downcase == 'p'
    first_child = node.children.first
    if first_child.name.downcase == 'br'
      first_child.unlink
      initial_linebreak_transformer.call options
    end
  end
}