Strip text from HTML document using Ruby

Asked 30/9, 2009 at 11:6 Answered 30/9, 2009 at 14:55

There are lots of examples of how to strip HTML tags from a document using Ruby, Hpricot and Nokogiri have inner_text methods that remove all HTML for you easily and quickly.

What I am trying to do is the opposite, remove all the text from an HTML document, leaving just the tags and their attributes.

I considered looping through the document setting inner_html to nil but then really you'd have to do this in reverse as the first element (root) has an inner_html of the entire rest of the document, so ideally I'd have to start at the inner most element and set inner_html to nil whilst moving up through the ancestors.

Does anyone know a neat little trick for doing this efficiently? I was thinking perhaps regex's might do it but probably not as efficiently as an HTML tokenizer/parser might.

Ike answered 30/9, 2009 at 11:6 Comment(2)

Are you going to have to deal with bad markup? (unescaped entities, etc.) – Hob 30/9, 2009 at 11:57

It's possible - the markup I'm processing comes from end users so can't be relied on. – Ike 30/9, 2009 at 13:9

This works too:

doc = Nokogiri::HTML(your_html)
doc.xpath("//text()").remove

Bowers answered 30/9, 2009 at 14:29 Comment(1)

Excellent! Toss a .to_s on the end to get a string and you're good to go! – Amblygonite 5/3, 2012 at 0:28

You can scan the string to create an array of "tokens", and then only select those that are html tags:

>> some_html
=> "<div>foo bar</div><p>I like <em>this</em> stuff <a href='http://foo.bar'> long time</a></p>"
>> some_html.scan(/<\/?[^>]+>|[\w\|`~!@#\$%^&*\(\)\-_\+=\[\]{}:;'",\.\/?]+|\s+/).select { |t| t =~ /<\/?[^>]+>/ }.join("")
=> "<div></div><p><em></em><a href='http://foo.bar'></a></p>"

==Edit==

Or even better, just scan for html tags ;)

>> some_html.scan(/<\/?[^>]+>/).join("")
=> "<div></div><p><em></em><a href='http://foo.bar'></a></p>"

Lajoie answered 30/9, 2009 at 12:13 Comment(0)

To grab everything not in a tag, you can use nokogiri like this:

doc.search('//text()').text

Of course, that will grab stuff like the contents of <script> or <style> tags, so you could also remove blacklisted tags:

blacklist = ['title', 'script', 'style']
nodelist = doc.search('//text()')
blacklist.each do |tag|
  nodelist -= doc.search('//' + tag + '/text()')
end
nodelist.text

You could also whitelist if you preferred, but that's probably going to be more time-intensive:

whitelist = ['p', 'span', 'strong', 'i', 'b']  #The list goes on and on...
nodelist = Nokogiri::XML::NodeSet.new(doc)
whitelist.each do |tag|
  nodelist += doc.search('//' + tag + '/text()')
end
nodelist.text

You could also just build a huge XPath expression and do one search. I honestly don't know which way is faster, or if there is even an appreciable difference.

Tabard answered 30/9, 2009 at 14:3 Comment(0)

I just came up with this, but @andre-r's solution is soo much better!

#!/usr/bin/env ruby

require 'nokogiri'

def strip_text doc
  Nokogiri(doc).tap { |doc|
    doc.traverse do |node|
      node.content = nil if node.text?
    end
  }.to_s
end

require 'test/unit'
require 'yaml'
class TestHTMLStripping < Test::Unit::TestCase
  def test_that_all_text_gets_strippped_from_the_document
    dirty, clean = YAML.load DATA
    assert_equal clean, strip_text(dirty)
  end
end
__END__
---
- |
  <!DOCTYPE html>
  <html xmlns='http://www.w3.org/1999/xhtml' xml:lang='en' lang='en'>
    <head>
        <meta http-equiv='Content-type'     content='text/html; charset=UTF-8' />
        <title>Test HTML Document</title>
        <meta http-equiv='content-language' content='en' />
    </head>
    <body>
        <h1>Test <abbr title='Hypertext Markup Language'>HTML</abbr> Document</h1>
        <div class='main'>
            <p>
                <strong>Test</strong> <abbr title='Hypertext Markup Language'>HTML</abbr> <em>Document</em>
            </p>
        </div>
    </body>
  </html>
- |
  <!DOCTYPE html>
  <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
  <head>
  <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  <title></title>
  <meta http-equiv="content-language" content="en">
  </head>
  <body><h1><abbr title="Hypertext Markup Language"></abbr></h1><div class="main"><p><strong></strong><abbr title="Hypertext Markup Language"></abbr><em></em></p></div></body>
  </html>

Mozza answered 30/9, 2009 at 14:55 Comment(0)

Recommended topics

Hot tags