Nokogiri <pre>, serialize() and whitespace failing spec

gistfile1.rb

require 'spec_helper'

def serialize(thing, format=:as_html)
  case thing
  when Nokogiri::HTML::Document
    document.serialize(:encoding => 'UTF-8') do |config|
      config.format.send format
    end
  when Nokogiri::HTML::DocumentFragment
    thing.serialize
  end
end

describe "Nokogiri significant whitespace preservation in <pre> and <code> blocks" do

  let(:html) do
      "
  <div>
    <pre>
      <code>
        <img src='about:blank'>
        <img src='about:blank'>
      </code>
    </pre>
  </div>
      "
  end

  let(:document) { Nokogiri::HTML html }

  let(:fragment) { Nokogiri::HTML::fragment html }

  context 'use a fragment (serialize with default args)' do
    it "should output the image tags on the same line" do
      serialize(fragment).lines.count{|l| l =~ /img/}.should == 2
    end
  end

  context 'use a document (serialize as XML)' do
    it "should output the image tags on the same line" do
      serialize(document, :as_xml).lines.count{|l| l =~ /img/}.should == 2
    end
  end

  context 'use a document (serialize as HTML)' do
    it "should output the image tags on the same line" do
      serialize(document, :as_html).lines.count{|l| l =~ /img/}.should == 2
    end
  end

end

gistfile2.txt

FFF

Failures:

  1) Nokogiri significant whitespace preservation in <pre> and <code> blocks use a document (serialize as XML) should output the image tags on the same line
     Failure/Error: serialize(fragment, :as_xml).lines.count{|l| l =~ /img/}.should == 2
       expected: 2
            got: 1 (using ==)
     # ./spec/ce/whitespace_spec.rb:41:in `block (3 levels) in <top (required)>'

  2) Nokogiri significant whitespace preservation in <pre> and <code> blocks use a document (serialize as HTML) should output the image tags on the same line
     Failure/Error: serialize(fragment, :as_html).lines.count{|l| l =~ /img/}.should == 2
       expected: 2
            got: 1 (using ==)
     # ./spec/ce/whitespace_spec.rb:47:in `block (3 levels) in <top (required)>'

  3) Nokogiri significant whitespace preservation in <pre> and <code> blocks use a fragment (serialize with default args) should output the image tags on the same line
     Failure/Error: serialize(fragment).lines.count{|l| l =~ /img/}.should == 2
       expected: 2
            got: 1 (using ==)
     # ./spec/ce/whitespace_spec.rb:35:in `block (3 levels) in <top (required)>'

Finished in 0.0541 seconds
3 examples, 3 failures

Failed examples:

rspec ./spec/ce/whitespace_spec.rb:40 # Nokogiri significant whitespace preservation in <pre> and <code> blocks use a document (serialize as XML) should output the image tags on the same line
rspec ./spec/ce/whitespace_spec.rb:46 # Nokogiri significant whitespace preservation in <pre> and <code> blocks use a document (serialize as HTML) should output the image tags on the same line
rspec ./spec/ce/whitespace_spec.rb:34 # Nokogiri significant whitespace preservation in <pre> and <code> blocks use a fragment (serialize with default args) should output the image tags on the same line

Maybe change the required libs so it says:

require 'rspec'
require 'nokogiri'

Needing Rails for this sucks :) I'll see about other people taking a look at this gist.

Let me see if I can organise my thoughts here.

The immediate problem we're having is Nokogiri is adding newlines to any tags, namely <span>, when outputting HTML. Which fucks with <pre> tags. Solving that, we'll be good, and I believe everything else will be ok too since I distinctly recall the content extractor doing a good job bookmarking code from sites like css-tricks.com.

So am I correct in assuming that finding how not to pretty-print when outputting stuff with Nokogiri would solve this?

No. The problem is tags with nothing but whitespace between them get put on one line.

Well, maybe it's a mixture of pretty-printing and stripping new lines. Not sure. I will continue on this tonight.