Question

I’ve struggling to try to do this with a simple regex but it’s never been very accurate. It doesn’t have to be perfect.

Source has a combination of
and

tags. I don’t want to count blank lines.

Old way:

  self.words = rendered.gsub(/<p>&nbsp;<\/p>/,'').gsub(/<p><br\s?\/?>|(?:<br\s?\/?>){2,}/,'<br>').scan(/<br>|<br \/>|<p/).size+1

New way (not working:
Tries to turn all the
+ into paragraphs, then throw it into nokogiri to count paragraph tags with more than 3 chars in them (I have no idea how? Counting 1 letter lines would be nice too, but this worked ok in javascript)

  h = rendered
  h.gsub!(/<br>\s*<br>/gi,"<p>")
  h.gsub!(/<br>/gi,"<p>") if h =~ /<br>\s*<br>/
  h.prepend "<p>" if !h =~ /^\s*<p[^>]*>/i
  h.replace(/<p>\s*<p>/g,"<p>&nbsp;</p><p>")
  Nokogiri::HTML(rendered)
  # find+count p tags with at least 1-3 chars?

  # this is javascript not ruby, but you get the idea
  $('p', c).each(function(i) { // had to trim it to remove whitespaces from start/end.
    if ($(this).children('img').length) return; // skip if it's just an image.
    if ($.trim($(this).text()).length > 3)
      $(this).append("<div class='num'>"+ (n += 1) +"</div>");
  })

Other methods are welcome!

Example poem ( http://allpoetry.com/poem/7429983-the_many_endings-by-Kevin )

<p>
    from the other side of silence<br>
    you met me with change and a pocket<br>
    of unhappy apples.</p>
<p>
     </p>
<p>
    <br>
    we bled together to black<br>
    and chose the path carefully to<br>
    france.<br><br>
    sometimes when you smile<br>
    your radiant footsteps fall<br>
    and all around us is silence:<br>
    each dream step is<br>
    false but full of such glory</p>
<p>
     </p>
<p>
    <br>
    unhappiness never made a student of you:<br>
    just two by two by two.  now three<br>
    this great we that overflows our<br>
    heart-cave<br><br>
    each jewel-like addition to the delicate<br>
    crown.  but flowers fall and dreams,<br>
    all dreams, come to and end with death.</p> 

Thank you!

Answer 1

For posterity, here’s what I’m using now and it seems to be quite accurate. Non latin chars cause some problems sometimes from ckeditor, so I’m stripping them out for now.

  html = Nokogiri::HTML(rendered)
  text = html.at('body').inner_text rescue nil
  return self.words = rendered.gsub(/<p>&nbsp;<\/p>/,'').gsub(/<p><br\s?\/?>|(?:<br\s?\/?>){2,}/,'<br>').scan(/<br>|<br \/>|<p/).size+1 if !text

  #bonus points to strip lines entirely non-letter. idk

  #d "text is", text.gsub!(/([\x09|\x0D|\t])|(\xc2\xa0){1,}|[^A-z]/u,'')
  text.gsub!(/[^A-z\n]/u,'')
  #d "text is", text
  self.words = text.strip.scan(/(\s*\n\s*)+/).size+1