From: "Nigel S.Ball" <nigelb@...>
Date: 2001-10-31T23:24:49+09:00
Subject: [ruby-talk:23950] CGI::pretty hack

The standard cgi.rb includes this method to make "pretty" html. As elegant as 
it is there is an unfortunate side effect - <pre> and <textarea> elements are 
also indented, not what is wanted.

  def CGI::pretty(string, shift = "  ")
    lines = string.gsub(/(?!\A)<(?:.|\n)*?>/n, 
"\n\\0").gsub(/<(?:.|\n)*?>(?!\n)/n, "\\0\n")
    end_pos = 0
    while end_pos = lines.index(/^<\/(\w+)/n, end_pos)
      element = $1.dup
      start_pos = lines.rindex(/^\s*<#{element}/ni, end_pos)
      lines[start_pos ... end_pos] = "__" + lines[start_pos ... 
end_pos].gsub(/\n(?!\z)/n, "\n" + shift) + "__"
    end
    lines.gsub(/^((?:#{Regexp::quote(shift)})*)__(?=<\/?\w)/n, '\1')
  end

I have made the following altered hack which overcomes the problem but am 
sure that there would be a better, more elegant, way of achieving the 
required result

  def CGI::pretty(string, shift = "  ")
    i = 0; preserve = {}
    lines = string.gsub(/\n*<pre.*<\/pre>\n*/inm) {
      t = "%PRE#{i}"; i += 1
      preserve[t] = $&
      t
    }
    lines = lines.gsub(/\n*<textarea.*<\/textarea>\n*/inm) {
      t = "%TEXTAREA#{i}"; i += 1
      preserve[t] = $&
      t
    }
    lines = lines.gsub(/(?!\A)<(?:.|\n)*?>/n, 
"\n\\0").gsub(/<(?:.|\n)*?>(?!\n)/n, "\\0\n")
    end_pos = 0
    while end_pos = lines.index(/^<\/(\w+)/n, end_pos)
      element = $1.dup
      start_pos = lines.rindex(/^\s*<#{element}/ni, end_pos)
      lines[start_pos ... end_pos] = "__" + lines[start_pos ... 
end_pos].gsub(/\n(?!\z)/n, "\n" + shift) + "__"
    end
    lines = lines.gsub(/^((?:#{Regexp::quote(shift)})*)__(?=<\/?\w)/n, '\1')
    preserve.each { |t, s|
      lines.sub!(/#{Regexp::quote(t)}/, s)
    }
    lines
  end
-- 
Nigel