Cleaning converted docx -> html file

sanitize.rb

require 'rubygems'
require 'sanitize'
require 'nokogiri-styles'

no_whitespaces = lambda {|env|
  node = env[:node]
  return unless node.elem?

  unless node.children.any?{|c| !c.text? || c.content.strip.length > 0 }
    node.unlink
  end
}

first_table_row = lambda {|env|
  node = env[:node]
  return unless node.elem?
  if node.name == 'table'
    tr = node.children.first
    tr.children.each do |td|
      td.name = 'th'
    end
    tr['class'] = 'gray'
  end
}

color_cell = lambda {|env|
  node = env[:node]
  return if !node.elem? or !['td', 'tr'].member? node.name
  node['class'] = 'blue' if node.styles['background'] == '#99CCFF'
  node['class'] = 'gray' if %w(#737373 gray).member? node.styles['background']
  node.delete 'style'
  return {:node_whitelist => [node]}
}

html = File.read('Contractor Monitoring Form.htm')

result = Sanitize.clean(
    html,
    attributes:{
      all: [:class, :colspan]
    },
    elements: ['html', 'body', 'table', 'tbody', 'tr', 'td', 'div', 'style', 'title'],
    transformers: [no_whitespaces, color_cell]
  )

File.open('converted.html', 'w') { |file| file.write result }