Uruguayan constitution markdown formatter

formatter.rb

# encoding: UTF-8

require 'open-uri'

years = %w{1997 1994 1989 1967 1952 1942 1934 1918 1830}
i = years.last

file = open("http://www0.parlamento.gub.uy/constituciones/const#{i.slice(1..-1)}.htm").read
t = file.encode('UTF-8', 'ISO-8859-1', universal_newline: true)

# Remove everything up to first h2 tag
t.gsub!(/.*?(?=<h2)/m, "")

# Remove everything after webbot comment
t.gsub!(/<!--webbot.*/im, "")

# Strip html comments
t.gsub!(/<!--(.*)-->\n/, '')

# Remove br tags
t.gsub!(/<br>\n?/, ' ')

# Set main header
t.gsub!(/<h2[^>]*>CONSTITUCION DE LA REPUBLICA<\/h2>/, '# CONSTITUCION DE LA REPUBLICA')

# Set explanatory text as 4th level header
t.gsub!(/<h4[^>]*>(CONSTITUCION .+)<\/h4>/, '#### \1')

# Remove line breaks in h4 tags
t.gsub!(/(<h4[^>]*>[^<]*)\n/, '\1 ')

# Set sections as second level headers
t.gsub!(/<h4[^>]*>(SECCION .*)<\/h4>/, '## \1')

# Set chapters as third level headers
t.gsub!(/<h4[^>]*>(CAPITULO .*)<\/h4>/, '### \1')

# Set rest of h4 as 4th level headers
t.gsub!(/<h4[^>]*>([^<]*)<\/h4>/m, '#### \1')

# Replace hr with ---
t.gsub!(/<hr[^>]*>/, '---')

# Remove p tags
t.gsub!('<p>', '').gsub!('</p>', '')

# Handle article names
t.gsub!(/<u><a[^>]*>(.*)<\/a><\/u>º?/, '**\1º**')

# Remove article links
t.gsub!(/<a[^>]*>(.*)<\/a>/, '\1')

# Format tables as numbered lists
t.gsub!(/<\/?table.*>\n/, '')
t.gsub!(/ *<tr>/, '')
t.gsub!(/ *<\/tr>/, '')
t.gsub!(/ *<td[^>]*>([0-9]+)(º|\.)<\/td>\n/, '\1.')
t.gsub!(/ *<td[^>]*>([^<]*)<\/td>/m, '  \1')

puts t