Convert posts from httrack of blogger/blogspot to jekyll format. Run from top folder.

blogger-to-jekyll.rb

#!/usr/bin/ruby
# convert posts from httrack of blogger/blogspot to jekyll format
# run from top folder.

require 'nokogiri'
require 'reverse_markdown'
require 'fileutils'
require 'date'

OUT_DIR = '_posts'
AUTHOR = nil # set default author-key or nil to use from post

# ---------------------------------

posts = Dir['20??/*/*']
FileUtils.mkdir_p "#{OUT_DIR}"

posts.each do |filename|
  p = Nokogiri::HTML(open(filename))
  date = p.css('h2.date-header')[0].text # Sunday, 18 August 2013
  date = DateTime.strptime(date, "%A, %d %B %Y").strftime('%Y-%m-%d')
  body = p.css('div.post')
  outfile = "#{OUT_DIR}/#{date}-#{File.basename(filename, '.html')}.md"
  
  date = "#{date} #{p.css('a.timestamp-link')[0].text} #{DateTime.now.strftime('%z')}"
  title = p.css('.post-title')[0].text.strip
  labels = p.css('span.post-labels')[0]
  if labels
    labels = '[' + labels.text.gsub(/\n?Labels:/, '').gsub(/\n/, '').gsub(',', ', ') + ']'
  else
    labels = '~'
  end
  author = AUTHOR ? AUTHOR : p.css('div.post-footer span.fn').text.strip
  puts "writing to #{outfile}"
  open(outfile, 'w') do |page|
  # frontmatter 
    page.puts "---"
    page.puts "layout:   post"
    page.puts "title:    \"#{title}\""
    page.puts "date:     #{date}"
    page.puts "comments: true"
    page.puts "categories: ~"
    page.puts "tags: #{labels}"
    page.puts "authors: [#{author}]"
    page.puts "---"
    page.puts ReverseMarkdown.convert(body.to_s)
  end
end