wp-xml-import.rb

require 'fileutils'
require 'date'
require 'yaml'
require 'rexml/document'
include REXML


doc = Document.new File.new(ARGV[0])

FileUtils.mkdir_p "_posts"

doc.elements.each("rss/channel/item[wp:status = 'publish' and wp:post_type = 'post']") do |e|
    post = e.elements
    slug = post['wp:post_name'].text
    date = DateTime.parse(post['wp:post_date'].text)
    name = "%02d-%02d-%02d-%s.markdown" % [date.year, date.month, date.day, slug]
    date_string = date.year.to_s + "-" + date.month.to_s + '-' + date.day.to_s
    
    # clean up years of category randomness in one fell swoop. 
    case post['category'].text
    when "diversions", 
         "elsewhere", 
         "family", 
         "health", 
         "life", 
         "links", 
         "meme", 
         "nerdliness", 
         "photography", 
         "random", 
         "relationships", 
         "social issues"
      category_string = post['category'].text
    else
      category_string = "life"
    end

    content = post['content:encoded'].text

    content = content.gsub(/<code>(.*?)<\/code>/, '`\1`')

    content = content.gsub(/<pre lang="([^"]*)">(.*?)<\/pre>/m, '')
    
    (1..3).each do |i|
        content = content.gsub(/<h#{i}>([^<]*)<\/h#{i}>/, ('#'*i) + ' \1')
    end

    puts "Converting: #{name}"

    data = {
        'layout' => 'post',
        'title' => post['title'].text,
        'date' => date_string,
        'comments' => 'false',
        'categories' => category_string,
        'author' => 'Mark'
     }.delete_if { |k,v| v.nil? || v == ''}.to_yaml

     File.open("_posts/#{name}", "w") do |f|
         f.puts data
         f.puts "---"
         f.puts content
     end
 
end

Modified gist to suit my needs. Added capture of WordPress category, and cleaned up categories into 12 tidy little piles, with "life" being the default category. Also capturing original posting date (not time) for posterity.