An excerpt from the Rakefile I use to deploy my blog. http://blog.alexbrowne.info

Rakefile

# ...

desc "Deploy website to s3/cloudfront via aws-sdk"
task :s3_cloudfront => [:generate, :minify, :gzip, :compress_images] do
  puts "=================================================="
  puts "      Deploying to Amazon S3 & CloudFront"
  puts "=================================================="

  # setup the aws_deploy_tools object
  config = YAML::load( File.open("_config.yml"))
  aws_deploy = AWSDeployTools.new(config)

  # get all files in the public directory
  all_files = Dir.glob("#{$public_dir}/**/*.*")

  # we want the gzipped version of the files, not the regular (non-gzipped) version
  # excluded files contains all the regular versions, which will not be deployed
  excluded_files = []
  $gzip_exts.collect do |ext|
    excluded_files += Dir.glob("#{$public_dir}/**/*.#{ext}")
  end

  # we do gzipped files seperately since they have different metadata (:content_encoding => gzip)
  puts "--> syncing gzipped files...".yellow
  gzipped_files = Dir.glob("#{$public_dir}/**/*.gz")
  gzipped_keys = gzipped_files.collect {|f| (f.split("#{$public_dir}/")[1]).sub(".gz", "")}

  aws_deploy.sync(gzipped_keys, gzipped_files,
          :reduced_redundancy => true,
          :cache_control => "max_age=86400", #24 hours
          :content_encoding => 'gzip',
          :acl => config['acl']
          )

  puts "--> syncing all other files...".yellow
  non_gzipped_files = all_files - gzipped_files - excluded_files
  non_gzipped_keys = non_gzipped_files.collect {|f| f.split("#{$public_dir}/")[1]}

  aws_deploy.sync(non_gzipped_keys, non_gzipped_files,
          :reduced_redundancy => true,
          :cache_control => "max_age=86400", #24 hours
          :acl => config['acl']
          )
  
  # invalidate all the files we just pushed
  aws_deploy.invalidate_dirty_keys  

  puts "DONE."

end

desc "Compress all applicable content in public/ using gzip"
task :gzip do

  unless which('gzip')
    puts "WARNING: gzip is not installed on your system. Skipping gzip..."
    return
  end

  @compressor ||= RedDragonfly.new

  $gzip_exts.each do |ext|
    puts "--> gzipping all #{ext}...".yellow
    files = Dir.glob("#{$gzip_dir}/**/*.#{ext}")
    files.each do |f|
      @compressor.gzip(f)
    end
  end

  puts "DONE."
end

desc "Minify all applicable files in public/ using jitify"
task :minify do
  
  unless which('jitify')
    puts "WARNING: jitify is not installed on your system. Skipping minification..."
    return
  end

  @compressor ||= RedDragonfly.new

  $minify_exts.each do |ext|
    puts "--> minifying all #{ext}...".yellow
    files = Dir.glob("#{$minify_dir}/**/*.#{ext}")
    files.each do |f|
      @compressor.minify(f)
    end
  end

  puts "DONE."

end

desc "Compress all images in public/ using ImageMagick"
task :compress_images do

  unless which('convert')
    puts "WARNING: ImageMagick is not installed on your system. Skipping image compression..."
    return
  end

  @compressor ||= RedDragonfly.new

  $compress_img_exts.each do |ext|
    puts "--> compressing all #{ext}...".yellow
    files = Dir.glob("#{$compress_img_dir}/**/*.#{ext}")
    files.each do |f|
      @compressor.compress_img(f)
    end
  end

  puts "DONE."

end

# ...

Thanks for this, but I wanted to point out an issue I came across:

If you don't specify a :content_type option in aws_deploy.sync, then you could end up with a Content-Type of "image/jpeg" for html files. This means that the sync method needs to have logic that detects what kind of file is being uploaded and add the appropriate Content-Type.

Good catch. Turns out S3 is setting the content-type for everything to be "image/jpeg." It doesn't seem to break anything, but I'll probably still fix it at some point for the sanity's sake.

Turns out the content type problem was causing problems on certain versions of IE, plus I can imagine that specifying an incorrect content type might have a performance penalty. I added a few lines to aws_deploy_tools.rb to automatically detect the content type based on the file extension. It uses part of Actionpack, which means you have to add rails to your gemfile.

I've also fixed a problem with cloudfront invalidations. It turns out that they work with paths instead of files. So, e.g., it's necessary to post invalidation requests for both "/index.html" and "/"

By the way– Anyone interested in seeing some of this converted to a gem?

This would make a useful gem.

I'll note that I had to replace "$public_dir" in the Rakefile excerpt with "public_dir"; otherwise the Dir.glob line tried to scan my entire hard drive since $public_dir is unset.

I would really like this into a gem! :)

I think there could be an issue with
def compress_img (files = [])
in that $img_options[:output_ext] is being forced to jpg by
$img_options = {
:output_ext => "jpg",
:max_width => "600",
:max_height => "1200",
:quality => "65",
:compress_type => "JPEG"
}

This means png files are being converted to jpg and deleted.