Question

Basically I want to stream data from memory into a tar/gz format (possibly multiple files into the tar, but it should NEVER TOUCH THE HARDDRIVE, only streaming!), then stream them somewhere else (an HTTP request body in my case).

Anyone know of an existing library that can do this? Is there something in Rails?

libarchive-ruby is only a C wrapper and seems like it would be very platform-dependent (the docs want you to compile as an installation step?!).

SOLUTION:

require 'zlib'
require 'rubygems/package'

tar = StringIO.new

Gem::Package::TarWriter.new(tar) { |writer|
  writer.add_file("a_file.txt", 0644) { |f| 
    (1..1000).each { |i| 
      f.write("some text\n")
    }
  }
  writer.add_file("another_file.txt", 0644) { |f| 
    f.write("some more text\n")
  }
}
tar.seek(0)

gz = Zlib::GzipWriter.new(File.new('this_is_a_tar_gz.tar.gz', 'wb'))  # Make sure you use 'wb' for binary write!
gz.write(tar.read)
tar.close
gz.close

That’s it! You can swap out the File in the GzipWriter with any IO to keep it streaming. Cookies for dw11wtq!

Answer 1

Take a look at the TarWriter class in rubygems: http://rubygems.rubyforge.org/rubygems-update/Gem/Package/TarWriter.html it just operates on an IO stream, which may be a StringIO.

tar = StringIO.new

Gem::Package::TarWriter.new(tar) do |writer|
  writer.add_file("hello_world.txt", 0644) { |f| f.write("Hello world!\n") }
end

tar.seek(0)

p tar.read #=> mostly padding, but a tar nonetheless

It also provides methods to add directories if you need a directory layout in the tarball.

For reference, you could achieve the gzipping with IO.popen, just piping the data in/out of the system process:

http://www.ruby-doc.org/core-1.9.2/IO.html#method-c-popen

The gzipping itself would look something like this:

gzippped_data = IO.popen("gzip", "w+") do |gzip|
  gzip.puts "Hello world!"
  gzip.close_write
  gzip.read
end
# => "\u001F\x8B\b\u0000\xFD\u001D\xA2N\u0000\u0003\xF3H\xCD\xC9\xC9W(\xCF/\xCAIQ\xE4\u0002\u0000A䩲\r\u0000\u0000\u0000"