NOTE: This library monkeypatches core classes and wa designed for Ruby 1.8 & 1.9. A redesign using refinements would be much better. Minimal support is provided.

If you need to do read and write binary data, there is of course Array::pack and String::unpack. The packable library makes (un)packing nicer, smarter and more powerful. In case you are wondering why on earth someone would want to do serious (un)packing when YAML & XML are built-in: I wrote this library to read and write FLV files…

Strings, integers & floats have long forms instead of the cryptic letter notation. For example:

["answer", 42].pack("C3n")

can be written as:

["answer", 42].pack({:bytes => 3}, {:bytes => 2, :endian => :big})

This can look a bit too verbose, so let’s introduce shortcuts right away:

Most commonly used options have shortcuts and you can define your own. For example:

:unsigned_long  <===> {:bytes => 4, :signed => false, :endian => :big}

IO classes (File & StringIO) can use (un)packing routines. For example:

signature, block_len, temperature = my_file >> [String, :bytes=>3] >> Integer >> :float

The method each also accepts packing options:

StringIO.new("\000\001\000\002\000\003").each(:short).to_a  ===>  [1,2,3]

It’s easy to make you own classes (un)packable. All the previous goodies are thus available:

File.open(“great_flick.flv”) do |f| head = f.read(FLV::Header) f.each(FLV::Tag) do |tag| # do something meaningful with each tag… end end

It’s also easy to define special shortcuts that will call blocks to (un)pack any class. As an example, this could be useful to add special packing features to String (without monkey patching String::pack).

First, ensure that you’re running at least RubyGems 1.2 (check gem --version if you’re not sure – to update: sudo gem update --system).

Add GitHub to your gem sources (if you haven’t already):

sudo gem sources -a http://gems.github.com

Get the gem:

sudo gem install marcandre-packable

That’s it! Simply require 'packable' in your code to use it.

Designed to work with ruby 1.8 & 1.9.

The library was designed to be backward compatible, so the usual packing and unpacking methods still work as before. All packable objects can also be packed directly (no need to use an array). For example:

42.pack("n")  ===>  "\000*"

In a similar fashion, unpacking can done using class methods:

Integer.unpack("\000*", "n")  ===>  42

Although the standard string formats can still be used, it is possible to pass a list of options (see example in feature summary). These are the options for core types:

bytes

Number of bytes (default is 4) to use.

endian

Either :big (or :network, default), :little or :native.

signed

Either true (default) or false. This will make a difference only when unpacking.

precision

Either :single (default) or :double.

endian

Either :big (or :network, default), :little or :native.

bytes

Total length (default is the full length)

fill

The string to use for filling when packing a string shorter than the specified bytes option. Default is a space.

repeat

This option can be used (when packing only) to repeat the current option. A value of :all will mean for all remaining elements of the array.

When unpacking, it is necessary to specify the class in addition to any option, like so:

"AB".unpack(Integer, :bytes => 2, :endian => :big, :signed => false)  ===>  0x3132

It’s easy to add shortcuts for easier (un)packing:

String.packers.set :flv_signature, :bytes => 3, :fill => "FLV"

"x".pack(:flv_signature)  ===>  "xFL"

Two shortcut names have special meanings: default and merge_all. default specifies the options to use when nothing is specified, while merge_all will be merged with all options. For example:

String.packers do |p|
  p.set :merge_all, :fill => "*"	# Unless explicitly specified, :fill will now be "*"
  p.set :default, :bytes => 8     # If no option is given, this will act as default
end

"ab".pack  ===>  "ab******"
"ab".pack(:bytes=>4)  ===>  "ab**"
"ab".pack(:fill => "!")  ===>  "ab"     # Not "ab!!"

A shortcut can refer to another shortcut, as so:

String.packers do |p|

p.set :creator, :bytes => 4 p.set :app_type, :creator end

"hello".pack(:app_type)  ===>  "hell"

The following shortcuts and defaults are built-in the library:

:merge_all      =>   :bytes=>4, :signed=>true, :endian=>:big
:default        =>   :long
:long           =>   {}
:short          =>   :bytes=>2
:byte           =>   :bytes=>1
:unsigned_long  =>   :bytes=>4, :signed=>false
:unsigned_short =>   :bytes=>2, :signed=>false

:merge_all      =>   :precision => :single, :endian => :big
:default        =>   :float
:double         =>   :precision => :double
:float          =>   {}

:merge_all      =>   :fill => " "

All IO objects (in particular files) can deal with packing easily. These examples will all return an array with 3 elements (a string, an integer and another string):

io >> :flv_signature >> Integer >> [String, {:bytes => 8}]
io.read(:flv_signature, Integer, [String, {:bytes => 8}])
io.read(:flv_signature, Integer, String, {:bytes => 8})
[io.read(:flv_signature), io.read(Integer), io.read(String, :bytes => 8)]

In a similar fashion, these have the same effect although the return value is different

io << "x".pack(:flv_signature) << 66.pack << "Hello".pack(:bytes => 8)  # returns io
io << ["x", 66, "Hello"].pack(:flv_signature, {} , {:bytes => 8})       # returns io
io.write("x", :flv_signature, 66, "Hello", {:bytes => 8})               # returns the # of bytes written
io.packed << ["x",:flv_signature] << 66 << ["Hello", {:bytes => 8}]     # returns a "packed io"

The last example shows how io.packed returns a special IO object (a packing IO) that will pack arguments before writing it. This is to insure compatibility with the usual behavior of IO objects: io << 66 ==> appends “66” io.packed << 66 ==> appends “000000000B”

We “cheated” in the previous example; instead of writing io.packed.write(...) we used the shorter form. This works because we’re passing more than one argument; for only one argument we must call io.packed.write(66) less the usual write method is called.

Since the standard library desn’t define the >> operator for IO objects, we are free to use either io.packed or io directly. Note that reading one value only will return that value directly, not an array containing that value:

io.read(Integer)  ===>  42, not [42]
io.read(Integer,Integer)  ===>  [42,43]
io << Integer     ===>  [42]

Including the mixin Packable will make a class (un)packable. Packable relies on write_packed and unpacking on read_packed. For example:

class MyHeader < Struct.new(:signature, :nb_blocks)
  include Packable

  def write_packed(packedio, options)
    packedio << [signature, {:bytes=>3}] << [nb_blocks, :short]
  end

  def self.read_packed(packedio, options)
    h = MyHeader.new
    h.signature, h.nb_blocks = packedio >> [String, {:bytes => 3}] >> :short
    h
  end
end

We used the argument name packedio to remind us that these are packed IO objects, i.e. they will write their arguments after packing them instead of converting them to string like normal IO objects. With this definition, MyHeader can be both packed and unpacked:

h = MyHeader.new("FLV", 65)
h.pack  ===>  "FLV\000A"
StringIO.new("FLV\000A") >> Signature  ===>  [a copy of h]

A default self.read_packed is provided by the Packable mixin, which allows you to define read_packed as an instance method instead of a class method. In that case, read_packed instance method is called with the same arguments and should modify self accordingly (instead of returning a new object). It is not necessary to return self. The previous example can thus be shortened:

class MyHeader
  #...
  def read_packed(packedio, options)
    self.signature, self.nb_blocks = packedio >> [String, {:bytes => 3}] >> :short
  end
end

Instead of writing a full-fledge class, sometimes it can be convenient to define a sort of wrapper we’ll call filter. Here’s an example:

String.packers.set :length_encoded do |packer|
  packer.write  { |packedio| packedio << length << self }
  packer.read   { |packedio| packedio.read(packedio.read(Integer)) }
end

“hello!”.pack(:length_encoded) ===> “000000000006hello!” [“this”, “is”, “great!”].pack(**3).unpack(**3) ===> [“this”, “is”, “great!”]

Note that the write block will be executed as an instance method (which is why we could use length & self), while read is a normal block that must return the newly read object.

A final note to say that packers are inherited in some way. For instance one could define a filter for all objects:

Object.packers.set :with_class do |packer|
  packer.write { |io| io << [self.class.name, :length_encoded] << self }
  packer.read  do |io|
    klass = eval(io.read(:length_encoded))
    io.read(klass)
  end
end

[42, MyHeader.new("Wow", 1)].pack(:with_class, :with_class).unpack(:with_class, :with_class) ===> [42, MyHeader.new("Wow", 1)]

packable is licensed under the terms of the MIT License, see the included LICENSE file.

Author

Marc-André Lafortune