German Velasco

Software Developer

Do you break your Elixir eggs on the big end or the little end?

This blog post was originally published in thoughtbot's blog. You can find the original at thoughtbot.com/blog/do-you-break-your-elixir-eggs-on-the-big-end-or-the-little-end.

It is allowed on all hands, that the primitive way of breaking eggs, before we eat them, was upon the larger end; but his present majesty’s grandfather, while he was a boy… happened to cut one of his fingers. Whereupon the emperor his father published an edict, commanding all his subjects, upon great penalties, to break the smaller end of their eggs. The people so highly resented this law, that our histories tell us, there have been six rebellions raised on that account; wherein one emperor lost his life, and another his crown. – Gulliver’s Travels

Most of the time when dealing with an Elixir app, you do not have to worry about how your binaries are being represented (at least, I didn’t have to). But recently, while implementing the proof-of-work algorithm for Ethereum called Ethash, I found myself having to care a lot about the endianess of binaries.

I discovered that Elixir uses big-endian format by default. What that means in practice is that when we write a binary as <<4, 3, 2, 1>>, we are assuming that the bytes are ordered from most significant to least significant from left to right. So 4 is the most significant byte, and 1 is the least significant byte.

If we wanted to represent the same binary in little-endian format, we would expect to see <<1, 2, 3, 4>>, where 1 is the least significant byte and 4 is the most significant byte.

To better understand this, let’s compare them by representing them as unsigned integers,

# big endian <<4, 3, 2, 1>> == little endian <<1, 2, 3, 4>>
iex> :binary.decode_unsigned(<<4, 3, 2, 1>>, :big)
67305985

iex> :binary.decode_unsigned(<<1, 2, 3, 4>>, :little)
67305985

# big endian <<1, 2, 3, 4>> == little endian <<4, 3, 2, 1>>
iex> :binary.decode_unsigned(<<1, 2, 3, 4>>, :big)
16909060

iex> :binary.decode_unsigned(<<4, 3, 2, 1>>, :little)
16909060

As you can see, those binaries can be decoded into the same unsigned integers. They are just represented differently.

Getting the least significant byte

Sometimes, we may only be interested in reading the least significant byte. When our binaries are represented in big-endian format, we need to get the last byte. But if we could represent our binary in little-endian format, we could just get the first byte. And that sounds easier.

How can we do this?

One way to deal with this is by turning our binary into a list, reversing that, and grabbing the first element.

<<4, 3, 2, 1>>
|> :binary.bin_to_list()
|> Enum.reverse()
|> hd()

# => 1

I don’t like this very much because there’s no indication that we’re reversing the binary to get the little-endian format.

A better way to deal with this is to decode the binary from big-endian into an unsigned integer and then encode it as little-endian (essentially using the equivalence we saw above),

<<4, 3, 2, 1>>
|> :binary.decode_unsigned(:big) # could omit :big since it's default
|> :binary.encode_unsigned(:little)

# => <<1, 2, 3, 4>>

Then we can do some binary pattern matching to grab the first byte (8 bits),

<<head::size(8), rest::binary>> =
  <<4, 3, 2, 1>>
  |> :binary.decode_unsigned(:big) # could omit :big since it's default
  |> :binary.encode_unsigned(:little)

head
# => 1

List of unsigned integers

Other times, we may be interested in representing a large binary as a series of unsigned integers (I had to do that a lot for the proof-of-work). Let’s see how we can do this assuming we want to turn a little-endian binary into a series of 32 bit unsigned integers.

If the binary in question is only four bytes long (32 bits), then the task is simple. We could use :binary.decode_unsigned/2,

:binary.decode_unsigned(<<1, 2, 3, 4>>, :little)
# => 67305985

If the binary is larger, and we do not know its size ahead of time, then we could follow a brute-force approach by turning our binary into a list, grabbing chunks of four, turning each of them back into binary, and decoding them.

Let’s use <<1, 2, 3, 4, 5, 6, 7, 8>> as an example,

<<1, 2, 3, 4, 5, 6, 7, 8>>
|> :binary.bin_to_list()
|> Enum.chunk_every(4)
|> Enum.map(&:binary.list_to_bin/1)
|> Enum.map(&:binary.decode_unsigned(&1, :little))

# => [67305985, 134678021]

That does the job, but there’s a better way! Binary pattern matching allows us to specify the unit and size of its parts, and it allows us to use modifiers so that we can express exactly what we’re trying to do.

Once again, let’s start with a binary that is only four bytes long (32 bits). Since we know the size of the binary, we can specify it via size and unit options, and we can use the unsigned and little modifiers,

# specific size and unit
<<number::size(4)-unit(8)-unsigned-integer-little>> = <<1, 2, 3, 4>>

number
# => 67305985

# or you can just use size * unit
<<number::size(32)-unsigned-integer-little>> = <<1, 2, 3, 4>>

number
# => 67305985

That last one reads particularly well (in my opinion) because it states exactly what we want, a 32 bit unsigned integer (from a little-endian binary).

And for larger binaries, we can combine what we know about binary pattern matching with a bitstring generator for a really succinct result!

binary = <<1, 2, 3, 4, 5, 6, 7, 8>>

for <<number::size(32)-unsigned-integer-little <- binary>> do
  number
end

# => [67305985, 134678021]

What’s next?

I hope that by looking at these brief examples you are pleasantly surprised with the excellent support Elixir has for dealing with binaries. I know I am!

If you found this interesting, I recommend reading the whole documentation section for the <<args>> macro. There’s some good stuff in there.

And last but not least… when dealing with endiannes, always be extra careful. It can get confusing, and you might just reverse a few things.

Until next time!

<<117, 110, 116, 105, 108, 32, 110, 101, 120, 116, 32, 116, 105, 109, 101>>
|> :binary.decode_unsigned()
|> :binary.encode_unsigned(:little)

"emit txen litnu"