Editing Dealing with binary data (section)

===Binary generation===

In contrast to parsing binary data, you might want to generate it. This is the
job of the <tt>Put</tt> monad. Follow along with the
[http://hackage.haskell.org/package/binary/docs/Data-Binary-Put.html documentation]
if you like.

The <tt>Put</tt> monad is another state-like monad, but the state is an offset
into a series of buffers where the generated data is placed. All the buffer
creation and handling is done for you, so you can just forget about it. It
results in a lazy bytestring (so you can generate outputs that are larger than memory).

Here's the reverse of our simple <tt>Get</tt> example:

<haskell>import qualified Data.ByteString.Lazy as BL
import Data.Binary.Put

serialiseSomething :: Put
serialiseSomething = do
  putWord32be 1
  putWord16be 2
  putWord8 3

main :: IO ()
main = BL.putStr $ runPut serialiseSomething</haskell>

And running it shows that it's generating the correct serialisation:

<pre>% runhaskell /tmp/example.hs| hexdump -C
00000000  00 00 00 01 00 02 03                              |.......|</pre>

If you want the output of <tt>runPut</tt> to be a strict bytestring, you just
need to convert it with <hask>B.concat $ BL.toChunks $ runPut xyz</hask>.

One limitation of <tt>Put</tt>, due to the nature of the <tt>Builder</tt> monad
which it works with, is that you can't get the current offset into the output.
This can be an issue with some formats which require you to encode byte offsets
into the file. You have to calculate these byte offsets yourself.