Personal tools

Binary IO

From HaskellWiki

(Difference between revisions)
Jump to: navigation, search
(style)
(just use Data.Binary)
Line 1: Line 1:
There are a number of binary I/O libraries available for Haskell.
+
== Data.Binary ==
  +
  +
There are a number of binary I/O libraries available for Haskell. The
  +
best to use is the new, semi-standard Data.Binary library:
  +
  +
* [http://hackage.haskell.org/cgi-bin/hackage-scripts/package/binary Data.Binary]
  +
  +
It's very simple to use, and provides a highly efficient, pure interface
  +
to binary serialisation.
  +
  +
A tutorial:
  +
  +
* [[Serialisation_and_compression_with_Data_Binary]]
  +
  +
== Other libraries ==
   
 
* JeremyShaw's update of HalDaume's NewBinary package (Cabalized): http://www.n-heptane.com/nhlab/repos/NewBinary
 
* JeremyShaw's update of HalDaume's NewBinary package (Cabalized): http://www.n-heptane.com/nhlab/repos/NewBinary
Line 10: Line 10:
 
For very simple serialisation, use <hask>read</hask> and <hask>show</hask>.
 
For very simple serialisation, use <hask>read</hask> and <hask>show</hask>.
   
If you have simple binary IO requirements, then Data.ByteString might be easiest -- you get a List-like interface to packed byte arrays (interface documented [http://www.cse.unsw.edu.au/~dons/fps/Data.FastPackedString.html here]). For more complex serialisation, NewBinary would be preferred. [[Lambdabot]] follows this rule: when serialising lists and maps, it uses Data.ByteStrings. For complex algebraic datatypes, NewBinary is used (see [http://www.cse.unsw.edu.au/~dons/code/lambdabot/Plugin/Seen.hs Plugin/Seen.hs] in Lambdabot).
+
If you have simple binary IO requirements, then Data.ByteString might be easiest -- you get a List-like interface to packed byte arrays (interface documented [http://www.cse.unsw.edu.au/~dons/fps/Data.FastPackedString.html here]). For more complex serialisation, Data.Binary would be preferred.
 
NewBinary is based on Binary.hs from GHC, and traces back to NHC's binary library, described in [ftp://ftp.cs.york.ac.uk/pub/malcolm/ismm98.html "The Bits Between The Lambdas"]. It is simple to use: for each value you wish to serialise/unserialise, you write an instance of Binary for its type. This can be automated with the DrIFT tool, which will derive binary for you (as is done in GHC's .hi files). You can rip out your own copy of Binary.hs from NewBinary -- just take Binary.hs and FastMutInt.lhs.
 
 
Lamdabot, hmp3 and hs-plugins are applications written using NewBinary, and it fairly simple and effective. Here, for example, is the binary serialisation code for the [http://www.cse.unsw.edu.au/~dons/code/hmp3.html hmp3] database type, an Array, and a user defined type: File. Easy!
 
 
<haskell>
 
instance Binary a => Binary (Array Int a) where
 
put_ bh arr = do
 
put_ bh (bounds arr)
 
mapM_ (put_ bh) (elems arr)
 
get bh = do
 
((x,y) :: (Int,Int)) <- get bh
 
(els :: [a]) <- sequence $ take (y+1) $ repeat (get bh)
 
return $! listArray (x,y) els
 
 
instance Binary File where
 
put_ bh (File nm i) = do
 
put_ bh nm
 
put_ bh i
 
get bh = do
 
nm <- get bh
 
i <- get bh
 
return (File nm i)
 
</haskell>
 
 
As an aside, SerTH lets you derive instance of a Binary-alike class automagically using TH, rather than requiring hand-written instances, or DrIFT.
 
 
So, in summary. There are a number of ways to do binary IO in Haskell, efficiently and simply. It's really not that hard at all.
 
   
 
[[Category:Tutorials]]
 
[[Category:Tutorials]]

Revision as of 04:50, 9 July 2007

1 Data.Binary

There are a number of binary I/O libraries available for Haskell. The best to use is the new, semi-standard Data.Binary library:

   * Data.Binary

It's very simple to use, and provides a highly efficient, pure interface to binary serialisation.

A tutorial:

   * Serialisation_and_compression_with_Data_Binary

2 Other libraries

For very simple serialisation, use
read
and
show
.

If you have simple binary IO requirements, then Data.ByteString might be easiest -- you get a List-like interface to packed byte arrays (interface documented here). For more complex serialisation, Data.Binary would be preferred.