On 1/24/12 9:43 AM, Felipe Almeida Lessa wrote:

Use cereal [1], usually it's fast and easy enough.

Out of curiosity, is binary no longer the recommended standard for such things? Cheers, Greg

On 1/24/12 5:51 PM, Vincent Hanquez wrote:

On 01/24/2012 07:33 AM, Gregory Crosswhite wrote:

...

On 1/24/12 9:43 AM, Felipe Almeida Lessa wrote:

...

Use cereal [1], usually it's fast and easy enough.

Out of curiosity, is binary no longer the recommended standard for such things?

binary got only an interface for processing lazy bytestring. cereal is able to do strict and lazy bytestring and got a partial interface like attoparsec (which is required to do proper network/io processing).

Fortunately it's very simple to convert between the two, since the actual serialization API is really close.

Features-wise, in my view, cereal is a superset of binary. the only thing missing that i've noticed is that you can't tell how many bytes you have processed with cereal.

Fair enough, it's just that I had gotten the impression that for some time the binary package was considered by the community to the community to be the "standard" way of serialization/deserialization values. Is this no longer the case? Cheers, Greg

It's not as efficient for Maps, but you might want to look at the swapper package: http://hackage.haskell.org/package/swapper It transfers Haskell data structures (any functors) directly to and from disk. Tom On 1/23/12, Krzysztof Skrzętnicki wrote:

From my experience I can recommend msgpack ( http://hackage.haskell.org/package/msgpack) as being extremely fast. It comes with optimized prepared instances for common data structures which is very nice, because you don't have to roll your own version with library like cereal (which is indeed very fast, but simply less convenient).

Best regards, Krzysztof Skrzętnicki

On Tue, Jan 24, 2012 at 00:37, Nick Rudnick wrote:

...

Dear all,

if you want to temporarily store haskell data in a file – do you have a special way to get it done efficiently?

In an offline, standalone app, I am continuously reusing data volumes of about 200MB, representing Map like tables of a rather simple structure,

key: (Int,Int,Int) value: [((Int,Int),LinkId)]

which take quite a good deal of time to produce.

Is there a recommendation about how to 'park' such data tables most efficiently in files – any format acceptable, quick loading time is the most desirable thing.

Thanks a lot in advance, Nick

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe