[Haskell-cafe] Re: Fwd: Semantics of iteratees, enumerators,
apfelmus at quantentunnel.de
Wed Sep 1 08:10:22 EDT 2010
Tilo Wiklund wrote:
> Daniel Fischer wrote:
>> Well, I just gave an example where one would want chunking for reasons
>> other than performance. That iteratees don't provide the desired
>> functionality is a different matter.
> In the case of hashing, wouldn't it be more reasonable to consider
> iterators over streams of fixed (or at least predictable) sized chunks
> (where a set of chunks can themselves be chunked), with the chunking
> behaviour being given by another iteratee over the original stream?
> It seems to me that one of the major points of iteratees is to provide
> an abstraction from the kind of chunking irrelevant to the parsing
> logic, otherwise I fail to see any difference (at least relevant to
> chunking) to plain strict IO.
I thought so, too, but I was informed that iteratees are just a small
step up the abstraction ladder. The difference compared to an ordinary
file Handle is that you can now reuse one and the same iteratee for
reading from a String , for instance, without changing the source code
of the iteratee.
Furthermore, iteratees can be suspended, which facilities resource
management like closing files handles after they've been read.
More information about the Haskell-Cafe