On Wed, Jun 27, 2012 at 8:15 PM, Christian Höner zu Siederdissen <choe...@tbi.univie.ac.at> wrote: >> IMHO conduit is the best you can get right now. It's much simpler to >> use than enumerator, which was already simpler than iteratee. > > Well, as will all packages one has to understand how to write certain > stuff. I have to say, in defense of iteratee, that once you know it, it > is very nice to use. Especially in a monadic environment. I still need > to learn conduit.
Understanding how to do something may not make it less painful ;-). I find that conduit model's is much easier to understand and put to use, especially since most of the time I'm not using these libraries (so I don't have all of their details on my mental Haddock cache). IME Iteratees are a pleasure, Enumerators are interesting and sometimes confusing while Enumeratees are just plain painful. Sources, Sinks and Conduits all are simple concepts with simple implementations of the same complexity. That's how I see their differences. On a side note, I probably wouldn't have refactored my Stockholm code into an enumerator or iteratee library just because of the pain in using Enumeratees =/. >> It's used in the real world by many people, unlike pipes which isn't >> being used anywhere I know of. It suffers a lot of bashing even from >> people who don't like it =), and Michael is a very responsive and nice >> maintainer. > > I'll not start any comments. I know neither conduit nor pipes, only the > older ones (iterIO, enumerator, iteratee). When I started iteratee won > due to "zip". That's a wise choice =). >> As selfish as it sounds, I don't plan on working outside conduit >> unless I need to =). > > Why would that be selfish? Unlike Nicholas, I don't intend to support a port of biostockholm or any other library to something other than conduit -- this kind of selfish =). >> > Finally, anybody (Felipe?) use conduit/list.zipSinks yet? >> >> Never needed to use it myself, did you find anything wrong with it? > > No, it's just that I have an important use case for it. Two or more > sinks creating data structures from multigigabyte files, not necessary > in lock-step. And it could just happen that I can /not/ go twice over > the data as I have a the program in a linux-pipe. Well, I can't say anything other than just "try it" =). Cheers, -- Felipe. _______________________________________________ Biohaskell mailing list Biohaskell@biohaskell.org http://malde.org/cgi-bin/mailman/listinfo/biohaskell
