Distributed Systems in Haskell
yager.ioDistributed Programming was one of the things flagged "Immature" in State of the Haskell Ecosystem [1]. Just out of curiosity, was that a motivation behind the project, or do you have any other thoughts on that post-project?
[1] https://github.com/Gabriel439/post-rfc/blob/master/sotu.md#d...
It was definitely a factor. When we decided to use Haskell on the first project, we were mostly just curious how viable it would be for us.
If the current state of Distributed Programming in Haskell is "Immature", then I can't wait to see "Mature" :)
Excellent write up. One point though: watch out for WriterT and RWST leaking memory:
http://stackoverflow.com/questions/25280852/space-leak-in-pi...
It's often better to just include the writer part you need in a larder State type (lenses make this really seamless)/
That's a good point, but it was also mentioned in the article, with a link to a useful mailing list thread[1] on the matter.
It seems to be a matter of laziness. If you're writing some small piece of state like an Int, this can be a real problem. If you're writing out a lazy data structure like a list (as happens in the article), I think the extra laziness should be fine or even desired. (I've never used Writer in practice, so take my comments with a grain of salt!)
[1]: http://comments.gmane.org/gmane.comp.lang.haskell.libraries/...
It was? I still can't see it...
AFAIR the problem with WriterT was the use of mappend within is always lazy, regardless of the strictness of the w in WriterT w m a.
Yep, I mentioned this at the bottom of "monad it".
Did you run into any bugs or strange beahviour in Cloud Haskell? I looked into using it a while ago and came to the (possibly hasty) conclusion that it's not stable enough to build a large project on. Thanks!
None. It was rock solid; never a hiccup. It's evolving, but it seems quite well written and correct. As with many Haskell libraries in "beta", it works perfectly even though it's not finalized.
Author here. Glad to see this on HN! Happy to answer any questions.
Thanks for the writeup! I am wanting to learn Haskell and this inspires confidence that one can develop larger projects using a more rigorous, less error prone approach. I like it.
Shouldn't the type of the monadic serverStep be "Message -> m ()" to match the type of the explicit "serverStep :: Config -> State -> Message -> (State, [Message])" ? Aka, serverStep isn't a monadic object, it's a Kleisli arrow.
You're right, that's a typo. I even wrote it out correctly later. I'll fix that. Thanks!
Great post. How much knowledge and experience of haskell did you have before starting this project?
I'd been dabbling with it for a year or two.
Probably my most significant previous use was in my second-semester computer architecture class as a substitute for Verilog. Using Christiaan Baaij's Cλash project, one can transform a (large) subset of Haskell into VHDL/Verilog/SystemVerilog and slap it on an FPGA. There's a rundown of that project on my blog too.
My project partner had never really used Haskell before, so he was learning as we went. By the end of the semester he was writing code no problem. "Immersion learning" seemed to work quite well here.
Why Haskell? What made you choose Haskell?
For the first project, curiosity. For subsequent projects, the desire to repeat our experience with the first project.
This was a great post. Really love how you walked through the example code.
Very interesting structure of explanation. I like your short sumarization (tl;dr) at beginning and "why?" reasoning. I think i will copy it in my next book :-)