On Sat, Jan 26, 2013 at 07:46:04AM -0200, Thiago Negri wrote:

Do you need advice on what? I didn't understand your last phrase.

IIUC, it has to do which how to interrupt a blocking call to ncurses event handling, from another thread (the calculation thread) to let the UI thread refresh the UI with something to display (the calculation). I've been confronted with the same problem in the past, and i haven't found a satisfactory solution with ncurses. Erik, I think the only way to do that with ncurses, is to wrap the ncurses getevent with a select loop on the terminal fd and an arbitrary fd to be able to "ping" the select loop for arbitrary processing. You can also replace the arbitrary fd with a SIGALARM to yourself which i believe would interrupt select too. I don't think there's any solution that integrate nicely with the haskell way, but i'ld be happy to be wrong on this. In my case, i switched to Vty which solved this problem, by allowing to send arbitrary strongly-typed event to widgets. -- Vincent

Erik de Castro Lopo wrote:

...

Do you need advice on what? I didn't understand your last phrase.

Well I have data from two sources, stdin and the calculation thread. If I was doing this in C, I'd probably use a pipe for the calculation data and then do select on the two file descriptors.

There is a select package:

http://hackage.haskell.org/package/select

but I was wondering if there was a more idiomatic Haskell way of dealing with inputs from more than one source.

Of course. Threads communicate through MVars, Chans and other abstractions. If you have more complicated scenarios like one thread communicating with multiple other threads in a selective fashion you may also be interested in software transactional memory (STM). If you are using separate threads to perform calculations, you should make sure that you only pass fully evaluated values. In many cases it suffices to evaluate to WHNF, in which case you can simply write: putMVar var $! x This writes the WHNF of 'x' to the MVar 'var'. If the values you pass around are of more complicated data types with non-strict parts (like (Integer, Integer) instead of just Integer), you can also evaluate to NF: import Control.DeepSeq putMVar var $!! x Another option is go without concurrency entirely. Since this is about parallel calculations, I suppose that your computations are actually pure. Let me give you an example: let xs :: [Integer] xs = map (^1000000) [2..1000] mapM_ print xs To parallelize this you can simply use parallel strategies: import Control.Parallel.Strategies let xs :: [Integer] xs = parMap rseq (^1000000) [2..1000] mapM_ print xs This calculates the list values in parallel and is as simple as replacing "map" by "parMap rseq" or "parMap rdeepseq" (if the elements have non-strict parts like (Integer, Integer) instead of just Integer). Remember that for pure values you can always just say, "please evaluate this in parallel". No need at all to mess around with threads. Greets, Ertugrul -- Not to be or to be and (not to be or to be and (not to be or to be and (not to be or to be and ... that is the list monad.