
Hello, cafe. Is it possible to read data from different concurrent sources, i.e. read data from source as soon as it become avaliable, e.g. runResourceT $ (source1 stdin $= CL.map Left) >=> (source2 handle $= CL.map Right) $= application $$ sink where >=> - stands for concurrent combining of sources It would be good if it can be sources of different types (handle or STM channel, etc..). Currently I've found no good way to handle with this situation, except of using STM Channels for collecting data source1 ---+ | | sink | output sink +---] Channel [-------> application----->] | source source2 ---+ | From this point of view application takes concurent data, but this implementation requires additional thread per data processing. Also in many cases it will require run additional runResourceT (see later example). So if there any possible simplifications? Or ideas how to make (>=>) operator. Example: So I've got next code in my network-conduit based application: main :: IO () main = do pool <- createDBPool "..." 10 let r = ServerInit pool forkIO $ forever clientConsole --read channel list and send "Left" flip runReaderT r $ runTCPServer (ServerSettings 3500 Nothing) (protoServer) myServer src sink = do ch <- liftIO $ atomically $ newTBMChan 16 initState <- lift $ ask _ <- liftIO $ fork . (flip runReaderT initState) $ runResourceT $ src $= C.sequence decode $= CL.map Right $$ sinkTBMChan ch sourceTBMChan ch $= process $= C.sequence encode $$ sinkHandle stdout But in this situation I don't know if freeing of all resources are guaranteed, because I'm running additional resourceT in main resourceT scope. So can you advice is it possible to make concurrent sources now with currenly implemented library? If it's not possible but worth of implementing, so I can make that functions? Is it correct to runResourceT inside another resourceT? -- Best regards, Alexander V Vershilov