News for hackers. Stuff that matters.

A comment on Haskell's Missing Concurrency Basics (2016). Parent comment

Re:Haskell's Missing Concurrency Basics (2016) by SlySherZ on Friday May 11, 2018 @04:37AM (#17045785)
Elixir newbie here. If I remember correctly IO is implemented as a process, which means that different write requests are processed sequentially in the order they arrive to the process. The following link has more information: http://erlang.org/doc/apps/stdlib/io_protocol.html
[ Parent | on HN ]
Re:Haskell's Missing Concurrency Basics (2016) by phoe-krk on Friday May 11, 2018 @05:37AM (#17045972)
Correct. Erlang (and therefore Elixir) stdio is implemented by means of sending messages to a process that does low-level writing to a console. When a single message is served, the other are queued up.

Therefore, when two processes (let's name them A and B) want to output something at the same time, the result is going to be AAAABBBB or BBBBAAAA (depending on the order in which the messages arrive in the mailbox), but never BBAABBAA or anything similar.

[ Parent | on HN ]
Re:Haskell's Missing Concurrency Basics (2016) by gmfawcett on Friday May 11, 2018 @09:53AM (#17047277)
...where "AAAA" and "BBBB" are two distinct messages, and not two sets of four messages each ("A", "A", "A", "A").

When I first read your example, it sounded like you were saying that the IO process would exhaust all messages from one source process before processing any messages from the other, no matter what order they arrived in.

[ Parent | on HN ]
Re:Haskell's Missing Concurrency Basics (2016) by phoe-krk on Saturday May 12, 2018 @05:43AM (#17053256)
Yes, correct. I wasn't clear enough.
[ Parent | on HN ]
Re:Haskell's Missing Concurrency Basics (2016) by amelius on Friday May 11, 2018 @07:12AM (#17046330)
It's very customary to let logging and such be done by a separate thread; this has also the benefit that the original (sending) thread can continue without waiting for the device.
[ Parent | on HN ]