02:47 <discocaml> <darrenldl> mccd: saw your posts about stitch, i gave it a try and it's pretty neat i have to say

02:59 <dh`> er, immutable, yes

02:59 <dh`> obviously, except apparently to my fingers

08:51 <discocaml> <dinosaure> miou has green thread but I prefer the `fiber` term which is, IMHO, is better defined. But miou, and this is where it differs from eio, automatically has a pool of domains & is able to launch a parallel task which is represented also as a promise. This is the diff between `Miou.call_cc` which launches a fiber and `Miou.call` which asks a domain to run the given task. Both return a `'a Miou.t` which the promise of the task. In other words, it

08:53 <discocaml> <dinosaure> It is described here: https://robur-coop.github.io/miou/#when-not-to-use-miou and if you want to understand how to implement a basic scheduler with effects, this book is perfect 🙂

10:38 <discocaml> <kayceesrk> Hey @dinosaure

10:38 <discocaml> <kayceesrk> > Computations are periodically interrupted

10:38 <discocaml> <kayceesrk> Does that mean there is preemptive multi-threading in miou? i.e, timer interrupts forcing a context switch? How does this work?

10:42 <discocaml> <dinosaure> I should probably ask what you mean by context-switch but to clarify according to your example, a `Miou_unix.sleep` will invoke `select(2)` and leave the opportunity for another pending task to run (according to the round-robin policy).

10:43 <discocaml> <dinosaure> basically, any effects do a suspension, put the suspended task in the end of the todo list, then miou lookup for system events and run the next pending task

10:43 <discocaml> <dinosaure> a special case exists for the cancellation but we are deeper in details 🙂

10:50 <discocaml> <dinosaure> ah and about domains, actually, I/O are local to domains - we don't share events across domains (it's too complicated and probably not really interesting)

10:51 <discocaml> <dinosaure> so a domain can sleep and, at the same time, another domain can run few tasks (and handle its events)

11:45 <anadon> Is there a good way to split strings taking into account escaped characters?

11:59 <octachron> It depends what you mean by "good way" and "escaped characters" (and on the nature of "strings"). If you have small delimiters, comparing normalized character subsequences with the delimiter is probably good enough.

12:09 <anadon> I'm trying to split $PATH correctly, and handling "\:" correctly, because _somebody_ is going to include a colon in their actual filepath eventually.

12:11 <discocaml> <limp.biskit> anadon: i would just write a regex will

12:12 <discocaml> <limp.biskit> something like [^\]:

12:12 <discocaml> <limp.biskit> something like [^\\]:

12:13 <anadon> I'm looking at that, and checking if I can have the assumption that results maintain ordering.

12:15 <anadon> Also, if I do that, I'd have to correctly infer when to re-merge on splits which would be a pain.

12:17 <anadon> It really looks like it'll just be easier to implement my own from scratch. I don't think it's worth the fuss of using the stdlib.

13:02 <discocaml> <kayceesrk> > context switch

13:02 <discocaml> <kayceesrk>

13:02 <discocaml> <kayceesrk> If a fiber is purely CPU bound (spins in an infinite loop), does it allow other fibers to run? From your answer, probably not. (Neither does Eio).

13:10 <discocaml> <dinosaure> ah yes, in this case, we the fiber keeps the exclusivity of the cpu but I'm not certain that it's possible to avoid this case in pure OCaml

13:10 <discocaml> <dinosaure> ah yes, in this case, yes the fiber keeps the exclusivity of the cpu but I'm not certain that it's possible to avoid this case in pure OCaml

13:27 <companion_cube> With threads you might still make progress

13:27 <companion_cube> Unless you have n fibers locked in a loop at the same time of course

14:44 <discocaml> <limp.biskit> sounds like you want something with preemptive multitasking

14:58 <Leonidas> anadon: also, what about "C:\path\with\colon":more:paths

15:31 <neiluj> hi! can this pattern be simplified? https://bpa.st/JJQA

15:37 <discocaml> <hockletock> `Ok y` can just be `Ok _`

15:57 <neiluj> yep! thx

16:56 <anadon> Leonidas: They need to be escaped, unless there is some other canonical way to split `Sys.getenv "PATH"` correctly.

16:57 <anadon> But seeing as this is for bash shell checking, I doubt Windows will ever have to be taken into account.