00:03 <companion_cube> I'm not sure, it doesn't really have the same notion of fiber does it? :/

00:04 <companion_cube> Or, at least, promise

01:11 <discocaml> <dinosaure> Note that if you want httpaf and a scheduler with cancellation, you have miou with https://github.com/robur-coop/httpcats 🙂 . As moonpool, you can cancel any fibers and we explain how to handle the sigint signal to clean-up and stop an echo server in our book here: https://robur-coop.github.io/miou/conditions_and_mutexes.html#conditions. `httpcats` is not released because we currently do a bug change on `mirage-crypto`/`tls` - and it's no

01:12 <discocaml> <dinosaure> Also, we did some small benchmark and we currently are twice faster than cohttp/eio (~200k req/sec where cohttp/eio handles only 100k req/sec)

01:14 <discocaml> <dinosaure> Of course, all of these is pretty experimental but, at this stage, we mostly need users to start an interactive loop to improve Miou and httpcats

01:17 <discocaml> <darrenldl> > eio's cancellation is confusing

01:17 <discocaml> <darrenldl> i share the sentiment, though in my scenario i couldn't tell if i was the stupid one or not so i didn't chase down the rabbit hole

01:18 <discocaml> <dinosaure> And yeah. Cancellation is a hard issue - even more when it's about domains. For Miou, we tried to specify it a lot with a certain vision: we call it asynchronous cancellation. It does not block your all process (we actually can execute a part of some of your tasks) but we prioritize it over anything. Also, not sur for riot but cancellation cancels all subtasks of a task too (structured concurrency).

01:19 <companion_cube> Yeah it's always something you have to cooperatively check

01:19 <companion_cube> If you don't have a specialized runtime like the beam

01:19 <companion_cube> (i think riot must do the same with supervision trees)

01:53 <discocaml> <struktured> has anyone seen a signature emitted by utop which has a line like ```val _ : int -> string```. The emphasis here is the underscore where I would expect a valid symbol name instead. I was wondering why that happens and what it means. Thanks

02:06 <discocaml> <limp.biskit> that is very very cool, might have a look at porting sometime

02:07 <discocaml> <limp.biskit> i just don’t get why they can’t expose cancellation to the user

02:07 <discocaml> <limp.biskit> you shouldn’t have to throw exceptions as the only way to kill something

02:10 <discocaml> <limp.biskit> do you know if there’s any caqti equivalent for miou yet?

02:11 <discocaml> <dinosaure> I don't think so 😕 but you can try 😉 it can be a nice oppportunoty to test it

02:33 <discocaml> <limp.biskit> also - any lwt integration :p

02:53 <discocaml> <dinosaure> Not the case too, I need to think about that but it requires lot of work I think

07:50 <discocaml> <darrenldl> ikr

07:51 <discocaml> <darrenldl> i presume there are deep technical reasons why it's not exposed, but anyway

07:52 <discocaml> <darrenldl> i struggled with cancellation seemingly not propagating through some combinators, but i have yet to make minimal examples to confirm anything

10:21 <discocaml> <limp.biskit> that being said i haven’t tried manually creating cancellation contexts. the docs seem to warn against it

11:37 <discocaml> <limp.biskit> i think manually creating cancellation contexts are impractical since children are still attached to a switch and inherit it’s cancellation contexts. behind the scenes i think it’s still raising Cancelled exceptions so atp it’s probably better to have a dedicated X_cancelled exception for intentional cancelled operations

11:37 <discocaml> <limp.biskit> such a strange way of doing things

12:04 <Armael> /w 15

19:17 <Anarchos> what is the dune branch/tag used when 'opam switch' is set on 4.14.2 (the LTR release of ocaml 4) ?

19:45 <discocaml_> <ilias1911> Hi guys! I'm looking for data structures for term-indexed maps, with terms potentially containing variables. I'd like to be able to perform queries returning all values associated to keys that are unifiable to a given key but I don't care about isos. I know of @octachron 's work (https://github.com/Drup/dowsing) but if I can, I'd rather read a paper describing a solution rather than a large amout of code! I'm considering using a trie index

19:47 <companion_cube> (this is Drup's, not octachron, btw)

19:47 <discocaml_> <ilias1911> Ah sorry, will fix my message 🤦‍♂️

19:47 <companion_cube> for unification you could use a discrimination tree, it's kind of like a trie

19:47 <discocaml_> <Kali> dont

19:47 <discocaml_> <Kali> edits resend on the irc bridge

19:47 <discocaml_> <Kali> and that's a big block of text

19:48 <discocaml_> <ilias1911> Oh, good to know. Will not fix then! Sorry for pinging the wrong person.

19:48 <discocaml_> <ilias1911> Thanks for the reference CCube

19:52 <discocaml_> <ilias1911> I see, so iiuc it's easy to have keys containing variables in the domain of the map and check they are instantiable to a query but harder to do the reverse. That'd match what I guessed from my small experiments

19:56 <companion_cube> you can query for all terms matching your query term

19:56 <companion_cube> or all terms that are instances of your query term

19:56 <companion_cube> but one of the two (I think the latter) is more expensive

20:02 <prgbln> is there an easy way to stop dune from complaining about unused constructors? can't prefix with an underscore unfortunately

20:03 <discocaml_> <contextfreebeer> disable that warning in your dune file maybe?

20:05 <prgbln> was hoping there's something akin to the underscore prefix, but i suppose no

20:06 <discocaml_> <hockletock> `[@warning "-37"]`

20:08 <discocaml_> <kakadu18> @ilias1911 It looks like you are implementing a tabling from logic programming. You could try to find some paper about it... We have implemented somthing similar for OCanren project

20:10 <prgbln> hockletock: this works when placed next to the type keyword, cheers

20:10 <discocaml_> <hockletock> can disable it for a whole module with `[@@@warning "-37"]` or add the line `(env (dev (flags :standard -warn-error -37)))` to your dune-workspace @prgbln

20:11 <prgbln> hockletock: thank you, didn't know about doing this for a module

20:46 <discocaml_> <ilias1911> Thanks Kakadu, will have a look as well!