07:12 <greenbagels> hmm im parsing data line-by-line from a flat text file and trying to json-format it

07:13 <greenbagels> is there a general preference for whether i should try to first map the data to like, records and then use atdgen to turn it into json-serialized data, or just stuff it into yojson's json polymorphic variant type as im reading?

07:15 <greenbagels> because i feel like the number of matches / `Assoc [(key, val); ...] is getting kind of cumbersome, but i think this might be indicative of an issue i'd run into with my parsing rather than what libraries im using...

08:01 <Anarchos> Is there an ocaml equivalent of htonl/ntohl integer conversions for network read/write ?

12:02 <companion_cube> Just use Bytes.get_int32_{le,be}

12:02 <companion_cube> And the likes

12:02 <companion_cube> greenbagels: you want to emit a list of json, or just a single json object?

16:42 <greenbagels> companion_cube: single object

16:42 <companion_cube> Probably easier to make a big yojson object then

16:43 <companion_cube> Or jsonm stream of tokens if you need better perf?

16:45 <greenbagels> ok ill keep playing around with yojson until i find a program structure that doesnt feel awkward :~)

16:46 <greenbagels> polymorphic variants are cool

17:03 <companion_cube> Maybe use combinators? :p

17:24 <sim642> companion_cube, "your" catapult library is nice, but too bad it doesn't let me make up thread IDs (and give them custom names) to have separate tracks in Perfetto

18:07 <orbifx> sim642: can you not form our own maps?

18:12 <sim642> orbifx, hm?

18:14 <orbifx> can you not have a map: your id -> catapult id ?

18:20 <sim642> That's not really the issue. Rather the file format has a way to associate names to thread numbers, but the library API doesn't allow doing that arbitrarily

18:21 <sim642> And I want my custom names in the produced file so it'd show up in all the tooling that reads those files

18:32 <companion_cube> sim642: it should? Hmm

18:33 <companion_cube> It captures thread ids automatically

18:33 <companion_cube> And process ids. But you can provide these by hand

18:33 <sim642> yeah, but I don't want it to capture tids automatically

18:33 <sim642> And if I manually pass a tid, then it only works if that's actually the right one

18:33 <sim642> and fails an assert if it isn't

18:34 <companion_cube> Oh, interesting. What's the use case? It can be changed I rink

18:34 <companion_cube> Think

18:34 <sim642> Right now I just hacked around it by using a fake pid instead, but I cannot name it still

18:35 <sim642> The use case is that I want a separate track of nested durations

18:35 <companion_cube> Hmm there is a meta event to name it

18:35 <companion_cube> Hmm you sure you don't want async events?

18:36 <sim642> I want the normal track to have the normal tracing stuff, but then a secondary track that visualizes the stack of the program under (static) analysis

18:36 <companion_cube> Do you know how to do it if you produce the json by hand?

18:37 <companion_cube> Oh, you're saying it's just using a fake thread id. Interesting.

18:37 <sim642> Maybe? I don't know because I didn't really understand all the business with asyncs and flows

18:37 <companion_cube> Please open an issue, it can easily be accommodated :).

18:38 <companion_cube> Yeah I've used them only a little and with the chrome viewer, not perfetto

18:38 <sim642> Another thing I was wondering is the possibility of using multiple backends at once

18:38 <companion_cube> As an aside I also love Tracy with its importer.

18:38 <companion_cube> Oh?

18:38 <sim642> Like the normal json file one and another custom one that does some aggregation on the go for a landmarks style printout also

18:39 <companion_cube> Maybe with a multiplexer. But please tell, it's probably doable.

18:39 <companion_cube> Nice

18:42 <companion_cube> Producing a landmark file from the raw json might be possible, too. It's an interesting question.

18:43 <sim642> I didn't investigate it too much, but that's some generic CTF stuff that might already exist somewhere

18:43 <sim642> Especially given all the Perfetto stuff besides the UI

18:43 <sim642> And all the SQL-whatever in the UI as well

18:46 <sim642> I don't yet know for sure, what and how far I'd want to take, but I was just looking for more modern and powerful replacements for CIL's old Stats module...

18:47 <sim642> Which is like 20 years old landmarks, without any of the bells and whistles

18:47 <companion_cube> What's CIL ?

18:47 <companion_cube> But yeah, I'm open to improving things!!

18:47 <sim642> It's an ancient C parser/frontend library in ocaml

18:48 <companion_cube> Oh right.

18:49 <sim642> I also found that dune publishes the chrome-trace package, which is more low level and just for constructing the json output with no restrictions

18:50 <sim642> But it collects all the events in memory so it can output them inside a CTF object

18:55 <sim642> I'm guessing all users of that library are internal?

18:58 <companion_cube> Yep

18:59 <companion_cube> (well tbh it's mostly me)

23:23 <nvaxplus> Got a menhir question (well I suppose it's somewhat of a LR(1) question). Suppose I have a language where you can have named function arguments like python syntax (`foo(regular, named = quux)`). Naively trying to write this in menhir produces shift/reduce conflicts. Is that syntax simply not in LR(1) or am I doing it wrong? (Or does Menhir have any workarounds)

23:34 <dh`> I haven't thought about it for more than five seconds but most likely you're doing it wrong

23:35 <dh`> in general with grammars there are lots of ways to express thngs and most of them don't work, same as how with most languages there are lots of things you can type and most of them are syntax errors :-)

23:36 <nvaxplus> yeah I figure I am doing it wrong, but having a mental freeze where I can think of how to paraphrase it right

23:39 <dh`> I just hacked up a quick test and it works fine

23:40 <dh`> most likely source of trouble is that ':' appears in your expression syntax somewhere

23:42 <nvaxplus> hmm why do you say?

23:43 <dh`> because if expr | IDENT COLON expr is ambiguous the most likely reason is that IDENT COLON is a valid prefix of expr

23:43 <dh`> such that it can't disambiguate those cases

23:44 <nvaxplus> hmm I see

23:45 <nvaxplus> dh`: do you mind pastebinning what you wrote?

23:46 <dh`> yeah was about to, need to make it tidy first

23:46 <dh`> what's a good pastebin here?

23:46 <dh`> actually n/m

23:48 <nvaxplus> thank you so much! and no worries about tidiness!

23:52 <dh`> now it actually compiles :-)

23:53 <dh`> https://www.netbsd.org/~dholland/tmp/stuff.mly

23:55 <nvaxplus> dh`: thank you!!!

23:55 <dh`> yw

23:57 <dh`> grammar stuff is very highly stylized and it takes practice to be fluent with it, few people ever bother IME

23:58 <dh`> nearly every published grammar you'll find will be ambiguous and most yacc grammars in production are full of both conflicts and ad hoc disambiguation rules