<< 27-07-2020 >>

---

00:10:21	*	vicfred quit (Remote host closed the connection)
00:10:44	*	vicfred joined #nim
00:12:49	*	oddp quit (Ping timeout: 264 seconds)
00:23:05	*	fredrikhr quit (Read error: Connection reset by peer)
00:23:30	*	fredrikhr joined #nim
00:24:10	*	fredrikhr quit (Read error: Connection reset by peer)
00:24:35	*	fredrikhr joined #nim
01:05:38	*	Tlongir joined #nim
01:08:19	*	Tongir joined #nim
01:10:25	*	Tlongir quit (Ping timeout: 240 seconds)
01:19:31	*	apahl quit (Ping timeout: 272 seconds)
01:21:07	*	apahl joined #nim
01:23:03	FromDiscord	<Anuke> sent a code paste, see https://play.nim-lang.org/#ix=2spM
01:24:31	voltist	Is anyone on here the developer of NiGui?
01:31:36	FromDiscord	<slymilano> jiro4989 just posted a github actions for cross-platform build/release. it's really great!
01:31:49	FromDiscord	<Yardanico> @Anuke that's the wrong command
01:31:58	FromDiscord	<slymilano> git tag v0.0.1 and boom you got your app for all three major's easily reachable for your end user.
01:32:00	FromDiscord	<Yardanico> Replace arm with arm64
01:32:01	FromDiscord	<slymilano> shmexy
01:32:09	FromDiscord	<Yardanico> arm is armv7
01:33:03	FromDiscord	<Yardanico> arm64 is aarch64 and is armv8a
01:33:37	FromDiscord	<Yardanico> @voltist yes, he's there sometimes, but it's better to open an issue in NiGui to ask something about it
01:33:46	FromDiscord	<Yardanico> !seen TrustableCode
01:33:47	disbot	TrustableCode never seen.
01:34:00	FromDiscord	<Yardanico> !seen trustable-code
01:34:01	disbot	trustable-code never seen.
01:36:26	FromDiscord	<Yardanico> Hmmm
01:38:19	bung	https://github.com/nim-lang/Nim/issues/5768 this should close
01:38:21	disbot	➥ db_sqlite can't correctly quote binary data ; snippet at 12https://play.nim-lang.org/#ix=2spS
01:41:27	bung	I've read all stdlib,js lablled issues till now
02:03:37	FromDiscord	<Yardanico> Why it should be closed? I don't understand
02:10:01	bung	https://github.com/nim-lang/Nim/pull/14408/files#diff-0285f3281e8cf19b0275e85bc908d119R774
02:10:02	disbot	➥ add bindParams to db_sqlite
02:10:11	*	muffindrake quit (Ping timeout: 272 seconds)
02:10:38	bung	db_sqlite now have bind params api, can bind binary data null etc.
02:11:27	FromDiscord	<impbox> nice
02:11:45	*	muffindrake joined #nim
02:25:48	*	outtabwz quit (Quit: leaving)
02:38:54	*	fredrikhr is now known as couven92
02:42:23	FromDiscord	<Varriount> impbox: Who did the music for smalltrek? It's nice.
02:42:38	FromDiscord	<Varriount> Reminds me of one of my favorite schmups
02:48:20	*	Cthalupa quit (Ping timeout: 244 seconds)
02:48:47	*	Cthalupa joined #nim
02:52:05	*	lritter quit (Ping timeout: 240 seconds)
02:52:58	*	lritter joined #nim
02:58:47	*	waleee-cl quit (Quit: Connection closed for inactivity)
03:03:31	*	arecacea1 quit (Read error: Connection reset by peer)
03:03:34	*	B4s1l3 joined #nim
03:04:08	*	arecacea1 joined #nim
03:06:23	FromDiscord	<Varriount> Reminds me music by [Slimegirls](https://slimegirls.bandcamp.com/)
03:06:42	FromDiscord	<Varriount> (edit) '[Slimegirls](https://slimegirls.bandcamp.com/)' => 'Slimegirls (https://slimegirls.bandcamp.com/)'
03:09:30	*	FromDiscord quit (Remote host closed the connection)
03:09:45	*	FromDiscord joined #nim
03:15:20	*	Tlanger joined #nim
03:18:17	*	Tongir quit (Ping timeout: 265 seconds)
03:19:20	FromDiscord	<impbox> @Varriount i did
03:19:42	FromDiscord	<impbox> it was very rushed on my gameboy
03:20:32	FromDiscord	<impbox> i see the resemblance, very gameboy sound
03:23:42	FromDiscord	<impbox> https://soundcloud.com/luna-and-the-spacebaes/spatial-anomaly a song me and my partner created out of the smalltrek music
03:23:50	shashlick	shared data across threads again - when is this going to be solved
03:24:24	FromDiscord	<Varriount> shashlick: When we finally come up with a solution for world hunger and global peace.
03:24:36	FromDiscord	<impbox> you can't share data across threads?
03:25:08	FromDiscord	<impbox> pretty sure i've been doing that
03:25:34	FromDiscord	<Varriount> Not without either using the Boehm GC, or pointers.
03:25:47	shashlick	how are you - cannot use any nim types
03:25:50	FromDiscord	<Varriount> Or copying data
03:26:07	FromDiscord	<impbox> hmm maybe i'm using boehm, is that the default?
03:26:21	shashlick	nope
03:26:37	*	B4s1l3 is now known as opDispatch
03:26:45	FromDiscord	<impbox> is this a recent change you're talking about?
03:27:06	shashlick	this is why arc/orc exist
03:27:08	FromDiscord	<impbox> i haven't played with threads lately, but nimsynth was sharing global data between threads
03:27:25	shashlick	cause the refc gc doesn't support it
03:27:28	FromDiscord	<Varriount> impbox: What method were you using to share data, channels?
03:27:36	FromDiscord	<impbox> nah, just accessing globals
03:27:39	FromDiscord	<impbox> with locks
03:27:58	FromDiscord	<Varriount> Did the globals have any reference types in them?
03:28:43	FromDiscord	<impbox> i'll check the code so i'm not giving misleading data
03:28:49	FromDiscord	<Varriount> Where you get into trouble is if you store a reference from thread A in a global, then copy it into thread B
03:29:09	FromDiscord	<impbox> hmm is it copying? i thought both threads could access the same data
03:29:13	FromDiscord	<Varriount> (copy the reference, not the data the reference points to)
03:29:45	shashlick	i cannot opt for boehm here, it works great for this
03:29:52	shashlick	have to use pure nim stdlib
03:30:05	shashlick	and cannot assume boehm is installed
03:30:58	FromDiscord	<Varriount> To put it another way, a thread cannot handle references (either directly, or indirectly through an object) that were not created in that thread
03:31:18	FromDiscord	<Elegant Beef> The default GC has seperate heaps per thread, right?
03:31:28	FromDiscord	<Varriount> Yes
03:32:53	shashlick	one option is to have one thread that manages the data and use channels back and forth with that thread
03:33:08	shashlick	just a mess though - just trying to parallelize some code
03:33:36	FromDiscord	<Varriount> shashlick: Launch separate processes. >:D
03:34:19	shashlick	doesn't help cause they still need to coordinate
03:34:37	FromDiscord	<Varriount> Pipes?
03:35:00	shashlick	basically i'm experimenting with parallelizing nimble
03:35:33	FromDiscord	<Varriount> Parallel downloads?
03:35:41	shashlick	ya and processing dependencies
03:35:51	shashlick	but cannot install the same dependency multiple times
03:36:07	shashlick	has to be some coordination
03:36:13	FromDiscord	<Varriount> Isn't that the kind of thing channels are good at?
03:36:25	shashlick	then i'll need a thread keeping track of this
03:36:33	FromDiscord	<Varriount> Have each thread push/pull jobs onto/from the channel
03:37:02	FromDiscord	<impbox> can you just use async instead of threads?
03:37:17	shashlick	i would have but am spawning processes which need to be parallelized
03:37:33	shashlick	for that i'll need asynctools which isn't stdlib
03:37:42	disruptek	so put it in.
03:38:01	shashlick	i might just ditch this experiment
03:38:33	FromDiscord	<impbox> is nimble such a bottleneck that it needs optimising?
03:38:36	shashlick	doubt i can improve performance very drastically without major rework
03:39:57	shashlick	let's just call it a fun experiment that originated from @PMunch's https://github.com/nim-lang/nimble/issues/722
03:39:58	disbot	➥ Concurrent downloads of packages
03:40:04	*	fredrik92 joined #nim
03:40:31	FromDiscord	<impbox> mmm concurrent downloads sounds like it shouldn't require threads
03:40:51	FromDiscord	<impbox> spawn a bunch of processes, wait for them all to finish
03:41:09	*	jeko quit (Quit: Leaving)
03:42:13	shashlick	one point worth consideration is that most packages only have 2-3 deps so the tree could get wide but you don't know child deps unless you download the parent
03:43:04	FromDiscord	<impbox> could be stored as metadata in the ... whatever it's called that nimble talks to
03:43:50	*	couven92 quit (Ping timeout: 260 seconds)
03:43:58	shashlick	the package repo only includes urls to repos, not deps and version info
03:44:06	FromDiscord	<impbox> so far!
03:44:16	shashlick	it isn't a db - just a json file
03:45:32	shashlick	performance is a distraction after all, let's just call it a useless Sunday and move on
03:45:44	shashlick	some day data sharing across threads will be easy
03:45:46	FromDiscord	<impbox> but even without threads, check when one process finishes and check if new deps are added by it
03:45:57	FromDiscord	<impbox> then spawn processes for those deps
03:46:03	FromDiscord	<impbox> and keep a queue of processes
03:48:19	shashlick	some day, will consider this again, for now need to focus on bugs again
03:50:31	shashlick	someone pick a defect
03:51:05	FromDiscord	<impbox> better error messages for nimble scripts
03:51:18	shashlick	that should be improved with recent changes
03:51:26	FromDiscord	<impbox> \o/
04:06:01	*	supakeen quit (Quit: WeeChat 2.8)
04:06:39	*	supakeen joined #nim
04:09:05	*	altarrel quit (Quit: Konversation terminated!)
04:17:06	*	narimiran joined #nim
04:17:54	disruptek	!repo cps
04:17:55	disbot	https://github.com/disruptek/cps -- 9cps: 11Continuation-Passing Style for Nim 🔗 15 15⭐ 0🍴 7& 1 more...
04:18:29	disruptek	bad news:
04:18:35	disruptek	tomorrow is shower sunday.
04:19:01	FromGitter	<kennymalac> Hey, out of curiousity, is it possible to override default keywords such as if, else, and =
04:19:38	disruptek	!eval var `if` = "woot"
04:19:40	NimBot	<no output>
04:21:13	FromDiscord	<impbox> kennymalac you can use them as non keywords by quoting with backticks i think
04:21:17	FromDiscord	<Elegant Beef> you can strop variables
04:21:17	FromDiscord	<Elegant Beef> Yea
04:21:19	FromDiscord	<impbox> not the same as overriding them
04:21:37	FromGitter	<kennymalac> I'm trying to compare nim to a lisper
04:22:10	FromGitter	<kennymalac> nim to Common lisp to a lisper*
04:22:18	disruptek	!repo belamenso/v
04:22:19	disbot	https://github.com/belamenso/v -- 9v: 11Write Nim only with 'v' 15 24⭐ 3🍴
04:25:40	FromDiscord	<Rika> *flashbacks to the last lisper who was here*
04:28:47	skrylar[m]	that is horrifying
04:45:38	*	opDispatch left #nim ("Konversation terminated!")
04:47:54	*	fredrikhr2 joined #nim
04:48:07	*	fredrikhr2 is now known as fredrikhr
04:51:52	*	fredrik92 quit (Quit: Client Disconnecting)
04:58:07	*	fredrikhr quit (Quit: The Lounge - https://thelounge.chat)
05:02:07	*	fredrikhr joined #nim
05:30:02	Zevv	disruptek: so, how many nim bugs have you summoned over the last weeks again?
05:55:01	*	maier joined #nim
06:09:38	*	Vladar joined #nim
06:29:13	*	jjido joined #nim
06:38:40	*	aurelius joined #nim
06:44:14	*	hoijui joined #nim
06:46:19	*	solitudesf joined #nim
06:55:00	*	lritter quit (Quit: Leaving)
07:11:00	Zevv	who wants a fun little puzzle
07:11:01	Zevv	http://ix.io/2sqi
07:11:20	Zevv	winners gets a large beer at next fosdem
07:17:36	*	nikita` joined #nim
07:18:09	FromDiscord	<Rika> what if we dont drink beer
07:20:55	FromDiscord	<Rika> Zevv: checking the two's "bits" by casting them into an int64 shows that theyre exactly the same in memory, it seems
07:21:26	Zevv	A kebab or falafel good for you?
07:21:30	FromDiscord	<Rika> so it seems that its impossible to distinguish them
07:21:46	Zevv	right so
07:21:53	FromDiscord	<Rika> uh i dont live in europe so its extremely unlikely you'll ever meet me
07:22:18	Zevv	Who cares. you have local deliveries I bet :)
07:22:21	*	Tongir joined #nim
07:23:13	FromDiscord	<Rika> wait, let me redo my "observation"
07:23:19	Zevv	haha
07:23:41	Zevv	the fun thin is: disruptek has a case where there is different behaviour depending on cast vs conversion
07:23:50	Zevv	but it's pretty convoluted
07:24:25	*	Tlanger quit (Ping timeout: 240 seconds)
07:24:31	FromDiscord	<Rika> so maybe its not minimized correctly
07:25:00	Zevv	sure, but then again. His code does A with casting and B when converting.
07:25:36	*	jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
07:26:54	FromDiscord	<Rika> then maybe its not just "when a type inherits another", maybe theres another thing in combination that causes the issue
07:27:03	Zevv	there must be.
07:27:27	Zevv	I think there's some funny AST going on in all the macros in the code
07:27:32	Zevv	but I don't understand that code :/
07:57:02	*	hoijui quit (Ping timeout: 260 seconds)
08:11:02	FromDiscord	<Varriount> Zevv: Could the behavior be C compiler related?
08:11:08	Araq	shashlick, awesome work on Nimble! Thank you!
08:11:10	Zevv	I don't know. There *is* different codegen
08:11:17	*	Araq is going through his github related emails
08:18:17	*	dzamo joined #nim
08:21:13	*	dzamo quit (Client Quit)
08:21:32	*	dzamo joined #nim
08:25:46	*	jjido joined #nim
08:26:22	Araq	Zevv, don't understand your snippet
08:26:34	Araq	if you explain it once again, I might help out
08:29:02	Zevv	well, disruptek "fixed" something in CPS yesterday by casting between derived object types
08:29:05	Zevv	instead of converting
08:29:21	Zevv	I don't understand his CPs code anymore, but my point was that if he needed to cast, something is wrong
08:29:35	Zevv	because everthing should be perfectly fine, typewise, after CPS transformations
08:29:48	Zevv	but he seems to have a case where his bug is "fixed" by casting instead of converting
08:29:56	Zevv	There is an actual difference in the codegen
08:30:13	Zevv	but I don't understand what the different behaviour could be between a casted and converted object
08:30:23	Zevv	because the end result is exactly the same, it is only a different path of going there
08:30:52	Zevv	But I strongly oppose between putting in any kind of casting in the CPS code, because we basically lose the help of the compiler there if we mess up
08:31:27	Zevv	so this snippet is just a way of showing two ways to get to the same result, and my question was if there is any distinguishable difference between these two objects - one casted and one converted
08:31:34	Zevv	because the C-code *is* different
08:31:44	Zevv	does that make sense?
08:34:23	FromDiscord	<Varriount> Zevv: Wouldn't the conversion code have conversion checks?
08:34:38	FromDiscord	<Varriount> (assuming you're not turning them off)
08:35:14	Araq	Zevv, I agree with you
08:35:31	Araq	and 'cast' shouldn't be used
08:35:52	FromDiscord	<impbox> just so i can follow along, what's CPS?
08:36:05	Araq	we have different code paths in the Nim compiler though because type conversions are checked at runtime
08:37:06	Araq	and I hunted a maybe related codegen bug for the last days
08:38:06	Zevv	that would make sense.
08:38:20	Zevv	The problem is that this bug is triggered after all this CPS juggling with the AST
08:38:46	Zevv	so it might also be that his problem is caused by funny AST which is not right but also not quite wrong
08:39:01	Zevv	I'm not sure how strict the compiler is on AST that it does not generate from the Nim parser
08:39:20	Zevv	I've seen the occasional 'ill formed' message, but I bet there's a million ways to get the C-gen to mess up with custom AST
08:39:34	Araq	it's strict but I suppose e.g. nkConv is usually not generated by a macro
08:39:47	Araq	and yeah, getting it to "strict" was a pita :-)
08:40:17	Zevv	makes perfect sense.
08:40:42	Zevv	Problem is that I do not understand the underlying stuff of disruptek issue. I do understand that I am right and he is wrong, but I needs some time to realize that
08:40:55	Zevv	s/he needs/
08:43:06	FromDiscord	<Varriount> @impbox I'll PM you, to avoid cluttering up the chat
08:45:46	*	xet7 quit (Ping timeout: 260 seconds)
08:47:05	*	aurelius quit ()
08:47:34	*	xet7 joined #nim
08:48:14	*	aurelius joined #nim
09:07:37	*	oddp joined #nim
09:14:12	FromGitter	<alehander92> ok
09:15:22	*	Tlanger joined #nim
09:18:13	*	Tongir quit (Ping timeout: 264 seconds)
09:20:23	*	ofelas quit (Ping timeout: 258 seconds)
09:40:56	*	dzamo quit (Quit: dzamo)
09:41:18	*	dzamo joined #nim
09:50:33	Zevv	Varriount: sure, but I turned everything off with -d:danger
09:54:49	*	Tongir joined #nim
09:56:59	*	Tlanger quit (Ping timeout: 240 seconds)
10:09:47	*	jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
10:14:12	FromDiscord	<Varriount> Araq: Regarding closures (and closure iterators) do you know why they are slower than regular function calls? Is it the accesses to the closure environment?
10:18:41	Araq	well it's a closure call, closure calls introduce an 'if' that good optimizers can optmize out but when I checked years ago they didn't
10:20:01	Araq	the environment and heap allocations don't help either but look
10:20:34	Araq	the real costs here come from the Future + callbacks + closure iterator interactions
10:21:07	Araq	if you could only use Futures, or only callbacks or only closure iterators it would be better
10:25:25	FromGitter	<alehander92> can you inline many calls
10:25:29	FromGitter	<alehander92> to just goto-s
10:25:39	FromGitter	<alehander92> like, can one write a simple compiler for no-async code
10:25:45	FromGitter	<alehander92> that just generates a single big call
10:25:56	FromGitter	<alehander92> without creating new frames etc
10:26:21	FromGitter	<alehander92> and just rearranging everything into different parts of the code (and limiting stack variables )
10:26:27	Araq	that's how my Lego Mindstorm Not quite C implementation worked
10:26:38	Araq	it's terrible and you lose recursion but it works
10:26:48	FromGitter	<alehander92> oh yeah recursion
10:26:54	Araq	it was a valid strategy for Fortran code before they added recursion to Fortran
10:27:04	FromGitter	<alehander92> interesting
10:27:13	FromGitter	<alehander92> but does it help somehow performance
10:27:32	FromGitter	<alehander92> or is it just a cool-that-it-can be done
10:27:40	FromGitter	<alehander92> thing
10:28:58	FromGitter	<alehander92> like, frames are just there, so it's not really a matter of creating them, and moving vars to heap is probably not cool
10:36:50	*	outtabwz joined #nim
10:39:57	FromDiscord	<Clyybber> Araq: Is there a way I can access the template scope of a template generated symbol?
10:40:26	FromDiscord	<Clyybber> In semProcAux
10:43:33	Zevv	wait what
10:43:35	Zevv	did you make NQC?
10:46:13	FromDiscord	<Yardanico> Implementation for Lego Mindstorm it seems
10:48:49	Zevv	yeah man I grew up with that stuff
10:49:13	Zevv	mindstorms software sucked big time, but luckily there was NQC
10:51:44	FromDiscord	<Yardanico> Oh nice so the closure leak was a real one
10:52:09	FromDiscord	<Yardanico> That's nice to know, I hope we won't have other leaks with async after that one is fixed
10:52:33	FromDiscord	<Yardanico> 1 line fix :D
10:56:35	FromDiscord	<Clyybber> that 1 line fix is only temporary afaict
10:56:54	FromDiscord	<Clyybber> because in general we still want to have the isDecl firstwrite optimization
10:57:37	FromDiscord	<Yardanico> Oh
11:00:03	FromDiscord	<Yardanico> Also @Clyybber is it fine to merge #15085 ?
11:00:04	disbot	https://github.com/nim-lang/Nim/pull/15085 -- 3Add test-cases for #12576 and #12523
11:00:20	FromDiscord	<Clyybber> Oh, I was just looking at it
11:02:03	*	NimBot joined #nim
11:03:02	FromDiscord	<Yardanico> And yeah, I can probably do some comment renamings in the test suite to our new scheme
11:03:11	FromDiscord	<Yardanico> I mean the # bug #12345 stuff
11:03:12	disbot	https://github.com/nim-lang/Nim/issues/12345 -- 3Confusing error with static parameters ; snippet at 12https://play.nim-lang.org/#ix=26CQ
11:03:18	FromDiscord	<Clyybber> not needed IMO, but if you want to go ahead :)
11:04:35	Araq	Zevv, ah no, that was poorly worded, it was "my Mindstorm"
11:04:43	Araq	but I didn't write the software
11:05:31	Araq	<Yardanico> that fix fixes your reduced test case
11:05:49	Araq	but the async leak remains, well it got a little better
11:06:48	Araq	a different reduction would be sweet ;-)
11:07:03	FromDiscord	<Yardanico> Oh okay I'll see
11:08:49	Araq	Clyybber: the 'isFirstWrite' rule is mostly unaffected
11:09:14	Araq	only closures were affected
11:10:45	Araq	Clyybber: alternatively you can work on making forward declarations obsolete
11:11:22	Araq	they make things more complex with almost no known benefits
11:15:08	FromDiscord	<Clyybber> yeah
11:15:38	FromDiscord	<Clyybber> Araq: But you changed the only call that sets isDecl to true in var/let sections
11:16:15	Araq	but for n.kind != nkSym which means it's a "lifted" variable
11:16:38	Araq	it's var env.v = 3 which should really be evn.v = 3
11:16:45	FromDiscord	<Clyybber> oh, I missed that part
11:16:49	FromDiscord	<Clyybber> right, makes sense
11:17:32	*	krux02 joined #nim
11:19:41	*	fredrikhr quit (Read error: Connection reset by peer)
11:20:48	Araq	alehander92, on a modern CPU it's not an optimization anyway. But thinking about it, it's how GPUs work
11:20:59	Araq	afaik
11:21:16	FromDiscord	<Yardanico> Also I forgot about -d:nimAllocStats , guess reducing would be easier with it
11:21:25	*	fredrikhr joined #nim
11:27:27	*	dulsi quit (Quit: Leaving)
11:29:49	*	fredrikhr quit (Read error: Connection reset by peer)
11:30:30	*	fredrikhr joined #nim
11:30:52	*	dulsi joined #nim
11:33:18	*	hoijui joined #nim
11:39:27	*	oriba joined #nim
11:45:07	*	fredrikhr quit (Read error: Connection reset by peer)
11:45:31	*	fredrikhr joined #nim
11:50:52	*	waleee-cl joined #nim
11:59:03	icyphox	any good examples for xmlparser?
12:01:03	FromDiscord	<Yardanico> Idk, it's not that hard
12:01:21	FromDiscord	<Yardanico> E.g. https://github.com/Yardanico/nim-snippets/blob/master/logcheck/main.nim#L82
12:01:51	FromDiscord	<Yardanico> https://github.com/Yardanico/nim-snippets/blob/master/tg_export_parse.nim#L6
12:02:03	icyphox	yeah i was wondering what to do with the XmlNode object
12:02:09	icyphox	found some procs i can use in xmltree
12:02:32	FromDiscord	<Yardanico> nimquery is a good lib
12:02:43	FromDiscord	<Yardanico> For using CSS selectors to parse html
12:02:56	icyphox	ah no, this is pure XML
12:03:01	icyphox	xmpp stanzas
12:03:32	FromDiscord	<Yardanico> It'll still work
12:03:50	FromDiscord	<Yardanico> Current html parser is built on top of the XML one
12:04:06	FromDiscord	<Yardanico> https://github.com/GULPF/nimquery
12:04:17	FromDiscord	<Yardanico> parseHtml returns a XmlNode
12:05:08	icyphox	nice
12:05:11	icyphox	i'll check this out
12:06:02	*	supakeen quit (Quit: WeeChat 2.8)
12:06:39	*	supakeen joined #nim
12:13:00	FromDiscord	<dom96> > if you could only use Futures, or only callbacks or only closure iterators it would be better↵@Araq[IRC]#0000 what makes you think you can't?
12:13:28	Araq	async's design uses all three
12:13:37	FromDiscord	<dom96> > Oh nice so the closure leak was a real one↵@Yardanico what's the context on this?
12:13:53	Yardanico	https://github.com/nim-lang/Nim/issues/15076 the last comment repro was fixed
12:13:54	disbot	➥ Memory leaks with async (closure iterators?) under ORC ; snippet at 12https://play.nim-lang.org/#ix=2snX
12:13:55	Araq	context is --gc:orc + async
12:14:11	Araq	and it's still leaking, I'm looking into the problem right now
12:14:16	FromDiscord	<dom96> Araq: async /await/'s design
12:14:26	Araq	(but it leaks less now, yay)
12:14:46	FromDiscord	<dom96> Nothing stops you from using futures on their own, except the mess that brings to your code 🙂
12:15:12	FromDiscord	<dom96> but maybe there is a way to reduce that mess by creating an async await macro which doesn't use closure iterators
12:15:40	*	fredrikhr quit (Read error: Connection reset by peer)
12:16:24	Araq	yeah but
12:18:08	Araq	in the end it doesn't matter much. I prefer to do things differently.
12:18:41	Yardanico	"(allocCount: 600092, deallocCount: 499967)" oh i see
12:18:44	Yardanico	from my first example hm
12:18:57	*	fredrikhr joined #nim
12:19:47	Yardanico	btw I was labeling some issues yesterday, we still have ~300 unlabeled ones (although some of them are hard to properly label)
12:19:59	Yardanico	i mean open unlabeled issues
12:20:15	FromDiscord	<dom96> Araq: that's a shame, because async is very stable and used in production by many.
12:21:52	Araq	dom96: well my way has the benefit that we might get rid of "closure iterators" entirely
12:22:03	Araq	and end up with a smaller Nim
12:23:21	Araq	with your proposed way all we get is refactorings for our async.
12:24:36	*	fredrikhr quit (Read error: Connection reset by peer)
12:25:18	*	fredrikhr joined #nim
12:25:49	Araq	or maybe in the end we understand closure iterators better and can write a decent spec and eliminate edge cases
12:26:48	FromDiscord	<dom96> > with your proposed way all we get is refactorings for our async.↵I guess, but you said that only using Futures would be an improvement. I assume that's a win in and of itself. Even if we get there by refactoring async.
12:34:51	Yardanico	ok I minimized the original leak to only importing asyncfutures :P
12:35:01	Yardanico	"(allocCount: 30020, deallocCount: 25018)" for 5k iterations
12:36:28	*	sagax quit (Remote host closed the connection)
12:37:24	Yardanico	ah wait
12:37:24	*	aurelius quit (Remote host closed the connection)
12:38:13	Yardanico	Araq: did you test the original snippet (the first one from the issue) so you said there's still a leak?
12:38:21	Yardanico	because it seems like this leak comes from debugging info of futures
12:38:56	Araq	Yardanico: yeah, valgrind suggests that
12:39:13	Yardanico	if you compile with -d:release you get "(allocCount: 500090, deallocCount: 499967)", although this might still mean a leak
12:39:22	Yardanico	and it does, valgrind still shows a bit of leaks
12:41:50	silvernode[m]	Tonight was my last night off work, I had 10 days off. I got about 6 more hours to kill before sleep time and need to motivate myself to work on code. -_-
12:43:38	Yardanico	oh lol
12:43:51	Yardanico	it might have to do something with Future[T] inheriting from FutureBase
12:44:03	Yardanico	if I replace every Future[void] with FutureBase (everywhere in the code), that leak goes away
12:44:29	*	sagax joined #nim
12:45:06	*	jjido joined #nim
12:45:50	Yardanico	leaks in - https://gist.github.com/Yardanico/694ae199eac7b864bc3fb822ae401d15, doesn't leak with - https://gist.github.com/Yardanico/a58949aea5653c4163c3f91b60a2b657, whole diff - https://gist.github.com/Yardanico/a3e1ba90f9d0e34e5edac07b88b59153
12:46:21	FromDiscord	<XxDiCaprioxX> Anybody here who used Discordnim?
12:46:30	Yardanico	you should use dimscord instead :)
12:46:32	Oddmonger	are there differents ways of writing a cast ?
12:47:16	Yardanico	Oddmonger: not really, there's only one way of writing a real "cast" in nim
12:47:27	Yardanico	@XxDiCaprioxX discordnim is unmaintained really
12:47:28	Oddmonger	ie i want to cast a cstring to a string, so i did (naively): print_string( (string) foo() )
12:47:29	FromDiscord	<XxDiCaprioxX> Okay, but is that a good library?
12:47:29	Yardanico	dimscord OTOH is
12:47:40	FromDiscord	<XxDiCaprioxX> I mean dimscord
12:47:42	Yardanico	@XxDiCaprioxX well the current IRC bridge uses it
12:47:45	Yardanico	the author is active here
12:47:54	Yardanico	Oddmonger: well that'll work
12:47:57	Yardanico	but it's not a "cast"
12:47:59	Yardanico	it's a conversion
12:48:03	Yardanico	conversions are safe, casts are not
12:48:13	Oddmonger	ah
12:48:14	Yardanico	and what you did is possible due to command syntax
12:48:16	Yardanico	e.g.
12:48:21	Yardanico	!eval (echo)5
12:48:23	NimBot	5
12:48:50	FromDiscord	<XxDiCaprioxX> Okay thank you @Yardanico[IRC]#0000
12:48:59	Oddmonger	whooo
12:49:00	Oddmonger	ok
12:49:10	Oddmonger	thank you
12:49:11	Yardanico	the more readable way is
12:49:21	Yardanico	print_string(string(foo()))
12:49:24	Yardanico	or foo().string
12:49:26	Yardanico	or string foo()
12:49:32	Oddmonger	ah
12:49:34	Yardanico	UFCS
12:49:43	Oddmonger	i did: cast[string](foo())
12:49:47	Yardanico	don't do that
12:49:59	Yardanico	there's a $ operator for cstring defined which converts it to a string
12:50:02	Araq	do not 'cast'!
12:50:17	Araq	read the tutorials once again please
12:50:41	Oddmonger	yes i think i need to
12:51:13	Oddmonger	usually i read once, understand everything, write something trickier than «hello world» and read it again
12:51:47	FromDiscord	<dom96> I've got a project on my ideas list to rate Nim packages in terms of how many `cast`, `ptr`, `pointer`, `addr` etc. are used in the code. Avoid these features as much as you can.
12:52:02	FromDiscord	<Rika> Cast only needed if you want to do super dumb shit like I do lol
12:52:06	Yardanico	I can do that I guess @dom96
12:52:12	Yardanico	because I already have the code for pkgraph
12:52:17	Yardanico	to clone all nimble packages :D
12:52:19	FromDiscord	<Rika> @dom96 so basically a code quality analyzer?
12:52:30	FromDiscord	<Rika> Or rather, one component to it
12:52:31	Yardanico	need to clone them again though since I had to delete them
12:52:43	Yardanico	also Araq btw I parsed all #nim IRC logs
12:52:59	Yardanico	from when nimbot started working
12:53:02	Yardanico	!ping
12:53:02	NimBot	pong
12:53:02	FromDiscord	<dom96> I mean if you want go for it
12:53:05	Yardanico	!lag
12:53:05	NimBot	19ms between me and the server.
12:53:09	Yardanico	fast
12:53:15	FromDiscord	<dom96> But I wasn't thinking about doing it for all packages
12:53:24	FromDiscord	<dom96> it's something that you should be able to run on specific packages
12:53:33	FromDiscord	<dom96> and it was inspired by a Rust program that looks for `unsafe`
12:53:37	FromDiscord	<dom96> can't remember it's name though
12:53:45	FromDiscord	<Rika> runsafe
12:53:49	FromDiscord	<Rika> I'd name a program that
12:54:02	FromDiscord	<Rika> Lol (I don't know the name, I'm just guessing)
12:54:05	*	Kaivo joined #nim
12:56:54	FromDiscord	<Varriount> Wow, lots of PRs popping up today
12:57:15	FromDiscord	<Clyybber> Araq: Hmm, WDYT about attaching scopes to symbols?
12:57:29	FromDiscord	<Clyybber> This would make template scope mixing perhaps easier
12:57:33	Araq	I don't like it
12:57:57	Araq	scopes are mutable
12:58:12	Araq	and poorly understood
12:58:53	FromDiscord	<Varriount> Poorly understood from an implementation perspective, or a conceptual perspective?
13:00:55	Yardanico	Araq: well I seem to have minimized it to less than 100 loc
13:01:12	Yardanico	"(allocCount: 30020, deallocCount: 25018)"
13:01:14	Yardanico	https://gist.github.com/Yardanico/817a7a7c71344ddb8f5aa1374ee466d0
13:01:35	Yardanico	if you just do Future and not Future[T] - leak disappears
13:01:41	Yardanico	maybe because then the type is just aliased
13:01:55	Yardanico	or if you use FutureBase directly
13:02:01	FromDiscord	<Clyybber> Araq: Should `template a: untyped = proc p() {.gensym.}; proc p() {.gensym.} = discard` work?
13:03:23	Yardanico	seems to be related to " stackTrace: seq[StackTraceEntry]" directly
13:03:34	Yardanico	if I remove that field and it's initialization - there's "almost" no leak
13:03:38	Yardanico	"(allocCount: 25018, deallocCount: 25017)"
13:04:17	Yardanico	yeah seems like it doesn't dealloc this stackTrace field
13:04:45	Araq	Clyybber: good question
13:05:20	Araq	so ...
13:05:41	Araq	Telegram tells me I can only use it until September the 1st unless I update my OS
13:05:46	Yardanico	wat
13:06:05	Yardanico	huh, do they want to not support win7 anymore?
13:06:08	Araq	problem is updating my OS (X) is impossible because my hard disk lacks the space for an update
13:06:12	Yardanico	oh osx
13:06:21	Yardanico	ah sorry you're on win10 right
13:06:37	Yardanico	there are two Telegram apps for macOS - one is telegram desktop, other is "native" one
13:06:38	Araq	and I cannot attach an external disk either because OSX doesn't support NTFS
13:06:52	Araq	any solutions?
13:07:17	Yardanico	which telegram do you use?
13:07:29	Yardanico	there's https://apps.apple.com/us/app/telegram/id747648890 which is the native telegram I was talking about
13:07:35	Yardanico	"Compatibility
13:07:35	Yardanico	OS X 10.11 or later, 64-bit processor"
13:08:16	Araq	I'll probably simply buy a new laptop tbh
13:08:57	Araq	don't know a single feature I'll miss from mac. certainly not the broken keyboard and the lack of games
13:09:04	Yardanico	hehe
13:09:38	FromDiscord	<Rika> Perhaps the trackpad
13:10:11	Araq	ah damn, will miss valgrind
13:10:30	Araq	maybe the sanitizers have become good enough though
13:12:16	Yardanico	well if I try to run that leak example with clang's memory sanitizer (with orc and -d:useMalloc), it exits on "use of uninitialized value" in orc.nim
13:12:23	Yardanico	in collectWhite
13:12:25	Yardanico	:D
13:12:39	FromDiscord	<dom96> valgrind works on Linux right?
13:12:43	Yardanico	not on windows
13:12:52	Yardanico	and 4raq doesn't like linux a lot :P
13:12:53	federico3	dom96: yes
13:13:04	FromDiscord	<dom96> Araq can use WSL
13:13:07	Yardanico	ah right
13:13:09	Yardanico	that's a thing now
13:13:47	FromDiscord	<Clyybber> I doubt it works with valgrind though
13:13:52	Yardanico	it does
13:14:01	Yardanico	wsl2 virtualises the whole kernel with HyperV
13:14:02	FromDiscord	<Clyybber> hardware watchpoints too?
13:14:21	Yardanico	not sure about that :P
13:14:23	Araq	Yardanico: do we use uninited data in collectWhite?
13:14:36	Yardanico	Araq: well clang errors on line 244 in orc.nim
13:14:43	Yardanico	"if t.color == colWhite and (t.rc and isCycleCandidate) == 0:"
13:15:55	Oddmonger	is it possible to have 0 padding with strformat ? Like {foo:03.2f} (which compiles but does not add zeros before)
13:15:56	FromDiscord	<Clyybber> gensymmed but untyped (as in no symbols but idents) doesn't really make sense does it?
13:16:00	Yardanico	maybe the stack trace is wrong
13:16:48	Oddmonger	strangely in the sample of strformat page, it seems to be the right way to do
13:17:19	FromDiscord	<Clyybber> Araq: Do we support such stuff? Gensymmed untyped idents?
13:17:27	Oddmonger	well, never mind. > is better
13:18:05	Araq	Clyybber: yeah
13:18:25	Araq	in fact, gensymed symbols usually start with "untyped" (sym.typ == nil)
13:18:43	Araq	as type checking is done after symbol lookup
13:19:36	FromDiscord	<Clyybber> Oh, yeah, I meant to say: Do we support gensymmed, non-symbol-lookuped idents
13:19:49	FromDiscord	<Clyybber> If that makes any sense :p
13:20:07	Araq	no, idents cannot be gensymed
13:20:13	FromDiscord	<Clyybber> Ok, good
13:20:43	Araq	Yardanico: so what's the best test case that leaks?
13:20:54	Yardanico	Araq: the one I posted in the issue
13:21:00	Yardanico	it's the one related to debug stack trace info
13:21:09	Yardanico	https://github.com/nim-lang/Nim/issues/15076#issuecomment-664382380
13:21:10	disbot	➥ Memory leaks with async (closure iterators?) under ORC ; snippet at 12https://play.nim-lang.org/#ix=2snX
13:21:39	Yardanico	if i rremove stack trace lines (the field and it's initialization) - there's basically no leak
13:22:12	Yardanico	also it only leaks with object inheritance, if you directly use FutureBase - it doesn't leak
13:24:03	*	hoijui quit (Ping timeout: 272 seconds)
13:25:56	Yardanico	i tried removing one of the closure iterators (or even just the closure proc), but they seem to be needed
13:26:45	Yardanico	but at least even now async leaks much, much less with orc (with -d:release or -d:danger)
13:28:13	*	pietroppeter joined #nim
13:28:34	Yardanico	with 100mil iterations of the first snippet in the issue (the one with async macro) with orc and danger - "(allocCount: 500080010, deallocCount: 499960007)"
13:29:05	Yardanico	btw, about getAllocStats - should allocCount and deallocCount always match for ARC, and always match for ORC after GC_fullCollect() ?
13:31:38	Araq	not yet as we don't free the gDisp variable
13:31:50	Yardanico	oh
13:32:00	Yardanico	I mean for cases other than async :)
13:32:15	Yardanico	and when there's no cycles (for arc)
13:32:34	Araq	yes but in general you need to beware that globals can keep things alive
13:32:45	Yardanico	ah right
13:33:18	FromDiscord	<dom96> what do these alloc/dealloc numbers look like for the default GC?
13:33:26	Yardanico	it doesn't support them I think
13:34:49	FromDiscord	<dom96> hrm, would it be difficult to introduce? Might be useful to see if A) the default GC also leaks B) the globals are to blame.
13:35:42	Araq	dom96: we do support -d:nimTypeNames and listing the alive objects with the old GCs
13:35:59	Araq	and we know it's quite free of leaks
13:36:26	FromDiscord	<dom96> sure, I know, I used that to find some leaks myself
13:36:42	Araq	but the numbers are never as precise as ORC because of the conservative stack tracing
13:37:01	FromDiscord	<dom96> would be nice to have the same alloc/dealloc numbers though
13:37:25	Zevv	disruptek awake yet?
13:37:31	Araq	well in theory the numbers should be the same, we don't elide allocations
13:37:39	Araq	(though we should...)
13:38:52	Yardanico	Araq: well for 100 thousand futures with refc getAllocStats() says "allocCount: 33"
13:38:58	Yardanico	and deallocCount: 25
13:39:16	Yardanico	which is kinda wrong I think
13:39:42	Araq	maybe it counts blocks requested from the OS
13:40:39	Yardanico	btw will when defined(gcOrc): still be available in the future?
13:41:01	Araq	sure
13:42:32	Yardanico	seems like the nimyaml creator is okay with supporting orc in the future (when it doesn't leak for sure), and they said they'll look into removing the "source" pointer used in the lexer
13:42:33	Yardanico	https://github.com/flyx/NimYAML/issues/85
13:42:34	disbot	➥ NimYAML compatibility with ARC/ORC.
13:43:34	FromDiscord	<dom96> Why don't we finally introduce a nice ``gc`` enum instead of ``gcOrc`` and whatever else?
13:43:44	FromDiscord	<Clyybber> we do have something like that
13:43:50	FromDiscord	<Clyybber> but its not an enum rn afaik
13:44:05	FromDiscord	<Clyybber> but a string
13:52:07	Araq	wow
13:52:16	Araq	if I comment out
13:52:23	Araq	#result.stackTrace = getStackTraceEntries()
13:52:34	Araq	the deallocation count is off by one only
13:52:44	Yardanico	exactly
13:52:47	Yardanico	:P
13:52:48	FromDiscord	<Clyybber> so its stackTraces not getting freed
13:52:55	FromDiscord	<dom96> nice
13:53:02	FromDiscord	<Clyybber> and if you comment that out, only the last one is not getting freed
13:53:15	Yardanico	I said the thing about stack trace before btw :P
13:53:29	leorize[m]	is @euantorano here?
13:53:46	leorize[m]	I'm not sure what's their irc handle
13:54:00	Araq	and cursor inference makes ORC so much faster ...
13:54:10	Araq	(in theory)
13:54:19	Araq	it's a pure beauty
13:54:45	Yardanico	did you look into the cursorifier issue I reported yet? although it's still not truly minimized (depends on lists module)
13:54:49	leorize	euantor: are you euantorano on github?
13:55:12	Araq	Yardanico: no, that's scheduled as my next bugfix
13:55:15	Yardanico	ah nice :D
13:55:39	FromDiscord	<dom96> does ARC work wellish with JS backend (yet)?
13:55:51	leorize	arc doesn't exist for js :P
13:56:03	leorize	the entire gc switch is now ignored on js fwiw
13:56:15	Yardanico	@dom96 it wouldn't make sense
13:56:21	Yardanico	you can't control memory in JS really
13:56:32	Araq	you can, ArrayBuffer etc
13:56:34	*	jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
13:56:35	Yardanico	oh
13:56:46	Yardanico	would that be faster than using the native JS's gc though?
13:56:53	leorize	obviously not :P
13:57:07	FromDiscord	<dom96> What I'm trying to get at is: are the ARC improvements going to enable less NimCopy's?
13:57:15	FromDiscord	<dom96> or rather, have they already?
13:57:20	leorize	already
13:57:29	FromDiscord	<dom96> I recall Araq saying that they would
13:57:35	Araq	yeah
13:57:43	Araq	but not yet implemented
13:57:54	leorize	a side effect of arc is that destructors and move semantics are developed further and optimized :P
13:57:59	Araq	maybe there is a fortunate interaction that I'm unaware of
13:58:19	FromDiscord	<dom96> leorize: `:P :P :P :P :P`
13:58:26	Araq	leorize: the JS backend doesn't take advantage of moves though
13:58:27	Yardanico	:P is nice!
13:58:40	FromDiscord	<Clyybber> :p
13:59:16	leorize[m]	it's a habit of mine to mask my poor english to avoid offending people xd
13:59:33	Yardanico	oh right about the ircord bridge - a few days ago I added irc formatting -> markdown and markdown -> irc, but didn't test it extensively though
13:59:41	Yardanico	since I have nim irc logs parsed, maybe I can try to test it on that
13:59:47	*	jjido joined #nim
14:00:04	FromDiscord	<dom96> Your English is indistinguishable from someone that's a native speaker IMO
14:00:24	Yardanico	yeah, leorize[m]'s english is pretty nice
14:00:29	Yardanico	mine is kinda worse :P
14:00:59	Araq	problem with ArrayBuffers is DOM interop
14:01:15	Araq	otherwise I would have tried them for Karax back then
14:01:17	FromDiscord	<dom96> Also I actually do use ArrayBuffers in my game
14:01:42	*	vicfred quit (Quit: Leaving)
14:02:02	*	vicfred joined #nim
14:02:05	FromDiscord	<dom96> Otherwise serializing the data structures into thrift binary would have been a bigger pain than it was
14:02:41	leorize[m]	I'm terrible at figuring out the tone though, so my messages might come across as offensive
14:02:45	leorize[m]	but thanks for the compliment :)
14:03:34	FromDiscord	<dom96> Maybe it's just me, but I'm starting to associate the use `:P` with a negative tone.
14:03:41	FromDiscord	<dom96> (edit) 'Maybe it's just me, but I'm starting to associate the use ... `:P`' => 'Maybe it's just me, but I'm starting to associate the useof'
14:03:43	FromDiscord	<Clyybber> the trick is to actively spout bad english so when you accidentally insult someone, they'll know its because of your bad english
14:04:04	Yardanico	thanks
14:04:08	Yardanico	i'll remember that
14:04:18	Yardanico	+
14:04:31	FromGitter	<Lecale> I think it depends on the size of the toungue :p or :P
14:04:32	Yardanico	sory diden't men to sent "+"
14:04:48	FromDiscord	<Clyybber> lol, its the opposite for me, I use `:p` to turn something that could be interpreted in a negative way into a tongue-in-cheek, nicey way :p
14:05:20	FromDiscord	<Rika> I think I'm slowly transitioning from :P to xd
14:05:31	Yardanico	that's worse
14:05:32	Yardanico	imo
14:05:47	Yardanico	and then you'll also start using "bruh"
14:05:50	leorize	@dom96 #15093 should address the getAppFilename issue by documenting it
14:05:51	disbot	https://github.com/nim-lang/Nim/pull/15093 -- 3os: documents potential security implications of getAppFilename
14:06:13	FromDiscord	<dom96> Obviously need to transition to :Þ
14:06:49	Yardanico	sure :P
14:06:52	Yardanico	what about :)
14:07:05	FromDiscord	<Clyybber> what about :))))))))))
14:07:17	Yardanico	but yes, I think I use ":P" so it doesn't sound harsh or rude too
14:07:24	FromDiscord	<Clyybber> fits the honeybadger
14:07:25	leorize	`:)` is context-dependent, the facebook version of that emoji looks like you're dead inside
14:07:38	FromDiscord	<Clyybber> appropriate emoji, thanks
14:07:40	Yardanico	well it's true
14:07:49	FromDiscord	<dom96> It doesn't matter what emote you use, if what you're saying sounds rude to someone then an emote won't help. It'll just cause an association between that emote and rudeness.
14:07:59	FromDiscord	<Recruit_main707> 🙂 discord one also looks pretty dead
14:08:00	FromDiscord	<Clyybber> fuck you :)
14:08:08	FromDiscord	<Clyybber> :p
14:08:14	FromDiscord	<Rika> Yardanico: okay.
14:09:19	Yardanico	@Clyybber fuck you /s :) :P :D xd bruh
14:09:32	FromGitter	<Lecale> An I know nothing about compilation question - is there an obvious solution to the issue I raised here https://github.com/halonium/halonium/issues/1 , short of wearing a tin foil hat :p :p xd stfuartfm
14:09:33	disbot	➥ Compilation Error
14:09:49	FromDiscord	<Clyybber> @Yardanico reported to the discord server admins
14:09:49	Yardanico	hrm
14:09:52	FromDiscord	<dom96> What I'm guessing happens is: you write a message that you know may be construed as rude, so you throw in a `:P`. The person that reads it can see that it's rude and the `:P` loses it's cheeky interpretation.
14:11:05	Yardanico	@Lecale I think you need libzip
14:11:16	Yardanico	but I'm not sure how zip package imports libzip
14:11:25	leorize	they might be missing zlib, not sure
14:11:55	FromDiscord	<Clyybber> @dom96 just out of curiousity, did you think that about this message https://discordapp.com/channels/371759389889003530/371759389889003532/736940642336374816 too ?
14:12:17	Yardanico	btw, as I said yesterday - I published #nim message sqlite db
14:12:29	Yardanico	in https://github.com/timotheecour/Nim/issues/126#issuecomment-663849153
14:12:32	disbot	➥ need a way to access raw db for gitter, irc, forum, enables more powerful search on local clients via raw db access ; snippet at 12https://play.nim-lang.org/#ix=2srB
14:12:42	Yardanico	we really need a DB for the forum too
14:12:50	Yardanico	it contains a lot of useful info which can be kinda hard to find
14:12:53	Araq	Yardanico: if I remove the [T] it also works
14:13:01	Yardanico	Araq: yes I know :P
14:13:03	Araq	so it's not inheritance
14:13:07	Yardanico	oh
14:13:10	Araq	it's generic inheritance
14:13:21	*	jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
14:13:34	FromGitter	<Lecale> cool
14:13:36	Yardanico	well I know that it works when removing [T], maybe I should've left that in the comments
14:13:49	*	fredrikhr quit (Read error: Connection reset by peer)
14:14:26	*	fredrikhr joined #nim
14:14:33	FromDiscord	<dom96> @Clyybber I wouldn't call it rude, but it definitely annoyed me at the time.
14:15:06	FromDiscord	<Clyybber> yeah, thats why I added the :p, because I thought it might be annoying
14:15:13	FromDiscord	<dom96> lol
14:15:14	FromDiscord	<Clyybber> didn't work :p
14:15:18	Yardanico	there are only 23k messages with :P in #nim
14:15:29	FromDiscord	<dom96> and hence my hatred for `:P` grew
14:15:32	Yardanico	out of ~1238755 total
14:15:33	FromDiscord	<Clyybber> lol
14:15:44	*	mbuchel joined #nim
14:15:53	Yardanico	https://i.imgur.com/Ho9O4fx.png
14:16:04	mbuchel	is there a way to have a circular reference?
14:16:06	mbuchel	in nim
14:16:07	Yardanico	yes
14:16:19	mbuchel	i mean recursive import
14:16:26	Yardanico	well it's partially possible
14:16:36	mbuchel	how would i do it?
14:16:43	FromDiscord	<dom96> Yardanico: You should see how many of those started showing up recently 😄
14:16:50	Yardanico	I know :(
14:17:08	Yardanico	there are 714 messages containing 🙂
14:17:10	FromDiscord	<Clyybber> lol :(
14:17:14	leorize[m]	mbuchel: you don't, the "partial" way is buggy at best and should be avoided
14:17:20	Yardanico	is it?
14:17:24	FromDiscord	<Clyybber> its not buggy
14:17:26	Yardanico	I think 4raq said it actually works pretty well
14:17:31	Yardanico	if you actually use it properly
14:17:33	leorize[m]	currently the best we can do is to bundle everything into one file
14:17:49	Yardanico	oh yeah I also found some nice troll messages in the DB
14:18:00	FromDiscord	<lqdev> mbuchel: see the example here https://nim-lang.org/docs/manual.html#modules
14:18:08	mbuchel	i am really against bundling everything inside one file, especially with the protocol i am working on currently
14:18:10	FromDiscord	<dom96> Monorepo? Let's go a step further and create monofiles 😄
14:18:12	leorize[m]	if `a <-> b` and you setup `b` so that `import a` works, `import b` explodes.
14:18:29	Yardanico	nimx uses circular imports btw
14:18:42	FromDiscord	<dom96> mbuchel: you're best off using the utils/types pattern in my experience
14:18:52	mbuchel	yep that is what i am doing
14:19:03	mbuchel	but the protocol i am working on was very poorly designed
14:19:08	mbuchel	and has recursive imports
14:19:18	Yardanico	does that matter really?
14:19:27	Yardanico	you can always kinda restructure it
14:19:30	leorize[m]	why does recursive import have anything to do with a protocol?
14:19:33	mbuchel	you have a file ei, which requires pyld, but pyld requires ei
14:19:49	mbuchel	i am trying to match the same as the xsd file they provide for the protocol
14:20:07	mbuchel	so it would be easier to expand in the future
14:20:55	Yardanico	english language be like https://i.imgur.com/7xWUxyk.png
14:20:56	mbuchel	also thanks for that modules section
14:21:04	mbuchel	there are some things from there i can use
14:21:24	Yardanico	oh no there's dom using :P in that screenshotr
14:21:25	Yardanico	busted
14:21:45	leorize	lol
14:22:05	Yardanico	uhh guess my english was influenced a bit by this chat :P
14:22:11	Yardanico	https://i.imgur.com/UpQcNFj.png
14:23:08	FromDiscord	<dom96> Yardanico: fake news
14:23:09	FromDiscord	<Clyybber> you turned the so yeah into well yes
14:23:17	Yardanico	well yes, well yes
14:23:41	FromDiscord	<dom96> I like how most of what I write ends with a period.
14:23:53	Yardanico	I wonder what can happen if I use markov chains on this DB
14:23:56	Yardanico	https://i.imgur.com/kFMdpWd.png
14:23:59	FromDiscord	<dom96> But yeah, I did use `:P` in the past
14:25:00	supakeen	:p
14:25:22	Yardanico	:) has been used more times than :P in total
14:25:27	Yardanico	only 1k message difference though
14:25:30	supakeen	dom96: what library did you rock for the star game or something by yourself?
14:25:59	FromDiscord	<dom96> supakeen: I wrote one: gamelight
14:26:11	FromDiscord	<Clyybber> Araq: This doesn't seem to be needed anymore https://github.com/nim-lang/Nim/blob/devel/compiler/semexprs.nim#L1206 (checked with the tests), is it ok to remove?
14:26:29	FromDiscord	<dom96> Ugh, I thought I could disable link previews in Discord
14:26:34	Yardanico	you can't :(
14:26:42	FromDiscord	<dom96> but it also disables them when someone pastes image links
14:26:53	FromDiscord	<dom96> pretty sure that's a bug because it displays a white space
14:26:58	FromDiscord	<dom96> and not the image link itself
14:27:20	Araq	clyybber, yeah if the CI says everything is fine it's safe to be removed
14:27:25	FromDiscord	<dom96> Maybe it's time I transition to IRC again
14:27:26	supakeen	Ah I see, I've been toying with `nico` :)
14:28:11	FromDiscord	<dom96> supakeen: yeah, nico is cool but has some specific limitations AFAIK
14:28:20	FromDiscord	<dom96> like I don't think I'd be able to build Stardust using it
14:28:26	Yardanico	well that's what nico was made for
14:28:28	Yardanico	something like pico8
14:28:37	Yardanico	create pixel games easily
14:28:42	FromDiscord	<dom96> yeah
14:28:46	supakeen	Yea I'm currently scratching my game-building itch for a dungeon crawler with nico and so far i'm ok.
14:28:55	supakeen	Love the simplistic pico8 api indeed.
14:29:47	FromDiscord	<dom96> Yeah, it's good to have restrictions. They help creativity especially with games.
14:29:48	Yardanico	well I have a small game-buildting itch, but either for an osu!mania clone or an RTS (which is much harder than an osu!mania clone)
14:30:10	FromDiscord	<dom96> Yardanico: just gotta c2nim 0 A.D. 😄
14:30:14	Yardanico	lol
14:30:25	Yardanico	maybe Zero-K instead? :)
14:30:35	supakeen	RTS sounds cool too.
14:30:42	Yardanico	There's an open source Spring engine for RTS
14:30:45	Yardanico	and Zero-K - https://github.com/ZeroK-RTS/Zero-K
14:30:46	supakeen	Any specific ideas about it if you just think freely without being realistic? :)
14:31:20	Araq	no matter what game you program
14:31:29	Araq	if you make it "text based" I'm gonna kill you
14:31:33	Yardanico	nonono
14:31:37	Araq	I hate this terminal bullshit
14:31:46	Yardanico	I was born after the text-based games era :)
14:31:47	FromDiscord	<Clyybber> you will kill *me*
14:31:48	Araq	even DOS had better stuff
14:32:05	FromDiscord	<dom96> nethack ftw
14:32:10	supakeen	nimhack
14:32:18	FromDiscord	<Clyybber> I don't know why, but theres something that intrigues me about text based games
14:32:36	FromDiscord	<dom96> now can we create a multiplayer nethack hmmmm
14:32:39	supakeen	what if i pretend to draw a terminal with sdl?
14:32:46	FromDiscord	<Clyybber> now that, that is shit
14:32:48	Yardanico	there's an imgui-like lib for terminal btw
14:32:49	FromDiscord	<Clyybber> that I hate
14:33:07	Araq	supakeen: that's the start of every good game engine with hot code reloading
14:33:12	Araq	:D
14:33:23	supakeen	i think that's how dwarf fortress does it nowadays
14:33:31	supakeen	though perhaps they still have different backends :)
14:33:33	FromDiscord	<Clyybber> dwarf fortress still has TEXT_MODE
14:33:52	Araq	press tab and get quake's console. and even that one is better than iTerm
14:33:53	FromDiscord	<Clyybber> but with the steam release it will get a nice graphical mode
14:34:11	Araq	(Nah, iTerm is ok)
14:35:07	supakeen	Is that the one that does DNS lookups for all the text you select?
14:35:48	supakeen	Ah yes, and the one that had code execution if you look at the wrong bytes (e.g. a logfile).
14:36:18	supakeen	I will not understand why that terminal is so complex at least they fixed column redraws on it.
14:40:18	euantor	leorize: Apologies for the delay; yes I'm `euantorano` on GitHub :)
14:40:26	Yardanico	i have to agree, some people try to add some stupid features to terminal emulators
14:40:31	Yardanico	like why would you need images in the terminal
14:40:33	Yardanico	use a proper GUI then
14:40:46	Yardanico	(looking at people who use it exclusively for neofetch)
14:40:57	FromDiscord	<Rika> :BlobSweatAnimated:
14:41:19	leorize	euantor: can you help me figure out why stdlib/tssl in #15066 just exit with error code 1 on freebsd? thanks :)
14:41:20	disbot	https://github.com/nim-lang/Nim/pull/15066 -- 3asyncnet, net: don't attempt SSL_shutdown if a fatal error occurred
14:41:30	FromDiscord	<Rika> Ngl I was considering using a term emu for that feature
14:42:23	euantor	leorize: I can certainly take a look, though I'm at work right now. I'll try have a quick look now and might have to get back to you later
14:45:16	*	FromDiscord quit (Remote host closed the connection)
14:45:31	*	FromDiscord joined #nim
14:50:13	FromDiscord	<exelotl> about text-based games... someone released a beautiful looking ascii tileset for my favourite roguelike https://github.com/PropFeds/terminal-blues
14:50:37	FromDiscord	<exelotl> I find it really hard to play though
14:50:48	FromDiscord	<exelotl> i've been playing with the default graphical tiles for years
14:52:12	FromDiscord	<exelotl> now I have to remember that a yellow "b" is a fire beetle, a pink "b" is a grid bug
14:59:12	FromDiscord	<Varriount> dom96: It's a shame there are no when-case statements. They would be helpful (at least internally for the compiler/standard library)
15:00:02	FromDiscord	<Varriount> exelotl: What is POWDER like? Hardcore and intricate, or more forgiving?
15:00:17	solitudesf	its not forgiving
15:00:43	solitudesf	its like a less complicated nethack, i suppose
15:00:45	FromDiscord	<Rika> well its a roguelike
15:02:40	disruptek	!last zevv
15:02:40	disbot	Zevv spoke in 12#nim 85 minutes ago 12https://irclogs.nim-lang.org/27-07-2020.html#13:37:25
15:02:42	disruptek	so.
15:02:54	Zevv	so.
15:03:02	disruptek	do we write ast that is correct or ast that works?
15:03:04	Zevv	this town is too small for the both of us
15:03:08	disruptek	lol
15:03:15	disruptek	g'day zevv
15:03:20	Zevv	moin disruptek
15:03:22	Zevv	slept well?
15:03:35	disruptek	too many bugs up here.
15:03:51	Zevv	know the feeling
15:04:01	*	audiofile joined #nim
15:04:09	disruptek	TIL nnkConv
15:04:21	Zevv	what is that stuff even
15:04:35	Zevv	you just put in a dotExpr and hoped that made sense, right
15:04:53	disruptek	no, it was nnkCall.
15:05:13	Araq	using nnkCall to construct type conversion is correct btw
15:05:17	disruptek	but, nnkConv doesn't seem to work.
15:05:24	disruptek	hence my query.
15:05:28	disruptek	do we write ast that is correct or ast that works?
15:05:45	Araq	nnkConv isn't correct, it's part of Nim's black hole
15:05:47	Zevv	let me draw a little venn diagram
15:05:53	Araq	which is sem'checked ASTs
15:06:16	Zevv	so, what would you recommend
15:06:59	disruptek	he's going to recommend typed.
15:07:11	disruptek	but we shouldn't have to go there yet.
15:07:25	Araq	construct untyped ASTs
15:07:40	disruptek	l0rd knows i'm trying.
15:09:05	*	altarrel joined #nim
15:10:25	*	maier quit (Ping timeout: 240 seconds)
15:10:34	Zevv	so, we could go with the cast, but only if we cna properly isolate the underlying problem of the conversion going wrong and file a proper nim issue for that
15:10:41	Zevv	because I would really hate to have the cast in cps
15:10:53	Zevv	I feep cps might end op being provable somehow, and with a cast you throw that away
15:11:03	disruptek	it's already proven.
15:11:16	disruptek	anyway, i don't care.
15:11:35	Zevv	cast away, then
15:11:37	disruptek	i'm trying to develop semantics. the machinery isn't important to me.
15:12:24	disruptek	oh, the root cast /was/ a dotexpr
15:12:39	Zevv	right
15:12:57	Zevv	not sure if nim understands your intention there.
15:13:04	disruptek	the child cast /is/ a call.
15:13:11	FromGitter	<alehander92> why cast
15:13:12	Zevv	should also be nnkConv
15:13:15	Zevv	WHY CAST?!
15:13:17	FromGitter	<alehander92> why is it such a problem
15:13:18	Zevv	WHYYYYYYY
15:13:19	disruptek	AIIEEEE!
15:13:19	FromGitter	<alehander92> to cast
15:13:22	FromGitter	<alehander92> i cast often
15:13:25	FromGitter	<alehander92> like i get it
15:13:27	Zevv	you cast, you fail
15:13:29	disruptek	OMGOMGOMGOMG
15:13:32	FromGitter	<alehander92> but sometimes he compiler should trust me
15:13:39	FromGitter	<alehander92> i mean i am a human
15:13:43	Yardanico	but why cast
15:13:45	Yardanico	if it's not needed
15:13:46	FromGitter	<alehander92> AND I NEED TO BE LOVED
15:13:52	disruptek	look, it's like when i used to teach esl to spanish kids in nyc.
15:14:03	Yardanico	==============================
15:14:08	disruptek	they were always so confused about the difference between `right` and `correct`.
15:14:12	FromGitter	<alehander92> can i interview yourself
15:14:21	FromDiscord	<dom96> plz dont cast bois
15:14:28	Zevv	screw you guys, I'm going home
15:14:34	FromGitter	<alehander92> but how can you have AST without CAST
15:14:38	disruptek	i'd tell them, listen, you can stick a banana up your ass -- it fits `right`, but it's not `correct`.
15:14:51	Zevv	alehander92: are you pulling legs?
15:15:09	Zevv	it's just a tree of sum types. Where is the casting?
15:15:17	FromGitter	<alehander92> yeah i have no idea
15:15:24	FromGitter	<alehander92> i am just trying to make puns
15:15:32	Zevv	ooooh I see it C-AST
15:15:33	Zevv	dude
15:15:41	Zevv	it's actually pretty good. Too bad I let you explain it
15:15:47	FromGitter	<alehander92> it's basic
15:15:49	FromGitter	<alehander92> but thank you
15:15:54	Zevv	:slow clap:
15:16:08	FromGitter	<alehander92> disruptek but what is correct
15:16:10	FromGitter	<alehander92> but not right
15:16:17	FromGitter	<alehander92> is it a correction facility
15:16:20	Zevv	anway disruptek: nnkConv does not solve it for you, so there might be a hole in nnConv - is that a safe conclusion of this adventure>?
15:16:40	FromGitter	<alehander92> because i think they might have imperfect work there
15:16:49	disruptek	Zevv: not really.
15:17:23	disruptek	per araq nnkConv is the domain of typed ast and not our baliwick.
15:17:39	disruptek	bailiwick, too.
15:18:37	disruptek	i thought my solution to shadowing proc params was pretty cute.
15:19:25	Zevv	oh i missed that, tell it to me
15:19:41	disruptek	just put the body in a block:
15:20:34	disruptek	i'm thinking that we can use the block to do something interesting.
15:20:42	disruptek	like exception handling.
15:21:34	Zevv	ooh right, why not
15:21:51	disruptek	remember that we know the return type and we can stuff the exception in on the way out.
15:21:58	Zevv	I spent a few minutes this afternoon to build some primitives like iterators and coroutines
15:22:08	disruptek	some day it will generate a nim compiler error, of course.
15:22:29	disruptek	where is your commit?
15:22:46	Zevv	not there yet
15:22:56	*	Zevv is now known as Zevv[m]
15:23:21	*	disruptek is now known as disruptek[o]
15:23:39	*	Yardanico is now known as yardanico[ok]
15:23:51	yardanico[ok]	ok
15:24:11	*	yardanico[ok] is now known as Yardanico
15:24:58	disruptek[o]	wtf someone stole my asd nick.
15:25:01	Yardanico	wtf
15:25:20	disruptek[o]	that's bs
15:27:06	*	vicfred quit (Quit: Leaving)
15:27:20	*	disruptek[o] is now known as disruptek
15:27:22	*	disruptek growls.
15:30:40	*	hoijui joined #nim
15:35:10	Zevv[m]	disruptek: https://github.com/disruptek/cps/blob/master/stash/iterator.nim
15:35:15	Zevv[m]	pretty basic stuff
15:38:56	Zevv[m]	ore reload, if you like Option[]
15:39:04	disruptek	looks good to me.
15:39:25	Zevv[m]	we need more trampolining then strictly necessary, maybe
15:39:28	Zevv[m]	but I don't care for now
15:39:36	Yardanico	btw, why you still need to specify the type ?
15:39:43	Yardanico	https://github.com/disruptek/cps/blob/master/stash/iterator.nim#L29 here for example
15:39:45	Yardanico	what's the reason
15:39:51	disruptek	the ast is untyped.
15:39:53	Yardanico	so?
15:40:00	disruptek	we cannot infer the type.
15:40:08	disruptek	we need the type in order to construct the continuation.
15:40:09	Yardanico	just use auto :P
15:40:19	disruptek	hard to define objects with auto.
15:40:55	disruptek	hmm, i wonder if we could use a fuckton of generics, though.
15:41:07	Yardanico	but i hope that this limitation will eventually be removed, right? :P
15:41:10	Yardanico	current way looks kinda ugly
15:41:24	disruptek	i'm not bothered.
15:41:24	Yardanico	also do you really need a while loop in each proc which uses the cps macro?
15:41:28	disruptek	no.
15:43:06	Zevv[m]	so, there is really no way to punch your code through two macros, right? First do an untyped pass, then a typed one after that
15:43:27	Zevv[m]	Yardanico: no, but that's one of the reasons you'd *want* to do this
15:43:29	disruptek	i haven't done it, but someone told me it was a thing.
15:43:56	Zevv[m]	Yardanico: but do note how cool this is: the concept of 'yield' and 'resume' is all in that file.
15:44:00	Zevv[m]	there is no magic
15:44:50	disruptek	it's a big deal. eventually, people will understand.
15:45:01	disruptek	it'll take time. people need to use this stuff.
15:45:04	Zevv[m]	it's like monads
15:45:48	*	rockcavera joined #nim
15:45:50	disruptek	it's like mescaline, but monads works, too.
15:46:24	Zevv[m]	disruptek: I'd like to talk to you
15:46:25	Zevv[m]	about rvalues
15:46:28	Zevv[m]	or is it too soon for that
15:46:31	Zevv[m]	in our relationship
15:46:48	Yardanico	about types
15:46:51	Yardanico	I know it's a bit of a hack
15:46:55	disruptek	we should think about this generics idea first.
15:46:56	Yardanico	but what about using typeof
15:47:15	Yardanico	var a: typeof(expression) = expression
15:47:16	disruptek	it's not possible because the types need to be defined before they are inferred.
15:47:19	Yardanico	var a: typeof(5) = 5
15:47:42	Zevv[m]	var a = doThis()
15:47:52	Zevv[m]	var a: typeof(doThis()) = doThis() ?
15:48:03	Yardanico	yes
15:48:53	Araq	this is so weird
15:49:00	Yardanico	the leak? :P
15:49:10	Araq	Future[T] = ref object of RootObj # about 30_000 allocs
15:49:21	Araq	Future[T] = ref object # about 25_000 allocs!
15:49:22	Yardanico	:DD
15:49:37	Araq	but it's inheriting from nothing, what causes 5000 more allocations?
15:49:54	disruptek	method names?
15:49:55	FromDiscord	<dom96> wat
15:50:00	FromDiscord	<dom96> it's inheriting from RootObj?
15:50:06	Yardanico	FutureBase does, yes
15:50:09	FromDiscord	<dom96> That's not nothing
15:50:25	Yardanico	disruptek: nim async doesn't use methods even if it uses inheritance
15:50:49	disruptek	the names are constants but we dispatch dynamically.
15:51:11	Yardanico	ah maybe the isObj thing you mean?
15:51:28	FromDiscord	<Shucks> ❤️
15:51:32	disruptek	i dunno, it's mere intuition.
15:53:15	Zevv[m]	so disruptek, what's next on the list
15:53:20	disruptek	generics.
15:53:56	Araq	dom96: RootObj is an empty object, inheriting from it doesn't cause more allocations
15:54:05	Zevv[m]	disruptek: how generics, tell me
15:54:10	Zevv[m]	what do you need to genericize
15:54:24	Zevv[m]	because I don't think I needed this yet?
15:54:26	disruptek	noroutine field values.
15:54:33	Zevv[m]	hmmm
15:54:58	Zevv[m]	is that something we need *now*
15:55:03	Zevv[m]	and please don't call'em that
15:55:14	disruptek	lol
15:57:03	Zevv[m]	yarn. floss. twine.
15:57:04	Zevv[m]	pick one
15:57:31	FromDiscord	<dom96> Araq: and yet it seems it does? 🙂
15:57:35	Zevv[m]	I think generics can wait. It's cool and all but will complicate your macros and will not bring essential stuff on the table now
15:58:01	disruptek	it brings type inference, which isn't nothing.
15:58:19	Zevv[m]	type inference for what parts?
15:59:06	disruptek	noroutine field values.
15:59:28	Zevv[m]	yeah but not for stuff on the environment
15:59:32	Zevv[m]	which is more painful
15:59:38	disruptek	yes, stuff on the environment.
15:59:44	Zevv[m]	mo way
15:59:47	Zevv[m]	var i = 3
15:59:50	Zevv[m]	'give i a type'
15:59:52	Zevv[m]	you won't solve that
16:00:24	Zevv[m]	you want this? http://ix.io/2ss2
16:01:21	Araq	dom96, yeah, it's not explainable
16:01:32	Yardanico	wait you can't find the cause of it?
16:01:39	Yardanico	i mean not yet :P
16:01:47	*	audiofile quit (Quit: Default Quit Message)
16:02:06	disruptek	i will solve that.
16:02:25	disruptek	Cont[typeof(rando)](fn: ..., rando: rando)
16:02:26	Yardanico	now where's our color change on github
16:02:29	Yardanico	a month passed
16:02:40	Zevv[m]	disruptek: sure, nice. but you wont solve local variables to get type by magic
16:02:56	disruptek	the local vars don't need a type; it's inferred.
16:03:18	Zevv[m]	no way.
16:03:22	disruptek	var rando = env_3(locals_3).rando
16:03:22	Zevv[m]	surprise me
16:03:36	disruptek	yard gave us the hint.
16:03:57	Zevv[m]	no that was wrong
16:04:04	Zevv[m]	I just tried, doesnt compute
16:04:08	disruptek	why don't you make us `:=`() mr. go lover?
16:04:16	Zevv[m]	bah
16:04:31	Zevv[m]	neverr
16:04:57	disruptek	you tried what?
16:05:17	Zevv[m]	var a: typeof(foo()) = foo()
16:05:53	disruptek	i dunno what that is.
16:05:55	Zevv[m]	var a: typeof(5) = 5 might
16:06:04	Yardanico	!eval import strutils; var a: typeof("a b".split()) = "a b".split(); echo a
16:06:04	disruptek	i'm talking about something different, i guess.
16:06:06	NimBot	Compile failed: /usercode/in.nim(1, 54) Error: type mismatch: got <seq[string]> but expected 'string'
16:06:27	disruptek	holdon, i will create an example.
16:06:34	Zevv[m]	yes sir please sir
16:06:47	Yardanico	disruptek: we're talking about the fact that you have to resort to specifying the type
16:06:50	Yardanico	like in that zevv's exampole
16:06:52	Yardanico	example*
16:06:55	Yardanico	which is IMO quite ugly :)
16:06:57	disruptek	i know, chucklehead.
16:07:12	disruptek	i mean, yardanico.
16:07:24	Zevv[m]	dude
16:07:26	Zevv[m]	rude
16:07:37	Zevv[m]	I will fry my fries and eat them not. bbl
16:07:45	Zevv[m]	s/not/now/
16:07:49	disruptek	http://ix.io/2ss4/nim
16:08:22	disruptek	typeless type inference.
16:09:51	*	FromDiscord quit (Remote host closed the connection)
16:10:08	*	FromDiscord joined #nim
16:13:42	FromDiscord	<jcs224> I have a question about convention. If I have a `proc` called `readLine`, which is already in the standard library, what would be the "proper" way to avoid a namespace collision with that existing function? There are quite a lot of keywords in Nim. Coming from JavaScript.
16:14:02	Yardanico	well, does your readLine proc have the same signature as the one in stdlib?
16:14:43	Yardanico	if signatures are different, there wouldn't be any collisions
16:14:54	FromDiscord	<jcs224> Yeah, it takes a `Socket` as the first argument, same as the stdlib one. That's the only argument.
16:14:57	Yardanico	if there are collisions, you can always specify the module name before the identifier
16:15:04	FromDiscord	<dom96> You don't have to do anything, your `readLine` will already be in a different module and therefore different namespace
16:15:23	disruptek	just don't import the stdlib version.
16:15:26	Yardanico	lol
16:16:29	FromDiscord	<jcs224> Ok thanks. I tried just naming it `readLine` and it works fine. I have to get used to the automatic nature of importing, importing in JS and PHP is much more explicit. I've used Python but very little.
16:17:18	Yardanico	you can always use the python way of importing but it's not really needed most of the time :)
16:19:56	FromDiscord	<jcs224> and thanks for the super-prompt responses. Loving this language so far!
16:24:21	disruptek	leorize: we're missing half the nightlies for devel last couple nights...?
16:24:53	leorize	disruptek: it's a bug caused by fusion, only fixed today
16:24:59	disruptek	cool.
16:25:12	disruptek	i'm loving this new fusion thing.
16:25:15	disruptek	really great.
16:25:47	disruptek	what else can i do to help? 😉
16:27:58	euantor	leorize: I've had a quick look and posted a comment on the PR. Looks like the process gets a `SIGSEV`
16:28:57	*	vicfred joined #nim
16:29:34	leorize	thanks for looking
16:29:45	leorize	seems to be a bug with threads
16:40:31	*	waleee-cl quit (Quit: Connection closed for inactivity)
16:41:48	Zevv[m]	whats fusion
16:42:18	FromDiscord	<dom96> (bad answers only)
16:43:10	disruptek	an innovation to shorten ci runs.
16:45:45	bung	old problem, how I test unittest ?
16:46:27	Yardanico	just test it ?
16:46:33	Yardanico	expected output
16:47:10	Yardanico	see e.g. https://github.com/nim-lang/Nim/blob/devel/tests/stdlib/tunittest.nim
16:47:58	bung	oh, that's easy, so I just copy paste my console outputs
16:51:25	bung	hmm, there's file name line column
16:51:29	Yardanico	yes
16:54:40	bung	how I hanlde the line column things ? hard code will make it hard to change
16:55:09	Yardanico	well don't handle it
16:55:21	Yardanico	just insert it in the test, and then tune so the output is the same
16:57:28	bung	oh, basically just change some line number ,that's ok
16:58:34	*	superbia1 joined #nim
16:58:39	*	someunknownuser joined #nim
16:59:37	someunknownuser	Hello
16:59:43	Zevv[m]	Hello
17:00:08	someunknownuser	does anyone know why osproc's hasData proc blocks until data is available?
17:00:23	someunknownuser	(at least on linux)
17:00:30	*	WilhelmVonWeiner joined #nim
17:00:45	*	superbia quit (Ping timeout: 240 seconds)
17:01:22	someunknownuser	It looks like it calls select internally, but passes nil for the timeout argument, which results in it blocking until data becomes available.
17:01:28	FromDiscord	<dom96> Wow. Julia is doing an AMA on the main IAmA subreddit https://reddit.com/r/IAmA/comments/hyva5n/we_are_the_creators_of_the_julia_programming/
17:01:46	*	pangey quit (Ping timeout: 256 seconds)
17:01:58	disruptek	someunknownuser: sounds like a bug.
17:02:58	someunknownuser	currently the behavior suggests that the proc should be named/called waitData (or something similar) instead.
17:03:13	*	pangey joined #nim
17:03:30	Zevv[m]	ej disruptek
17:03:45	someunknownuser	The bool return values becomes basically useless, since obviously data is available after it waited for it to become available xD
17:04:08	Zevv[m]	we now have 'cps foo()' where foo is `proc foo(c: MyCont, ...): MyCont` right. CPS is a magic keyword
17:04:23	disruptek	hai
17:04:26	Zevv[m]	what if we make cps an implicit variable in cps functions which is just the current cont
17:04:33	Zevv[m]	then we do `cps.foo()`
17:04:39	Zevv[m]	and there is less magic.
17:04:52	disruptek	jesus it almost makes sense.
17:05:32	Zevv[m]	the good thing is that the cps closure is available in the .cps. proc. But the bad thing is that the cps closure is available in the .cps. proc.
17:05:37	disruptek	why make it implicit, though? we'll just use cps as our locals.
17:06:01	Zevv[m]	yeah, but the user does not declare 'cps' somewhere, or do we take it a argument #1
17:06:04	Zevv[m]	in .cps. procs
17:06:17	disruptek	we rewrite them.
17:06:27	disruptek	it solves the problem really elegantly.
17:06:32	Zevv[m]	yeah but where does the ident "cps" come in, where is it declared?
17:06:34	disruptek	see, i knew there was a reason i kept you around.
17:06:37	Zevv[m]	Write me an example of how it would look
17:06:53	Zevv[m]	Well, I have acres to sniff, but if you put me outside I just bark
17:07:06	*	maier joined #nim
17:07:13	disruptek	`proc after_17860210(locals_17860216: Cont): Cont` turns into `proc after_17860210(cps: Cont): Cont`
17:07:22	Zevv[m]	yeah, right, but 'cps' is implicit for the user
17:07:32	Zevv[m]	You don't see where `cps` is declared
17:07:36	disruptek	correct.
17:07:40	Zevv[m]	right, that's what I mean
17:07:49	disruptek	you want to make it explicit?
17:07:56	Zevv[m]	not sure. What does it add or solve?
17:08:11	Zevv[m]	The good thing is that the state is available from within your .cps. procs, if you like that kind of thing
17:08:28	Zevv[m]	but I'm not sure if that's opening pandoras box
17:08:31	Zevv[m]	'
17:08:37	disruptek	i agree, it's a little creepy.
17:09:12	Zevv[m]	But then again. Imagine our Big Best eventqueue, which will eventualle become the default in the stdlib, is a derivable type
17:09:28	Zevv[m]	you can just make your own "Foo as EventQueue"
17:09:31	Zevv[m]	put stuff in it
17:09:33	Zevv[m]	and have it there
17:09:40	Zevv[m]	it's powerful, but maybe a bit too
17:10:00	Zevv[m]	All the parent stuff is hidden because not exported, but your stuff is
17:10:41	disruptek	i mean, it's not a problem. you have that power anyway -- just call `cps myproc()` and you'll get the continuation. just, not in the same scope.
17:10:52	Zevv[m]	right
17:10:59	disruptek	once we give you cps, we don't care what you do with it.
17:11:04	Zevv[m]	it's Nim, right
17:11:10	disruptek	pretty sure, yes.
17:11:11	Zevv[m]	I'll take the red one
17:11:42	Zevv[m]	I do like explicit over implicit, I think. Problem is that it will not be used in 95% of the cases, probably
17:11:58	Zevv[m]	only as a thing to hang your `cont.func()` calls on
17:12:05	*	maier quit (Ping timeout: 240 seconds)
17:12:08	disruptek	it's easier for people to grok.
17:12:14	Zevv[m]	and wait, there's more: can you recognize a cps call if its not a fixed string?
17:12:18	disruptek	or is it? i mean, it's magical.
17:12:32	Zevv[m]	you need to save the id of arg #1 and match for that
17:14:00	*	rockcavera quit (Remote host closed the connection)
17:14:35	*	waleee-cl joined #nim
17:15:51	Zevv[m]	and probably handle the case where stupid/smart people shadow that, etc
17:16:04	Zevv[m]	because you're not *really* calling that thing, you can't know because you are untyped
17:16:15	Zevv[m]	it's just a nnkDotExpr
17:16:24	disruptek	well, we'll know if it's shadowed.
17:16:52	disruptek	but, yeah, it's tricky. i'm wondering if we're going about this all wrong.
17:17:19	Zevv[m]	probably.
17:17:20	disruptek	now that early return is a thing, maybe we want `return cps.foo()` to be supplied by the user.
17:17:33	Zevv[m]	argh
17:17:53	disruptek	well, it's explicit.
17:17:58	Zevv[m]	yeah but but
17:18:35	Zevv[m]	wow btw, what do we do with real 'returns' now, do you handlet them?
17:18:42	disruptek	of course.
17:18:47	Zevv[m]	because you shouldn't change return semantics
17:18:53	Zevv[m]	That'll confuse the hell out of people
17:19:28	Zevv[m]	bah then better keep it like it is
17:19:33	Zevv[m]	and find a good keyword for 'cps'
17:19:35	*	pietroppeter quit (Quit: Connection closed for inactivity)
17:19:48	disruptek	you idea is good.
17:19:53	disruptek	your idea is good, too.
17:20:10	Zevv[m]	yeah, but I now start to crawl back
17:20:22	Zevv[m]	like it is now we make 'cps' a keywordy kind of thing
17:20:34	disruptek	we can ban shadowing, which is fine. we can always ban renaming the 1st arg. it's cps, always.
17:20:35	Zevv[m]	so you can look at code and know for sure that *there* it will go out
17:20:41	*	ofelas joined #nim
17:20:56	Zevv[m]	if you go cps.yield(), it doesn't show. if cps is called 'foo' in my proc, there's 'foo.this()' and boom you're out
17:21:10	disruptek	we ban renaming.
17:21:23	Zevv[m]	I can hardle read other peoples code or my own code after two days. but this will be messy
17:21:26	disruptek	anyway, "you're out" isn't a thing.
17:21:28	FromGitter	<alehander92> are we in USSR
17:21:28	FromGitter	<alehander92> now
17:21:40	Zevv[m]	no, forget all this. It's a baaad idea mkayy
17:21:44	*	pangey quit (Ping timeout: 260 seconds)
17:21:51	Zevv[m]	we need to treat it as a keyword
17:22:01	Zevv[m]	ask leorize to make it stand out for us
17:22:09	FromGitter	<alehander92> CCP or CPS
17:22:10	Zevv[m]	because you should read code and know funny stuff is happening
17:22:24	disruptek	mmm nah
17:22:28	FromGitter	<alehander92> what is this ban renaming thing
17:22:32	Zevv[m]	hmmm jah
17:22:35	Zevv[m]	read up boi
17:22:57	disruptek	the reason we don't use continuations in imperative code is that it's too hard to follow.
17:22:59	Zevv[m]	it's not important anyway, because it was a baaaahd idea
17:23:10	FromGitter	<alehander92> keywords sound good
17:23:19	FromGitter	<alehander92> i believe they are a good idea
17:23:23	disruptek	what we're doing is bringing easy style to a hard design.
17:23:26	Zevv[m]	like, in general
17:23:30	Zevv[m]	you're not the lisp kind of guy
17:24:02	FromGitter	<alehander92> disruptek is not a lisp kind of guy
17:24:13	Zevv[m]	disruptek: yeah, but it's just too special to hide it.
17:24:14	disruptek	if we do our job right, no one has to know how this shit works.
17:24:18	Zevv[m]	it is *like* a return
17:24:26	Zevv[m]	or a `yield` or a `await`
17:24:29	FromGitter	<alehander92> i was such a guy, but then i realized one can quote non-homoiconic code
17:24:33	FromGitter	<alehander92> in nim
17:24:46	FromGitter	<alehander92> did you decide to do monads after all
17:24:59	disruptek	i'm doing nomads.
17:25:00	Zevv[m]	ohh wait I see what you mean
17:25:10	Zevv[m]	you want to go so far as to hide the fact you're scheduled out
17:25:11	Zevv[m]	at all
17:25:15	disruptek	i have a sign on the door of my rv that says, "knock if you're horny"
17:25:43	FromGitter	<alehander92> if i even knock, dont trust the sign
17:25:45	Zevv[m]	dude, it *was* a good idea
17:25:48	Zevv[m]	but then we must hide the return
17:26:03	Zevv[m]	doing magic io, you do 'magicio.recv()'
17:26:05	Zevv[m]	and you get your result
17:26:14	disruptek	that's how it works now.
17:26:14	Zevv[m]	you don't want to make the 'return magicio.recv()'
17:26:18	*	arecacea1 quit (Remote host closed the connection)
17:26:19	Zevv[m]	right
17:26:25	FromGitter	<alehander92> really sounds like do notation
17:26:28	Zevv[m]	because return will just return from your innermost function
17:26:31	Zevv[m]	don't go changing that
17:26:41	*	arecacea1 joined #nim
17:27:08	Zevv[m]	well, I learned that with choices I can't choose from, there's a *reason* i can not choose.
17:27:20	Zevv[m]	Either it doesn't make a difference whatever I choose, or I just have not enough information
17:27:28	*	endragor quit (Remote host closed the connection)
17:27:34	Zevv[m]	this is case #2. I'd like to give it a go
17:27:35	disruptek	it's usually dom96's fault, let's be honest here.
17:27:55	*	endragor joined #nim
17:28:37	Zevv[m]	disruptek: leave dom alone and go type that in
17:28:46	Zevv[m]	make it explicit, and renamable
17:28:54	disruptek	did you read any donald norman, ever?
17:29:06	Zevv[m]	I read a Donald once
17:29:12	*	sputny joined #nim
17:29:20	disruptek	everyone should read `the design of everyday things` at least.
17:29:20	Zevv[m]	the design of things something
17:29:23	Zevv[m]	right, that one I read
17:29:44	disruptek	so one of the ideas is that people need to form a mental model of how the thing works.
17:30:01	disruptek	but in our case, it doesn't need to be based upon reality.
17:30:21	disruptek	we give them `cps.foo()` to hang their hat on, but it's just a magical device.
17:30:43	Zevv[m]	that was about people smashing into glass doors, right?
17:30:52	Zevv[m]	I just remember the parts with blood & violence
17:31:03	disruptek	that was the stallone remake.
17:31:17	Zevv[m]	right, I do agree. Give them `cps.foo()`, but make `cps` whatever you pass
17:31:38	Zevv[m]	because it differs. If it is an event queue I'll call it `evq`. If it is an iterator I'll call it `it`, and for coroutines I'll call it `co`
17:32:15	*	endragor quit (Ping timeout: 256 seconds)
17:32:18	disruptek	okay, that probably does make some sense.
17:32:24	disruptek	but we deny shadows?
17:32:38	Zevv[m]	nope
17:32:48	Zevv[m]	if you shadow it and call it, it will just be caught by the compiler
17:32:52	Zevv[m]	you're just calling the wrong thing
17:33:08	disruptek	you could shadow it with the same type, though.
17:33:14	Zevv[m]	for good reasons, probably
17:33:22	Zevv[m]	it's not differnet then shadowing any var
17:33:24	*	pangey joined #nim
17:33:28	disruptek	fair enough.
17:33:32	Zevv[m]	dude, am I smart or what
17:33:52	disruptek	let me get back to you on that.
17:33:56	Zevv[m]	do
17:34:24	Zevv[m]	also, I don't want to see the word 'cps' in everyday use of this thing
17:34:40	Zevv[m]	you implement a coroutine with cps, and from then on you're working with coroutines. Not with cps
17:35:04	Zevv[m]	also, beacuse I keep typing csp and can't see the difference between the two
17:35:59	disruptek	make some packages, then.
17:36:24	Zevv[m]	no, I mean, when you're using the coroutine thingy, you don't need to know what "cps" is
17:36:34	Zevv[m]	when you *make* it you do, of course, but not when you *use* it
17:36:39	disruptek	i know, so publish a coroutine package named whatever.
17:36:46	Zevv[m]	sure, that's what I mean
17:36:52	Zevv[m]	but I don't want coroutine users to type "cps" all the time
17:37:06	disruptek	then don't make them create their own coroutines.
17:37:11	disruptek	provide them a new symbol.
17:37:24	Zevv[m]	right
17:38:12	disruptek	there are a lot of people that want a properly broad iterator package based on cps.
17:38:38	*	ForumUpdaterBot quit (Remote host closed the connection)
17:38:45	*	ForumUpdaterBot joined #nim
17:39:05	Zevv[m]	yeah sure, but you get what I mean
17:39:07	disruptek	okay, there are 3 people.
17:39:13	Zevv[m]	they work with 'it', not with 'cps'
17:39:14	Yardanico	"we can now pass an int and a float variable with devil compiler. Its fine of course, see" lol
17:39:27	Zevv[m]	one day we will say "there's dozens of us", just like the Nim users
17:41:00	Zevv[m]	I want 'for' to understand this stuff
17:41:41	disruptek	i hope we can rewrite it.
17:41:55	Zevv[m]	no, the other way around. Not this stuff to understand `for`
17:42:03	Zevv[m]	I want to be able to use `for` to run my iterator
17:42:26	disruptek	hmm, i don't know how hard that is.
17:42:29	Zevv[m]	I can already do `var a = 3 .. 7`: http://ix.io/2ssv
17:42:41	Zevv[m]	can't do that without support from the language, `for` is not reprogrammable
17:42:50	Zevv[m]	unless you put the whole caller block into a macro call
17:43:32	disruptek	for loop macros are a thing.
17:43:37	Zevv[m]	wait waht
17:43:47	Yardanico	Zevv[m]: yes
17:43:49	*	endragor joined #nim
17:43:50	Zevv[m]	dude!
17:43:51	Yardanico	we have for loop macros
17:43:52	Yardanico	experimental
17:44:15	Zevv[m]	\o/
17:44:15	Yardanico	only in devel I think
17:44:23	Zevv[m]	good enough for me. Let me whip that up
17:44:32	Yardanico	https://nim-lang.github.io/Nim/manual_experimental.html#case-statement-macros-for-loop-macros
17:44:40	disruptek	honestly, these are not limitations afaic.
17:44:48	disruptek	i mean, consider the alternative.
17:44:59	Yardanico	no one stops you from writing "myfor" though
17:45:11	disruptek	!repo foreach
17:45:12	disbot	https://github.com/disruptek/foreach -- 9foreach: 11A sugary for loop macro with syntax for typechecking loop variables 15 6⭐ 0🍴
17:45:36	Zevv[m]	yeah, but you cnat do "myfor a in myit:"
17:45:40	Yardanico	yes
17:45:47	Yardanico	it's valid nim syntax
17:45:52	Zevv[m]	oh dang
17:45:54	Yardanico	so you can make it into real working stuff with macros
17:45:59	Zevv[m]	you just ignore the "in"
17:46:11	disruptek	foreach a, b in pairs(foo) as int, string:
17:46:20	Zevv[m]	let me try the for loop macro
17:46:25	Yardanico	Zevv[m]: https://i.imgur.com/9MKXrwE.png
17:46:48	Zevv[m]	Yardanico: right, like I say, you ignore the 'in'
17:46:57	Zevv[m]	but then you have to call it from within a macro, that's not fair
17:47:01	Zevv[m]	oh no that's not true
17:47:06	Zevv[m]	right
17:47:50	Zevv[m]	hmm but for loop macros require something explicit
17:48:02	Zevv[m]	if my iterator is 'it', i need to make a 'each' and do `for a in each(it)`
17:48:05	Zevv[m]	I can't make `for a in it:`
17:48:08	disruptek	anyway, i'm gonna finish 22 and maybe manage to fix 9 at the same time.
17:48:18	Zevv[m]	nooo gimme my explict cps.call()
17:48:46	Zevv[m]	oh, or, right fix #22 and maybe #9 at the same time
17:49:27	disruptek	give me {.cps: Cont.} which also gives us `cps Cont:`
17:49:48	disruptek	and i think `cps foo()` out-of-context should be a trampoline, too.
17:50:06	Zevv[m]	yeah, I played with that, but that won't work because 'cps' is taken
17:50:13	Zevv[m]	or can you distinguish?
17:50:17	disruptek	why not?
17:50:36	disruptek	one is a procdef, one is a call, one is a type.
17:51:15	Zevv[m]	anyway, that will go away when we do `cps.foo()`
17:51:20	Zevv[m]	there is no free-floating cps anymore
17:51:22	Zevv[m]	which is good
17:51:27	disruptek	sure there is.
17:51:37	Zevv[m]	only the .cps.
17:51:50	disruptek	the .cps. is what does cps Cont: and cps foo().
17:52:06	Zevv[m]	right, only that. But no fake keyword.
17:52:11	disruptek	but whatever, i can write it if it's too hard for you. 😘
17:52:21	Zevv[m]	how is `{.cps: Cont.}` better?
17:52:46	disruptek	i forget.
17:52:50	*	dmi0 quit (Remote host closed the connection)
17:52:51	Zevv[m]	I guess it is. I could only half grasp all this when I wrote it myself
17:52:55	Zevv[m]	but now you wrote it I lost it
17:53:01	*	endragor quit (Ping timeout: 264 seconds)
17:53:17	*	dmi0 joined #nim
17:54:22	disruptek	it seems less important now, but the idea was that foo() out-of-context would yield useful error messages.
17:54:34	Zevv[m]	that's something different
17:54:43	disruptek	now that we are explicit with input and output, it may not be worth worrying about.
17:54:45	Zevv[m]	as it should. But if we do cps.foo(), you can not even call it out of context
17:54:48	Zevv[m]	because there is no out-of-context
17:54:54	Zevv[m]	right'o
17:55:21	disruptek	well, cps.foo() would need to be discarded.
17:55:22	Zevv[m]	co.resume(), it.produce(), evq.recv()
17:55:53	Zevv[m]	oh. Yeah, technically it doesn't be cause it is a fake call
17:55:58	Zevv[m]	but if you look at the signatures, it makes sense
17:56:08	Zevv[m]	and I still want my rvalue
17:56:16	Zevv[m]	I want co.resume() to return something
17:56:26	Zevv[m]	but that's not in the signature of resume()
17:56:30	Zevv[m]	but wait
17:56:31	Zevv[m]	there's more
17:56:40	Zevv[m]	because resume is .cps. we can make that
17:56:45	disruptek	okay, lemme get to work on this other stuff.
17:56:57	Zevv[m]	yeah but I'm just getting started
17:56:59	disruptek	i don't think `let foo := cps.bar()` is too hard.
17:57:11	disruptek	i just want to do exceptions, too.
17:57:39	Zevv[m]	if you put the type in the .cps:Thing., you can drop the explicit return type of the proc
17:57:44	Zevv[m]	and put it on by magic, right
17:58:10	Zevv[m]	so `proc foo(): Cont {.cps.}` would become `proc foo() {.cps:Cont.}`, is that true?
17:58:29	disruptek	maybe we do our casting in the trampoline instead.
17:58:47	Zevv[m]	because I was hoping we could make something like `proc foo(): int {.cps: Cont.}`
17:59:02	Zevv[m]	and then do my `let a = it.foo()`, somehow
17:59:17	Zevv[m]	but indeed, one step at a time
17:59:19	disruptek	it seems really doable to me.
17:59:23	Zevv[m]	\o/
17:59:28	Zevv[m]	Lua coroutines!
17:59:51	disruptek	we need an issue for while: for: while: for: break somewhere
17:59:57	Zevv[m]	what you pass to `resume(...)` goes to `yield`, what you pass to `yield(...)` goes to resume
18:00:20	disruptek	maybe we just use the yield keyword instead of return.
18:00:28	disruptek	it's kinda what we want.
18:00:39	Zevv[m]	true, you can just hijack it in the .cps. I guess
18:01:04	Zevv[m]	but then again, do we need it
18:01:31	disruptek	i think it helps the mental map.
18:01:32	Zevv[m]	`let msg = evq.recv(fd)` or `let msg = yield evq.recv(fd)`
18:01:37	Zevv[m]	then use `await` instead
18:01:43	disruptek	no, we won't use yield there.
18:02:10	Zevv[m]	So there's a choice to be made
18:02:13	disruptek	hmm, maybe we should.
18:02:25	Zevv[m]	will we tell the user what's happening here, make stuff clear and explicit "here you are scheduled out"
18:02:36	Zevv[m]	or hide it all as in goroutines. You want to recv? You go recv.
18:02:44	ForumUpdaterBot	New thread by Serge: Linking neo to openblas.dll ?, see https://forum.nim-lang.org/t/6605
18:03:19	disruptek	if it's magical and there's a crack in the mirror, the illusion is broken.
18:03:36	Zevv[m]	right. So that's the choice
18:03:54	disruptek	yield can be optional, too.
18:03:58	Zevv[m]	bwaah
18:04:21	*	outtabwz quit (Quit: Lost terminal)
18:04:34	Zevv[m]	baby steps
18:05:42	Zevv[m]	can I ask one more thing and then shut up?
18:06:35	Zevv[m]	I think we should drop the concept and make an (empty) base type all CPS's derive from, instead of RootObj
18:07:05	Zevv[m]	it's more clear what you are making there (a CPS thing), and if we ever want to put things in CPS for like debugging etc, we can put it somewhere
18:07:49	*	oriba quit (Quit: https://quassel-irc.org - Chat comfortably. Anywhere.)
18:09:24	disruptek	eh we have a concept.
18:09:26	disruptek	it works.
18:09:32	Zevv[m]	listen to me
18:10:05	Araq	what? CPS needs a base class
18:10:07	Zevv[m]	The concept does not add anything. All iterators will be a ref object that's inheritable anyway
18:10:26	disruptek	the concept adds expectation.
18:10:35	disruptek	cps doesn't need a base class.
18:10:44	disruptek	you supply whatever you want.
18:10:50	disruptek	we use it. we love it.
18:10:53	disruptek	it's elegant.
18:11:02	Araq	need to study it
18:11:05	Zevv[m]	right. But now you or I want to add some -d:cpsDebug logic to the CPS impelmentation, and you want ot be able to store stuff somewhere
18:11:17	disruptek	and i do.
18:11:19	Zevv[m]	if we have a base class (which is just empty usually)
18:11:21	Zevv[m]	we have a place
18:11:26	disruptek	witness cpsDebug and cpsTrace.
18:11:35	Zevv[m]	No, that's for debugging *your particular event queue*
18:11:37	Araq	last time I argued successfully it's never gonna work without inheritance (or closures but these use inheritance too)
18:11:38	Zevv[m]	not for CPS itself
18:11:47	Zevv[m]	it's gonna work
18:12:02	Zevv[m]	disruptek: ok, nevermind, it was just a loose idea, it's not important for now
18:12:05	disruptek	it works without inheritance (but i think inheritance will ultimately make it faster).
18:12:19	Zevv[m]	if you don't use inheritance, you'll need to cast :)
18:12:19	disruptek	also, it currently uses inheritance.
18:12:26	Araq	ah
18:12:47	disruptek	more assumptions are always good when we can get them for free.
18:12:54	Oddmonger	i don't find the != equivalent in nim /o\
18:13:04	Oddmonger	i find only doc with ==
18:13:11	disruptek	!= is a template in system.
18:13:21	disruptek	it's impl as not(a == b)
18:13:51	Oddmonger	ah so != is valid… i was sure it wasn't
18:13:54	Oddmonger	thank you
18:14:10	disruptek	the two things i want to change about cps, that actually matter, are scopes and generics.
18:14:12	FromDiscord	<Rika> if it isnt then maybe the type has no `==`
18:14:41	disruptek	generics buys us type inference and scopes buys us simplicity and correctness, not to mention speed.
18:14:57	disruptek	but anyway, i need to get to work. bbiab
18:15:13	*	someunknownuser quit (Quit: someunknownuser)
18:15:14	Araq	Oddmonger: also covered in our tutorials
18:16:13	disruptek	Zevv[m]: one last thing... did you actually play with the stack traces? it's pretty cool.
18:16:20	Zevv[m]	no, how?
18:16:40	*	tane joined #nim
18:16:48	disruptek	send stuff to my evq trampoline and enable --cpsTrace and raise an exception at the end of a bunch of calls.
18:17:05	Zevv[m]	oh but I never use your trampoline :)
18:17:07	Zevv[m]	Will take a look!
18:23:42	Zevv[m]	hm stupid question - is there a way to "manually" run a Nim closure iterator?
18:23:55	disruptek	sure.
18:24:01	Zevv[m]	how?
18:24:17	disruptek	there's some syntax, it's like python.
18:24:56	Zevv[m]	oh magical finished()
18:25:06	Zevv[m]	got it
18:25:19	Zevv[m]	ah I knew that
18:25:47	Oddmonger	Araq: i'm sure , but didn't find it in the «test» section of tutorial 1
18:26:18	Araq	tutorials are documents you're supposed to read from top to bottom
18:26:58	disruptek	holy shit, i been doin' it dumb all this time.
18:27:17	Zevv[m]	*it*
18:29:09	Zevv[m]	our iterators are only 15 times slower then native nim closure iterators
18:29:28	disruptek	no wonder our web-server is so fast.
18:29:50	Araq	fixed the leak!
18:29:57	Araq	now testing for regressions...
18:32:00	Zevv[m]	Soo, what caused the leak?
18:32:34	Araq	stupid convoluted logic inside liftdestructors.nim
18:33:29	bung	https://github.com/nim-lang/Nim/issues/6136 this should be close, not reproduciable
18:33:31	disbot	➥ Hash function can't handle leading zeroes ; snippet at 12https://play.nim-lang.org/#ix=2ssM
18:37:45	shashlick	what's this observable stores warning
18:38:16	Zevv[m]	yeahy what's this observable stores warning
18:38:40	disruptek	it means your result is mutating despite a possible exception. usually.
18:38:48	Zevv[m]	shashlick: https://github.com/nim-lang/RFCs/issues/230
18:38:51	disbot	➥ Nim's NRVO is C++'s inplace construction (placement new) ; snippet at 12https://play.nim-lang.org/#ix=2oor
18:39:23	Zevv[m]	ooh burn: "The message by the compiler in the latest stable release is now Warning: observable stores to 'x', which is even more cryptic than the original message suggested in this RFC. No link to an article is provided." :)
18:39:34	*	endragor joined #nim
18:39:45	disruptek	they knew just how to hurt me.
18:40:18	Araq	yeah, I know
18:40:26	Araq	will improve the warning message
18:40:39	Araq	could also disable it again since it never found any bug anywhere
18:40:42	Zevv[m]	ooh burn: "The message by the compiler in the latest stable release is now Warning: observable stores to 'x', which is even more cryptic than the original message suggested in this RFC. No link to an article is provided." :)
18:40:46	Zevv[m]	dang sorry
18:40:51	Araq	but let's all be paranoid about everything
18:40:57	shashlick	still I'm too dumb/lazy to know what to do here
18:41:00	disruptek	there was a bug in nimble.
18:41:31	Araq	disruptek: hmm maybe
18:41:47	Araq	nimble relied on the behaviour that we then didn't fix because it caused performance regressions
18:42:08	Araq	so arguably nimble relied on sticky stuff
18:44:05	FromDiscord	<dom96> what did Nimble rely on?
18:44:31	FromDiscord	<Skaruts> is it easy to embed lua with nim? I was looking at this example code, but I can't tell how to adapt to nim, if possible: https://gist.github.com/Vesnica/3672425
18:44:33	disruptek	mutating a result prior to an exception.
18:44:40	disruptek	and then consuming the result.
18:44:42	FromDiscord	<Skaruts> (edit) 'https://gist.github.com/Vesnica/3672425' => '<https://gist.github.com/Vesnica/3672425>'
18:44:55	FromDiscord	<dom96> ahh yes, I remember that one
18:45:08	*	endragor quit (Ping timeout: 256 seconds)
18:48:22	Araq	Skaruts: It's pretty simple, there should be examples in the Nim Lua wrapper
18:49:27	FromDiscord	<Skaruts> @Araq this one? <https://github.com/jangko/nimLUA>
18:49:40	Araq	iirc there is more than one
18:49:52	Zevv[m]	disruptek: were too slow, really too slow
18:50:35	Zevv[m]	outrun 15 times by closure iterators. I expected 3, maybe 5. but 15
18:50:55	FromDiscord	<Skaruts> at a glance I only found that one and one fork of it
18:50:59	FromDiscord	<Rika> 3 times 5 😉
18:51:38	Araq	Zevv[m]: huh? shouldn't it be faster than closure iterators?
18:51:52	Zevv[m]	nope. with --arc we're 7 times slower still
18:53:41	Zevv[m]	well, we trampoline one time too many, due to the way the stuff is split for now, so that might be solvable
18:53:59	Zevv[m]	but of course for each in/out jump we do a fresh continuation alloc
18:54:25	Zevv[m]	which is kind of sucky
18:54:44	Zevv[m]	we should be reusing these somehow
18:54:48	disruptek	see, i had hoped that with inheritance we could do a realloc and then omit some assignments.
18:55:18	Zevv[m]	that's not up to us. We just create a new thingy at the end of each split part
18:55:30	disruptek	no, it's up to us.
18:55:32	disruptek	why is it not?
18:55:42	disruptek	we can pass whatever we want.
18:55:55	Zevv[m]	oh so if your env groes if you go deeper, you realloc and add stuff
18:56:03	disruptek	of course.
18:56:09	Zevv[m]	of your env shrinks if you go up, you can even choose not to realloc
18:56:37	disruptek	so if only control flow changes, we just `result.fn = foo; return`
18:56:38	*	Jesin quit (Quit: Leaving)
18:56:40	disruptek	that's pretty fast.
18:56:46	Zevv[m]	if you know how deep you can go, you should realloc the whole thing on the first invication
18:56:49	disruptek	faster than an alloc. 😉
18:56:53	Araq	here is my advice: focus on the performance right now. otherwise you repeat my mistake where you debug a slow ARC implementation, then add optimizations and then debug it all once again
18:57:16	disruptek	sure thing, boss.
18:57:20	disruptek	well anyway, that's the idea of the inheritance.
18:57:36	*	NimBot joined #nim
18:57:46	disruptek	we'll figure out how best to use it eventually.
18:57:51	disruptek	but it clearly buys us something.
18:58:03	disruptek	and remember, this is an object that the user defined. completely.
18:58:19	FromDiscord	<dom96> what performance are you expecting out of this?
18:58:20	disruptek	just think about that, i mean, really.
18:58:22	*	nikita` quit (Read error: Connection reset by peer)
18:58:30	disruptek	pretty damned fast.
18:58:33	Zevv[m]	disruptek: euh wait, you gather the contents of that thing during lifting, right?
18:58:40	Zevv[m]	it's the locals
18:58:47	disruptek	right.
18:58:51	Zevv[m]	so how is that "user defined"
18:59:05	disruptek	the user provided the Cont type, wholecloth.
18:59:13	disruptek	it can be anything they want.
18:59:14	Zevv[m]	yeah right, but not "the whole thing"
18:59:23	Zevv[m]	it's that, plus what you put on top of it
18:59:47	disruptek	i mean, technically, we could add a pointer to the object definition and now the continuation is one piece of memory.
19:00:07	Zevv[m]	that's an extra dereference?
19:00:11	disruptek	right.
19:00:13	Zevv[m]	don't
19:00:16	disruptek	to wherever the continuation is destined.
19:00:34	disruptek	wait, it's not an extra dereference.
19:00:42	Zevv[m]	then I don't understand yet
19:00:56	disruptek	well, a continuation right now.
19:00:58	disruptek	what is it?
19:00:59	Zevv[m]	right now
19:01:02	disruptek	yes.
19:01:15	disruptek	like, simple answer.
19:01:16	Zevv[m]	it's an object of type X wich has children with extra genes
19:01:24	Zevv[m]	ref object
19:01:32	Zevv[m]	and a pointer to a proc
19:01:40	disruptek	no, it's a ref object that inherits and has a ptr to a fn.
19:01:52	Zevv[m]	right that's what I mean
19:01:54	Zevv[m]	said
19:01:55	disruptek	so it's an object that can be extended, and it points to some code.
19:02:08	Zevv[m]	right. The env extends it
19:02:11	disruptek	but, like, you don't extend this object.
19:02:14	Zevv[m]	I do
19:02:19	disruptek	no, you truly do not.
19:02:20	Zevv[m]	oh
19:02:21	disruptek	i promise you.
19:02:23	Zevv[m]	tell me
19:02:31	disruptek	it's just a concept.
19:02:37	disruptek	the concept is the words on the screen.
19:02:42	disruptek	same as i've repeated to you.
19:02:46	Zevv[m]	repeat it once more
19:02:55	disruptek	it's a ref to an obj that inherits and has a ptr to a piece of code.
19:03:16	disruptek	a piece of code that knows how to run with nothing more than some data as input.
19:03:23	Zevv[m]	ok but then again
19:03:24	disruptek	some data it may or may not mutate.
19:03:30	disruptek	add to.
19:03:35	disruptek	rewrite.
19:03:43	Zevv[m]	if you go deeper into the original stack, you inherit and inherit and keep adding stuff to the inherited type
19:03:46	Zevv[m]	is that right or wrong
19:03:58	disruptek	you can strip stack whenever you pass it.
19:04:04	Zevv[m]	is that right or wrong
19:04:06	disruptek	you only pass what you want to use.
19:04:07	disruptek	that's good.
19:04:23	disruptek	obviously, you pass the fn.
19:04:27	disruptek	so that's 8 bytes.
19:04:28	Zevv[m]	no go back
19:04:33	Zevv[m]	if you go deeper into the original stack, you inherit and inherit and keep adding stuff to the inherited type, is that right
19:04:38	disruptek	no.
19:04:49	Zevv[m]	so where am I wrong. I mean as it is *now*
19:04:51	disruptek	whenever you create a proc, that is when you define the environment.
19:05:04	disruptek	the environment that you're working in has whatever you want in it.
19:05:09	disruptek	any subset of your input.
19:05:23	disruptek	put anything at all in this env; it's yours to keep.
19:05:32	Zevv[m]	you totally lost me, sorry
19:05:33	disruptek	no one will have access to it.
19:05:48	disruptek	zevv, YOU are `fn`.
19:05:51	Zevv[m]	back. Talking as it is *now* or talking as you want it to *be*
19:05:57	disruptek	this is how it is now.
19:05:59	Zevv[m]	ok
19:06:14	disruptek	so you, YOU, are fn.
19:06:20	Zevv[m]	ok
19:06:20	disruptek	i give you some input.
19:06:23	disruptek	a banana
19:06:25	Zevv[m]	yeah
19:06:26	disruptek	a coconut.
19:06:28	Zevv[m]	also
19:06:38	disruptek	a gallon of rum.
19:06:45	Zevv[m]	yeah yeah go on
19:07:23	disruptek	you give me the name of a friend and a cocktail that i can give to them.
19:07:38	Zevv[m]	sure
19:07:43	disruptek	then we see if i'm getting lucky tonight.
19:07:59	disruptek	and so on and intercourse.
19:08:03	*	maier joined #nim
19:08:03	disruptek	er, so forth.
19:08:46	disruptek	let's say i don't use the banana. i give the banana back to you.
19:09:02	disruptek	look, i say, don't go to this friend. go to this other friend; she's love the banana.
19:09:20	disruptek	she gets the banana, i get the cocktail, you get fucked.
19:09:22	disruptek	or something.
19:09:27	disruptek	who cares, it doesn't matter.
19:09:28	Zevv[m]	dude
19:09:37	disruptek	the point is that you can use it or lose it or pass it.
19:10:15	Zevv[m]	ok. real example, gimme a sec
19:10:24	disruptek	maybe you took the banana and gave someone else your nut.
19:10:33	disruptek	whatever peels your banana, dude.
19:10:45	disruptek	i am not going to judge you on your banana.
19:10:50	disruptek	i'm more of a coconut man, myself.
19:10:55	Zevv[m]	real stuff: I have a proc with a few nested scopes, right
19:10:58	Zevv[m]	outer scope has only int i
19:10:58	*	Jesin joined #nim
19:11:00	disruptek	right.
19:11:03	Zevv[m]	one scope deeper int j comes in
19:11:05	*	dzamo quit (Quit: dzamo)
19:11:13	disruptek	intj's, know 'em. love 'em.
19:11:15	Zevv[m]	halfway first cps call I have env_1, which contains only int i
19:11:23	disruptek	right.
19:11:28	Zevv[m]	now we go one deeper, we get env_2, which inherits env_1 and adds int j
19:11:30	disruptek	i'm right there whitcha, boss.
19:11:30	Zevv[m]	still true?
19:11:40	disruptek	yes, yes, yes.
19:11:47	Zevv[m]	so the deeper you go, the further you inherit, the more you add
19:11:51	Araq	seriously? bananas and nuts?
19:11:53	disruptek	you put the lime in the coconut.
19:11:56	Zevv[m]	still true?
19:11:57	disruptek	you shake it all up.
19:12:01	disruptek	yes, yes.
19:12:07	Araq	stop it. use better terms
19:12:15	Zevv[m]	ok, earlier when I asked that, you said that was wrong
19:12:26	Zevv[m]	so you add to your inherited environment by inheriting deeper and adding members
19:12:38	Zevv[m]	you go a few scopes up, and you have another branch of inheriting
19:12:39	disruptek	you have i and j. but when you tell me the result, you can tell me as much or as little as you want.
19:12:46	disruptek	you can tell me i.
19:12:48	disruptek	you can tell me j.
19:12:52	Zevv[m]	shut up
19:12:52	disruptek	you can tell me k.
19:12:54	Zevv[m]	you go a few scopes up, and you have another branch of inheriting
19:12:59	disruptek	fine.
19:13:03	Zevv[m]	so, first call we construct env_1 with only i on it
19:13:09	Zevv[m]	gets malloced and on the heap
19:13:14	disruptek	fine.
19:13:17	Zevv[m]	we go to the first split fn, and pass that thing in
19:13:23	disruptek	stop.
19:13:25	*	maier quit (Ping timeout: 264 seconds)
19:13:26	Zevv[m]	it just throws it away, and constructs a new env_2
19:13:27	Zevv[m]	ok
19:13:34	disruptek	we got to the first split fn.
19:13:37	disruptek	where is that fn?
19:13:44	disruptek	where is the ptr to that proc?
19:13:45	Zevv[m]	the pointer to it is in the continuation
19:13:52	disruptek	okay, so it's in "our" memory.
19:13:59	disruptek	the memory we know how to find.
19:14:02	disruptek	right?
19:14:08	disruptek	we know the continuation.
19:14:11	disruptek	it was passed to us.
19:14:16	disruptek	so we know where the fn is on it.
19:14:17	disruptek	right?
19:14:19	disruptek	with me?
19:14:28	disruptek	the fn, of course, is ourselves.
19:14:35	disruptek	so we can identify ourselves.
19:14:42	disruptek	and we know, what stuff, in the world is around us.
19:14:44	disruptek	memory.
19:14:49	disruptek	we know what's in that memory.
19:14:52	disruptek	it's variables.
19:14:54	disruptek	we have the list.
19:14:56	Zevv[m]	right
19:14:58	disruptek	it's in our type.
19:15:01	Zevv[m]	right
19:15:10	disruptek	in fact, everything we're going to speak of, is in our type.
19:15:19	disruptek	all symbols that might be interesting to know about.
19:15:27	disruptek	any name->value sorta thing.
19:15:27	Zevv[m]	but as we pass the continuation to the next, we create a whole new continuation
19:15:35	Zevv[m]	that's what happens *now*
19:15:36	disruptek	but also anything in global scope that's like that.
19:15:42	disruptek	yes.
19:15:45	disruptek	that happens now.
19:15:47	Zevv[m]	right
19:15:51	Zevv[m]	that is what I wanted to say
19:15:51	disruptek	and what do we put into that object?
19:16:07	disruptek	do we put the whole world into it? all the memory we've amassed?
19:16:07	Zevv[m]	all the stuff that is in scope of the next one to receive this
19:16:19	disruptek	but who chooses what that is?
19:16:23	Zevv[m]	the receiver
19:16:25	disruptek	how do you pick?
19:16:29	Zevv[m]	it was picked for me
19:16:31	Zevv[m]	I don't know
19:16:43	disruptek	okay, so someone says to you, here, let's do business.
19:16:51	disruptek	they tell you what they want.
19:16:56	disruptek	you give them a result.
19:17:04	disruptek	right?
19:17:07	Zevv[m]	right
19:17:08	disruptek	they tell you.
19:17:34	disruptek	they say, "i don't care where you're from, who you are. or what you know."
19:17:41	disruptek	"i want x, y, z."
19:17:46	disruptek	"that's all i want. i don't care."
19:17:57	Zevv[m]	right, so I create another env type and put that stuff in there
19:17:59	disruptek	"give me x, y, z. and i will give you the world."
19:18:05	disruptek	right.
19:18:07	Zevv[m]	but that new env, is new and goes on the heap
19:18:08	Zevv[m]	another alloc
19:18:15	disruptek	right, but it could be a realloc.
19:18:25	disruptek	in fact, aren't there cases where it doesn't need to change?
19:18:33	Zevv[m]	right, that is where I want to go
19:18:38	Zevv[m]	but you went on your coconut tour
19:18:40	disruptek	it's it the case that sometimes you will get a request and just pass it somewhere else?
19:18:45	Zevv[m]	and also, if you know the deepest you could ever get
19:18:53	disruptek	it doesn't matter.
19:18:54	Zevv[m]	you could make 1 thing that you just pass all over, and mutate it
19:19:02	disruptek	sure, but it doesn't matter.
19:19:03	Zevv[m]	you alloc once, and everybody only uses what they need
19:19:06	Zevv[m]	no reallocs
19:19:09	Zevv[m]	no growing, no shrinking
19:19:11	disruptek	sure, whatever.
19:19:26	Zevv[m]	it can still be a hierarchy of different objects, but you can *start with the biggest possible thing
19:19:28	Zevv[m]	*
19:19:34	disruptek	if you want to work is a certain sized swimming pool, one way to do that is to get a lot of water in one place at the same time.
19:19:54	Zevv[m]	dude talk nim, I'm too stupid for that stuff
19:20:10	disruptek	you can do whatever you want with this memory.
19:20:22	Zevv[m]	yeah yeah but as it is *NOW*
19:20:23	Zevv[m]	it is no good
19:20:30	disruptek	why?
19:20:34	Zevv[m]	all the frigging reallocs
19:20:35	disruptek	arc sinks these values for us.
19:20:43	disruptek	oh, as it is NOW.
19:20:45	disruptek	yes, yes.
19:20:47	disruptek	of course.
19:20:47	Zevv[m]	YEEESSSS
19:20:50	disruptek	sorry.
19:20:50	Zevv[m]	LISTEN YOU MORON
19:20:52	Zevv[m]	I typed *NOW*
19:20:54	Zevv[m]	right
19:20:55	Zevv[m]	soooo
19:20:55	disruptek	sorry.
19:20:57	Zevv[m]	I love you too
19:20:59	Prestige	Is there a way to have a variable be a sort of alias for another var that is a ref object?
19:20:59	Zevv[m]	need to read the kids
19:21:06	Zevv[m]	doing quantum mechanics tonight
19:21:08	disruptek	okay.
19:22:19	Prestige	Trying to replicate this behavior: https://github.com/avahe-kellenberger/dwm/blob/master/src/dwm.c#L962 - my Client is a ref object of RootObj
19:24:36	disruptek	why do we even bother with arc?
19:24:49	disruptek	why don't we just manage the memory for you?
19:24:55	disruptek	we're doing that anyway.
19:25:09	disruptek	i mean, we're alloc'ing this stuff.
19:25:11	disruptek	we're copying it.
19:25:21	disruptek	we know it by name.
19:25:28	disruptek	we know what it is.
19:25:32	disruptek	it's not a guess, we KNOW.
19:26:00	leorize	Prestige: I'm not sure how this is supposed to work
19:27:01	Prestige	I believe that c->mon->clients is being reassigned
19:27:35	Prestige	clients is just a Client that has a pointer to the next client
19:28:24	Prestige	So I think tcc is sort of like an alias here, for c->mon->clients
19:29:11	leorize	yes
19:30:26	Prestige	I'm wondering if this is possible in Nim, where my Client is a ref object
19:31:05	leorize	yes
19:32:22	leorize	you just need a `ptr Client` to serve as your alias
19:32:33	leorize	`std / byaddr` have a pragma to do just that
19:33:21	disruptek	Zevv[m]: i think we can do ideal-alloc with just a single pointer. should be like one line of code.
19:33:38	Prestige	Thanks, I'll give it a try
19:38:40	Zevv[m]	'ideal-alloc' meaning the right size so we do not have to realloc any time during the lifetime of the whole chain of continuations?
19:39:11	*	aurelius joined #nim
19:39:15	disruptek	not, like, literally ideal, but based upon name clashes.
19:39:57	Zevv[m]	ok, so where you want to go: instead of throwing away the continuation with its env, you want to mutate it and pass that on
19:39:59	disruptek	so, like, it's not about size, it's about allocs.
19:40:01	Zevv[m]	so we don't realloc, right?
19:40:19	Zevv[m]	no, "so we dont throw away and allocate a whole new one"
19:40:22	disruptek	right.
19:40:41	Zevv[m]	right.
19:40:42	disruptek	i mean, currently.
19:40:58	Zevv[m]	but now we *do* make a new one every call
19:41:09	disruptek	yeah.
19:41:12	Zevv[m]	ok
19:41:42	disruptek	how can we get the static sizes of all this stuff?
19:41:48	Zevv[m]	so, do we make this thing just a block of memory which we pass along and grow when needed, and just cast it to the type it happens to have during this place in its lifetime?
19:42:08	Zevv[m]	or do we figure out the whole possible tree of inheritance, and pre-alloc the largest of those
19:42:13	Zevv[m]	and pass that around, so that any other type would fit in
19:42:44	disruptek	we can compute size at runtime, certainly.
19:43:04	Zevv[m]	if my mental image is right, every var that goes into an env has the same location in the memory block in all possible tree-permutations
19:43:15	disruptek	so we'll be able to make the assumption at compile-time that we got the size right.
19:43:15	Zevv[m]	so we never have to move stuff around. The outer scope 'i' will always be at the same offset
19:43:41	disruptek	that's fine.
19:43:51	disruptek	i'm just thinking that it doesn't matter.
19:43:55	Zevv[m]	The only downside will be that if you stash away a million of these things for later calling, you possible stash away a lot of unused memory
19:43:56	*	bung quit (Quit: Lost terminal)
19:43:58	disruptek	if i know how big i am, i can tell you where i end.
19:44:14	Zevv[m]	but I think araq is right, we should make this work soon
19:44:14	disruptek	so whatever.
19:44:41	*	vicfred_ joined #nim
19:45:02	disruptek	really, it's just a chunk of memory Cont[N].
19:45:38	Zevv[m]	right'o
19:45:46	disruptek	we setup the vars for you. you can use any string to name them.
19:45:56	*	vicfred_ quit (Client Quit)
19:46:05	*	aurelius quit (Ping timeout: 240 seconds)
19:46:14	*	vicfred_ joined #nim
19:46:43	*	altarrel quit (Quit: Konversation terminated!)
19:47:28	*	vicfred quit (Ping timeout: 256 seconds)
19:48:04	*	MightyJoe joined #nim
19:50:34	*	cyraxjoe quit (Ping timeout: 265 seconds)
19:50:40	Zevv[m]	I still don't know what you're saying here, but I think we mean the same thing about the memory block used for continuations
19:51:25	*	endragor joined #nim
19:51:30	Zevv[m]	does it make sense to alias locals to directly use the stuff stored in the env, or is nim smart enough to understand we're peeking it from there and poking it back at the end?
19:51:38	Zevv[m]	otherwise that could be a good optimization as well.
19:51:52	Zevv[m]	just rewrite `i` to `myenv.i`
19:51:56	disruptek	i'm thinking that anything stropped.
19:52:11	disruptek	so it's `i` versus 'i'.
19:52:32	disruptek	we just rewrite those. 🤷
19:52:44	Zevv[m]	eh wait. My original code does "i = 3"
19:52:48	Zevv[m]	waht do you rerwrite that to
19:52:55	disruptek	myenv.i = 3
19:52:59	Zevv[m]	right
19:53:04	Zevv[m]	where does the stropping come in
19:53:15	disruptek	so, i mean if it's stropped.
19:53:24	disruptek	i = 3 means, local i = 3.
19:53:37	disruptek	`i` = 3 means myenv.i = 3
19:53:37	Zevv[m]	ok
19:53:57	disruptek	it's cute, too.
19:54:09	Zevv[m]	so, we can keep the inheritance, or make a tons of objects that are tree-wise compatible
19:54:16	Zevv[m]	which at the C-level is about the same thing
19:54:49	disruptek	you make a fair point.
19:54:55	*	MightyJoe quit (Ping timeout: 265 seconds)
19:55:06	*	cyraxjoe joined #nim
19:55:52	Zevv[m]	but to make sure: we have an outer scope with i and j, next scope uses i, j and k so we add one. One deeper j is not used, so it only needs i and k. We still pass the whole thing, but the hole were j was is just not accessed
19:55:56	Zevv[m]	does that make sense
19:56:10	Zevv[m]	it's just the same memory block, but we must make sure that layout-wise k is always on the same address
19:56:20	disruptek	yes, of course.
19:56:33	Zevv[m]	so if you go to deeper scope, you still need to put in stuff from outer scopes, evne it if it is not used
19:56:55	Zevv[m]	there's a beauty to this stuff
19:57:03	disruptek	yes, but remember it's not just about used.
19:57:10	disruptek	you have to worry about the clashes, too.
19:57:13	Zevv[m]	I see these things in my minds eyes, but I couldn't draw a picture of it if I had to
19:57:19	Zevv[m]	clashes of what
19:57:23	Zevv[m]	names?
19:57:25	disruptek	yeah.
19:57:29	Zevv[m]	yust number them?
19:57:33	Zevv[m]	i_1, i_2, i_3?
19:57:42	Zevv[m]	in your 3d part you rewrite `i` to `env.i_3`
19:57:45	Zevv[m]	or whatever
19:57:53	disruptek	maybe.
19:57:54	Zevv[m]	they all live on the exact same place in the memory block
19:58:04	Zevv[m]	just with differetn names
19:58:18	*	endragor quit (Ping timeout: 265 seconds)
19:58:26	disruptek	but now you're saying that we'll find the largest chunk that your continuation ever has and make that the size of every continuation.
19:58:52	Zevv[m]	maybe. It avoids realloc during its life time, but it is overhead you might need to keep stashed away for a longer time
19:59:45	Zevv[m]	the good thing is that if you have a bunch of the same continuations, we could use some kind of presized memory pool and get 0 fragmentation
20:00:01	Zevv[m]	In real life, how big is your typical stack frame?
20:00:09	disruptek	well, we can certainly do that.
20:00:12	Zevv[m]	in my C time these were always pretty small, handfull of KB's
20:00:21	Zevv[m]	but with nim I tend to put much more stuff on the stack, so they might grow
20:00:51	Zevv[m]	but then again, as soon as your data structures get more complex to store more stuff, you're more likely to use refs and things go to the heap
20:01:12	Zevv[m]	so in practice I think closure size will exponentially go down pretty fast
20:01:26	Zevv[m]	most will be handful or tens of bytes, for simple iterators and stuff
20:02:15	disruptek	heap or not, you have the size of the memory to think about.
20:02:44	Zevv[m]	we could code it up so the user has a choice, somewhere somehow
20:02:51	disruptek	you cannot predetermine what people will do.
20:03:02	disruptek	why they will do the things they do.
20:03:03	Zevv[m]	either pre-alloc the full size, or grow and/or shrink during its life time
20:03:17	disruptek	the terrible, naughty, kinky things they do.
20:03:28	Zevv[m]	I try to understand the things mratsim does sometimes, and I think he's all for preallocs and pools and all that
20:03:33	Zevv[m]	you want to go fast, you pay
20:03:36	Zevv[m]	makes perfect sense
20:03:47	disruptek	i know.
20:03:53	disruptek	nothing is more expensive than memory.
20:04:03	Zevv[m]	so a continuation will be as large as the original funcions worst case stack size
20:04:19	disruptek	but i'm not comfortable with that.
20:04:26	Zevv[m]	then you need to realloc in flight
20:04:28	disruptek	why do you want to give that up?
20:04:31	Zevv[m]	but that'll cost performance
20:04:36	disruptek	hol' up.
20:04:39	disruptek	who gets to choose?
20:04:46	disruptek	why can't i say when i pay that penalty?
20:04:47	Zevv[m]	the implementer of the thing built on CPS
20:04:58	Zevv[m]	or the end user of the thing
20:05:04	disruptek	it's it the same to say that you can always go smaller?
20:05:12	disruptek	it's that safer?
20:05:20	Zevv[m]	don't understnd the question
20:05:26	Zevv[m]	"huh"
20:05:33	disruptek	you can pass memory the same size or smaller.
20:05:54	Zevv[m]	lost you again - who passes to who
20:06:16	disruptek	you, the cont.fn can return memory the same size, or smaller.
20:06:27	Zevv[m]	right
20:06:36	Zevv[m]	or larger
20:06:49	Zevv[m]	so either we start out at max size
20:06:56	Zevv[m]	or we do reallocs on every pass
20:07:03	Zevv[m]	that's a compile time choice in your macro
20:07:09	Zevv[m]	just inject a realloc or not
20:07:33	disruptek	hmm, i dunno.
20:07:45	Zevv[m]	¿problem?
20:07:58	disruptek	i dunno if i want that to be compile-time.
20:08:02	Araq	it makes me mad
20:08:10	disruptek	i mean, look at araq.
20:08:13	Araq	5000 allocations because of 'of RootObj'
20:08:25	disruptek	that's a lot.
20:08:29	Araq	so strange, valgrind says everything is not corrupted
20:08:43	disruptek	what was the actual problem?
20:17:36	Zevv[m]	disruptek: what do you propose then
20:17:57	Zevv[m]	so either you have the max size prealoced or keep it, or you realloc shrink/grow while traveling
20:18:18	disruptek	i think when you add to an env, you set your parent's "ideal" to yourself. clearly.
20:18:26	*	MightyJoe joined #nim
20:18:40	disruptek	there's no way to delete from envs, so it seems pretty correct.
20:18:52	disruptek	it just recurses.
20:19:14	Zevv[m]	you can't delete from envs, but when you get down the end of the call chain, where the original fucntion would bubble up scopes, you could lose the tail
20:19:15	FromDiscord	<Clyybber> whats crakcin
20:19:28	Zevv[m]	shrinking back to 0 as you ascend back to the surface
20:19:47	Zevv[m]	Clybber: we're way to slow and need to get rid of the environment allocs for every pass
20:19:47	disruptek	bubble up scopes?
20:19:56	Zevv[m]	wrong wording
20:20:07	Zevv[m]	a proc has a bunch of nested scops. Ifs, blocks whiles
20:20:13	Zevv[m]	if you are deep, you have a large stack = large environment
20:20:13	FromDiscord	<Clyybber> Zevv: Hmm, there is a PR for putting them on the stack
20:20:26	disruptek	yeah?
20:20:29	Zevv[m]	if you get out of one of these nested scopes, you lose a bit of your steack, smaller env
20:20:30	FromDiscord	<Clyybber> oh, eh
20:20:46	Zevv[m]	so the continaiton env can shrink sometimes
20:21:02	disruptek	how do i lose a bit of my env?
20:21:08	*	cyraxjoe quit (Ping timeout: 256 seconds)
20:21:13	Zevv[m]	the tail
20:21:27	Zevv[m]	it's as big as the stack would be if you didn't chop up the proc
20:21:31	Zevv[m]	stacks grow and shrink all the time
20:21:37	disruptek	these don't.
20:21:56	Zevv[m]	you nest into an if and declare a var, you env grows. You get out of the if block, that var is no longer in scope, your stack shrinks.
20:22:03	Zevv[m]	So any calls happening after that have a smaller env
20:22:49	disruptek	no, when i setup the continuation, i map every growth to point every prior env to myself.
20:23:14	disruptek	ie during the process where i register locals.
20:23:24	FromDiscord	<Clyybber> Araq: Hmm why does the inheritance cause allocs?
20:23:28	*	oriba joined #nim
20:23:29	FromDiscord	<Clyybber> Maybe because of the RTTI?
20:23:29	FromGitter	<alehander92> Araq so can't you diff
20:23:34	*	vicfred_ quit (Quit: Leaving)
20:23:46	FromGitter	<alehander92> the alloc callsites
20:23:48	Araq	found it
20:23:55	Zevv[m]	tell it
20:23:56	*	audiofile joined #nim
20:24:02	disruptek	yeah, jeeze.
20:24:04	*	vicfred joined #nim
20:24:04	Araq	when we use inheritance the cycle collector assumes there can be cycles
20:24:17	FromDiscord	<Clyybber> And that causes leaks/
20:24:19	FromDiscord	<Clyybber> ?
20:24:19	Araq	and that causes init/deinit calls for its internal data structure
20:24:27	FromDiscord	<Clyybber> or are you chasing leaks?
20:24:30	Araq	it doesn't leak, but it allocates more
20:24:34	FromDiscord	<Clyybber> ah ok
20:24:39	Araq	the leaks are fixed in my other PR
20:24:46	FromDiscord	<Clyybber> oh nice, gotta check it out
20:25:02	Araq	so now I know. only the cursor inference left and then I can focus on bootstrapping
20:25:12	Araq	4 days left
20:25:20	disruptek	lol
20:25:27	disruptek	you can do it, dude.
20:26:13	Araq	hmm I might want to optimize this though
20:26:17	disruptek	maybe if we try hard we can keep from filing more arc bugs between then and now.
20:28:43	FromDiscord	<dom96> Araq: don't burn yourself out. Now that you know the cause it would be a good time to stop and rest until tomorrow 🙂
20:29:25	FromDiscord	<Clyybber> nah, arc fixing is a drug
20:29:32	FromDiscord	<Clyybber> a drug with benefits
20:29:34	FromGitter	<alehander92> and drug addiction is bad
20:29:46	Zevv[m]	all drugs have benefits
20:29:53	FromGitter	<alehander92> say no to drugs clybber
20:29:55	FromGitter	<alehander92> choose life
20:30:01	FromDiscord	<Clyybber> a drug without disadvantages
20:30:25	FromDiscord	<Clyybber> drug: no
20:30:38	FromDiscord	<Clyybber> there I said it
20:30:49	FromDiscord	<Clyybber> even pinged the douche
20:31:05	FromGitter	<alehander92> choose refining optimization
20:32:09	FromDiscord	<Clyybber> Araq: Is there already a proc to compare idents discarding the gensym part?
20:32:53	FromDiscord	<Clyybber> I think equalParams needs that
20:33:54	shashlick	all calls in the stdlib to tables.add need to be cleaned up - stdlib is grumbling about itself
20:37:57	disruptek	wow, this would give me my `while: else:` feature for free.
20:38:35	Zevv[m]	what would
20:39:52	FromDiscord	<Clyybber> oh no
20:39:55	disruptek	the fact that break is not flow-through control flow.
20:40:39	FromDiscord	<Clyybber> inb4 while: else: gets built into cps
20:41:52	FromDiscord	<KingDarBoja> Ar4q must be happy after seeing how Python 3.10 is going to implement Pattern matching (because he was right)
20:41:58	FromDiscord	<KingDarBoja> 🤔
20:42:13	FromDiscord	<Clyybber> Araq: Should we actually gensym params of procs that are {.inject.} ?
20:42:28	disruptek	oh
20:42:42	disruptek	that would be awesome.
20:43:44	FromDiscord	<Clyybber> what would?
20:45:52	*	hoijui quit (Ping timeout: 260 seconds)
20:46:03	*	rockcavera joined #nim
20:47:52	*	tane quit (Quit: Leaving)
20:48:55	disruptek	well, you get more than break.
20:49:36	disruptek	you get continue, no-start, and post-work exit, right?
20:49:48	Araq	KingDarBoja: I don't care what Python does, it's a silly scripting language
20:50:24	Araq	Clyybber: ask me again tomorrow, cannot think right now, it's late
20:50:35	FromDiscord	<Clyybber> ok, haha
20:54:03	FromDiscord	<Clyybber> gn8
20:54:09	disruptek	cps lives in the corners of your while loops.
20:54:15	disruptek	it comes for you. it comes.
20:54:38	FromGitter	<alehander92> no please
20:54:51	FromGitter	<alehander92> gentrification
20:55:00	disruptek	just a tiny programming-language virus. i swear it'll be tiny.
20:55:05	FromGitter	<alehander92> we've had enough of this
20:55:19	disruptek	nim lives matter.
20:56:04	FromGitter	<alehander92> your life matters
20:58:08	*	endragor joined #nim
21:02:06	*	lritter joined #nim
21:02:36	Zevv[m]	yay, finally I have goto's in Nim: https://github.com/disruptek/cps/blob/master/stash/goto.nim
21:05:04	Zevv[m]	this is all so silly
21:05:06	*	narimiran quit (Ping timeout: 256 seconds)
21:07:25	*	endragor quit (Ping timeout: 264 seconds)
21:08:56	*	maier joined #nim
21:10:45	FromDiscord	<Clyybber> thats ugh
21:10:52	FromDiscord	<Clyybber> fucking amazing haha
21:11:43	Zevv[m]	I know I should be able to build a try/catch/throw with this, but I can't
21:13:45	*	maier quit (Ping timeout: 240 seconds)
21:19:05	*	dulsi_ joined #nim
21:19:40	*	sputny quit (Remote host closed the connection)
21:20:57	*	apahl quit (Ping timeout: 272 seconds)
21:21:12	*	dmi0 quit (Quit: Leaving)
21:21:23	*	apahl joined #nim
21:21:55	*	dulsi quit (Ping timeout: 265 seconds)
21:22:45	*	Trustable joined #nim
21:25:13	Zevv[m]	disruptek: you here still
21:26:27	FromDiscord	<Avatarfighter> disruptek is lost in the for loops
21:26:40	Araq	try and catch is easy, Zevv[m]
21:26:54	Zevv[m]	yeah, it's easy but it doesn't fit our current infra
21:26:56	Araq	see how --exceptions:goto do it
21:26:59	Zevv[m]	I need state to carry stuff over closures
21:27:03	Zevv[m]	it's just an implementation thing
21:27:06	disruptek	it does fit.
21:27:16	Zevv[m]	it fits, but I can't store my state anywhere right now
21:27:17	disruptek	it kills two birds with one stone, actually.
21:27:37	Zevv[m]	I want to store something in my Cont that carries over to all successive conts
21:27:45	disruptek	go ahead.
21:27:46	Zevv[m]	a ref to something
21:27:54	Zevv[m]	yeah but not sure how to nicely pass it in initially
21:27:54	disruptek	yeah, it's called e.
21:28:05	disruptek	the same way you pass anything.
21:28:45	disruptek	we gensym "fn" and "e".
21:29:02	Zevv[m]	oh right
21:29:03	disruptek	i don't feel bad about this.
21:29:08	Zevv[m]	i didn't know
21:29:12	disruptek	throw result in there, too.
21:29:22	Zevv[m]	dude your documentation is *so* bad
21:29:34	disruptek	i'm trying to write some now.
21:29:54	Zevv[m]	e is not in your concept
21:30:09	disruptek	no, neither is result.
21:31:41	Zevv[m]	where do I new() my e?
21:31:54	disruptek	we can probably put `punch` in pretty trivially.
21:32:14	disruptek	prettier nimions new it for you.
21:32:31	disruptek	or not at all.
21:32:37	disruptek	who's to say?
21:32:42	FromDiscord	<Clyybber> Araq: Kind of related, wdyt about making the gensym hash part of an ident consistent per template?
21:33:31	Araq	what does that mean
21:33:59	disruptek	it's caught from below.
21:34:01	FromDiscord	<Clyybber> It means that i and i in the same template will be gensymmed to the same
21:34:47	FromDiscord	<Clyybber> i`gensymSAME
21:35:59	Zevv[m]	disruptek: you're speaking in riddles
21:36:01	Zevv[m]	with your 'e'
21:36:06	FromDiscord	<Clyybber> This would allow us to get rid of the symbol searching we do
21:36:50	FromDiscord	<Clyybber> to make `when ...: var a else: var a\necho a` work
21:37:43	*	dulsi_ is now known as dulsi
21:37:44	FromDiscord	<Clyybber> and would make gensymmed forward decls work too without adding a mode that ignores gensym param name differences to searchForProc
21:38:31	Araq	listen to me
21:40:28	Araq	touch the gensym implementation as little as possible it took me at least 3 iterations to get into its current, semi-working state
21:41:26	Zevv[m]	Aller guten Dinge sind drei
21:48:48	*	solitudesf quit (Ping timeout: 256 seconds)
21:49:48	*	Vladar quit (Quit: Leaving)
21:52:20	*	endragor joined #nim
21:52:54	Zevv[m]	disruptek: https://github.com/disruptek/cps/blob/master/stash/trycatch.nim
21:53:12	Zevv[m]	tell me where I should store 'err', 'msg' and 'where'.
21:56:58	*	endragor quit (Ping timeout: 246 seconds)
22:04:25	*	endragor joined #nim
22:12:13	*	endragor quit (Ping timeout: 264 seconds)
22:13:23	Zevv[m]	#20 is in a PR, I'm off to Zzz
22:13:36	disruptek	g'night old man.
22:15:49	FromGitter	<bung87> `result = prefix & result` araq how this can without copy ?
22:17:07	FromGitter	<bung87> and when assign a char still a copy ? i dont know how to optimize string manipulation
22:20:25	Araq	well you start by scanning the prefix
22:20:36	Araq	while you do that you can already append to result
22:20:49	Araq	and then there is no need for 'result = prefix & result'
22:21:16	Araq	the slice into 's' can be avoided with a simple prefixLen integer variable (you almost have it inside 'idx' but its scope is off)
22:21:26	Araq	and that's about it I think
22:24:40	FromGitter	<bung87> let me think of it
22:34:24	FromGitter	<bung87> then `result = newString(L)` no need ? I just do all .add char ?
22:37:48	Araq	you can use result = newStringOfCap(L) and .add chars
22:39:35	*	audiofile quit (Quit: Default Quit Message)
22:40:33	FromGitter	<bung87> L depends on partsLen, I cant declare it then add prefix char to it
22:41:56	FromGitter	<bung87> in the first loop
22:45:01	FromGitter	<bung87> I've modified follow your suggestion, except the end line
22:45:57	Araq	L is only an estimate
22:46:08	Araq	it doesn't have to be exact
22:50:00	Araq	oh the cursor bug looks easy to do tomorrow
22:50:27	Araq	but it'll make the optimization less effective
22:55:03	FromGitter	<bung87> ok I got it
22:57:30	FromGitter	<bung87> i have to change to countup easier for understanding when use .add
23:01:41	Araq	bung87: btw, exceptional work with your PRs, please continue
23:02:12	Araq	good night
23:03:22	FromGitter	<bung87> Thanks ! I done the work, only result string now.
23:04:21	FromGitter	<bung87> always get benefit from code reviews
23:06:33	*	krux02_ joined #nim
23:09:23	*	krux02 quit (Ping timeout: 260 seconds)
23:09:47	*	maier joined #nim
23:15:13	*	maier quit (Ping timeout: 264 seconds)
23:15:59	*	aurelius joined #nim
23:20:58	*	aurelius quit (Ping timeout: 246 seconds)
23:34:17	*	endragor joined #nim
23:36:42	Yardanico	ok so there's still some async leak left, but it's much smaller
23:36:59	Yardanico	from a nim compiler 1 month ago, after stress-testing asynchttpserver with wrk, the RAM usage was 1.5gb
23:37:03	Yardanico	now it's 100mb
23:37:12	Yardanico	that's for ~400-500k requests
23:41:37	*	vicfred quit (Quit: Leaving)
23:48:09	Yardanico	seems like it leaks in processBasicCallbacks
23:48:45	*	oddp quit (Ping timeout: 240 seconds)
23:50:30	Yardanico	oh it has some shallowCopy in there
23:57:17	Yardanico	for ~80k requests in debug mode "(allocCount: 6908857, deallocCount: 6558581)"
23:58:54	FromDiscord	<Clyybber> niice
23:59:05	Yardanico	yeah it's still much less than before
23:59:12	Yardanico	~100mb leak for asynchttpserver for ~400k requests
23:59:13	disbot	no footnotes for `100mb`. 🙁
23:59:22	disruptek	might create some real competition here.