<< 27-11-2018 >>
00:28:00FromDiscord_<treeform> dom96, thanks, I will try it out.
00:45:23*abm quit (Ping timeout: 268 seconds)
01:03:20*krux02 quit (Remote host closed the connection)
01:15:35*kapil____ joined #nim
01:55:45*vaios joined #nim
01:56:05*vaios left #nim (#nim)
02:05:59*ng0 quit (Quit: Alexa, when is the end of world?)
02:09:44*vlad1777d quit (Ping timeout: 250 seconds)
02:10:28*zachk quit (Quit: Leaving)
03:06:25*PMunch quit (Remote host closed the connection)
03:07:40*banc quit (Quit: ZNC - http://znc.in)
03:16:24*nif quit (Ping timeout: 252 seconds)
03:17:01*nif joined #nim
03:17:05*flyx quit (Ping timeout: 268 seconds)
03:18:36*sknebel quit (Ping timeout: 252 seconds)
03:20:10*flyx joined #nim
03:23:28*banc joined #nim
03:30:46*sknebel joined #nim
03:33:51*dddddd quit (Remote host closed the connection)
03:36:15FromGitter<Varriount> @mratsim Maybe the implied calling convention of a procedure?
03:38:02FromGitter<Varriount> @mratsim https://github.com/nim-lang/Nim/blob/eb919c35b29afee8fb26a2145d88f4db9e6c332b/compiler/ast.nim#L109
04:22:29*demond joined #nim
04:22:38*demond left #nim (#nim)
04:31:03*Jesin joined #nim
04:32:53*Jesin quit (Remote host closed the connection)
04:54:24FromGitter<AchalaSB> how to solve the issues. here is my code ⏎ ⏎ ```code paste, see link``` [https://gitter.im/nim-lang/Nim?at=5bfcce0097a8982b9a6ebcc6]
04:59:39leorizewell, the issue is literally written in the error message...
04:59:48leorizehave you `import os`?
05:00:57leorizealso, you might want to fix your indentation
05:06:43*Jesin joined #nim
05:06:59FromGitter<AchalaSB> Yes. There is no error for indentation
05:08:00leorizeofc, but context wise, you should indent the for loop
05:08:06FromGitter<AchalaSB> The keyword `fileDir` and `path` etc are user defined for that Im getting error
05:08:09leorizeit's currently outside of that proc with then indentation
05:08:15leorizethat*
05:11:04leorizeAchalaSB: the `for` loop and that `if` is outside of the proc
05:11:18leorizeyou have to indent it for it to be part of the proc
05:11:22leorizeso it could access `path`
05:11:44leorizeas of the functions, `import os` should be suffice
05:12:45FromGitter<AchalaSB> ```code paste, see link``` [https://gitter.im/nim-lang/Nim?at=5bfcd24df59704348e0dd1b4]
05:15:15FromGitter<AchalaSB> Now Im getting indentation errors. and the previous errors are not solved. ⏎ another error ⏎ ⏎ ```code paste, see link``` [https://gitter.im/nim-lang/Nim?at=5bfcd2e3e25cc274054ce551]
05:19:18leorizeplease, indent them correctly
05:19:28leorizeit should solve your problems
05:25:55*nsf joined #nim
06:18:18*nif quit (Ping timeout: 268 seconds)
06:19:30*nif joined #nim
06:29:24*nif quit (Ping timeout: 268 seconds)
06:29:59*nif joined #nim
07:27:58FromGitter<narimiran> @AchalaSB you can set up in your code editor that pressing Tab will indent your code X amount of spaces (for nim, X = 2), so you shouldn't have indentation problems anymore
07:40:26*jjido quit (Remote host closed the connection)
07:44:13*vqrs quit (Ping timeout: 245 seconds)
07:52:59*vqrs joined #nim
08:06:00FromGitter<gogolxdong> How to generate 4..18 size random bytes?
08:12:00FromGitter<gogolxdong> ```code paste, see link``` [https://gitter.im/nim-lang/Nim?at=5bfcfc507d26db1e8301d0c9]
08:12:32FromGitter<gogolxdong> Is this a good method? How about Windows?
08:15:20Araqimport random
08:15:30Araqvar s: seq[byte] = @[]
08:15:55Araqfor i in 4..18: s.add byte(random([0..255]))
08:17:52FromGitter<gogolxdong> ah, yes, what I was thinking!
08:23:12*stefanos82 joined #nim
08:32:15*zakora joined #nim
08:32:56*krux02 joined #nim
08:35:22FromGitter<gogolxdong> ```code paste, see link``` [https://gitter.im/nim-lang/Nim?at=5bfd01caa115c91ef79d3c21]
08:41:46FromGitter<gogolxdong> Should randmize be used only once globally?
08:42:52leorizeit should be used anytime you need to reseed the generator
08:43:21leorizebut for most usecase related to it, yes, executing it once should be suffice
08:43:43FromGitter<gogolxdong> ok
08:55:26*nsf quit (Quit: WeeChat 2.3)
08:56:21*nsf joined #nim
09:18:14*leorize quit (Quit: WeeChat 2.2)
09:36:10*abm joined #nim
09:42:52FromGitter<mratsim> this reminds me of the threads with the OP getting unbelievable head or tails result with Nim random
09:43:23FromGitter<mratsim> it was because he was reseeding the PRNG just before rand, everytime.
09:44:17*dom96_w joined #nim
09:57:27FromGitter<alehander42> this is always confusing with random api-s :(
10:03:57Araqdunno how so, if it were (randomize(); random()) then why would randomize() even exist
10:04:36*ng0 joined #nim
10:04:53Araqthe only logical conclusion is that randomize() usually needs to be called in a different scope
10:05:00Araqlike "once at program startup"
10:06:37*dddddd joined #nim
10:06:39FromGitter<alehander42> yeah, but then people forget to call randomize at all
10:07:07FromGitter<alehander42> there must be something that forces them to either `randomize()` or `dontRandomize()` initially :D
10:36:31*floppydh joined #nim
11:00:07*Vladar joined #nim
11:01:00FromGitter<narimiran> that something is "we don't have to do it in python"
11:04:23*vlad1777d joined #nim
11:06:43Araqtrue but real systems programming languages (TM) require it explicitly
11:10:34FromGitter<narimiran> i have no problem with that, but i think "python way" is the reason for the behaviour @alehander42 mentions
11:27:02serialdev[m]Do you guys have any areas in which, there is a lack of good libraries / none in the Nim ecosystem atm? just wondering where you are seeing lacks to for example fully adopt Nim, also to guide what to work on on the side as I master the language
11:28:05FromGitter<bung87> there’s a repo for this https://github.com/nim-lang/needed-libraries
11:28:52FromGitter<narimiran> serialdev[m] i would like to see "one lib to rule them all" when it comes to plotting (think of it like a sane matplotlib with normal API)
11:29:41Araqserialdev[m], many are asking for good Qt wrappers
11:29:49FromGitter<narimiran> i know we have couple of small plotting libraries in nim, but none of them convinced me to use them. so i still switch back to python and MPL when i need to plot something
11:30:37serialdev[m]That is really nice! thanks guys, would interop with arraymancer be a must :P?
11:30:42FromGitter<alehander42> I think I've heard also requests for a better http lib, e.g. similar to python requests
11:31:09FromGitter<narimiran> nothing is a must ;)
11:31:10Araqplotting is something I think we should do without a wrapper
11:31:26Araqbase it on putPixel and SDL2
11:31:40serialdev[m]yeah python requests is really well made, I agree plotting could be a killer feature too
11:31:40serialdev[m]cross compiling plots can be so useful
11:32:08FromGitter<narimiran> scikit-learn clone would also be nice, but i doubt we'll see it (feel free to prove me wrong ;))
11:32:36serialdev[m]putPixel and SDL2 sounds good!, I will eventually go for the machine learning stuff (its my day job) but atm wanna get really good at nim first
11:33:54FromGitter<bung87> something like tensorflow also is a good target.
11:34:10Araqit's what I did with Python back then fwiw
11:34:40AraqI threw away my matplotlib code and wrote my own charts renderer with putPixel
11:34:45FromGitter<narimiran> @bung87 i think @mratsim has some big plans with Arraymancer :)
11:34:57serialdev[m]yeah I'm awaiting the arraymancer hdf5 pull request atm to see if we can use Nim to target embedded systems
11:35:11serialdev[m]and just convert keras-2-nim
11:35:17FromGitter<bung87> @narimiran that’s cool!
11:36:18FromGitter<narimiran> regarding plotting: as i said, there are lots of small plotting libraries, people scratched their own itch i guess. but if we want "one big plotting lib", it should be a group effort, not a single man project
11:36:50serialdev[m]should we start one then? and settle on SDL2 + putPixel?
11:37:34FromGitter<narimiran> i would like to hear more opinions before settling on anything ;)
11:38:20serialdev[m]which one would be the best way to target both C and Wasm/js
11:38:34FromGitter<mratsim> nim-plotly
11:38:38serialdev[m]then worry about other targets I think cause we basically cover ground with that
11:38:58FromGitter<mratsim> I’ve also started a quick proof of concept using Vega here: https://github.com/numforge/monocle
11:39:27FromGitter<mratsim> And link to nim-plotly: https://github.com/brentp/nim-plotly
11:39:43FromGitter<mratsim> the hdf5 PR was merged normally
11:39:53FromGitter<narimiran> btw, i've recently discovered Plotnine — it is based on the famous ggplot2. i haven't used it yet, but it looks good: https://plotnine.readthedocs.io/en/stable/
11:41:49serialdev[m]these ones are nice, but should we target some sort of "native" nim libs aswell? since depending on another can become a hassle
11:42:30FromGitter<mratsim> In monocle I’m embedding Vega.
11:42:53FromGitter<mratsim> see here: https://github.com/numforge/monocle/tree/master/src/monocle/vega
11:43:49FromGitter<xmonader> is there a way to access variables in `try` block in the `execpt` block instead of declaring them before the try block?
11:43:53FromGitter<mratsim> Idem for display, I’m using the native webview on all platforms (and no it’s not Electron)
11:44:25FromGitter<mratsim> I’m using this to bind the webview: https://github.com/zserge/webview
11:44:51FromGitter<mratsim> so it’s super tiny, it’s just a 2k lines header.
11:46:22FromGitter<mratsim> but I didn’t work on it in the past 2 months because I don’t like JS and frontend and I got busy with low-level optimizations :D
11:47:44FromGitter<narimiran> classic Mamy :)
11:48:24FromGitter<narimiran> btw @mratsim, I would subscribe to your posts/thoughts/snippets, if you offered such a thing!
11:48:43FromGitter<mratsim> me too :P
11:48:53FromGitter<narimiran> lol
11:49:40FromGitter<mratsim> Here is one that can be a blog post, I think caching JIT code fragment is a similar problem to dynamic vs static framework in machine learning: https://github.com/mratsim/jitterland/blob/master/jit/jit_datatypes.nim#L90
11:52:23serialdev[m]loop fusion could be a really interesting one too!
11:53:32FromGitter<mratsim> done here: https://github.com/numforge/laser/blob/master/benchmarks/loop_iteration/iter_bench_prod.nim#L87-L90
11:54:10FromGitter<alehander42> while implementing `[]=` for distinct seq, I can't typecheck ` castseq[T (b)[c.int] = e`
11:54:26FromGitter<alehander42> what is the correct way to do that
11:54:34FromGitter<mratsim> There is also an advanced example with loopfusion for parallel reduction here: https://github.com/numforge/laser/blob/master/examples/ex05_tensor_parallel_reduction.nim
11:55:30serialdev[m]Thanks! that will be fun to work through
11:55:32FromGitter<mratsim> but yeah lots of stuff to document. btw Laser will be the future backend of Arraymancer
11:55:55FromGitter<narimiran> @mratsim make that one a proper blog post, please
11:56:11FromGitter<alehander42> yep, we need more blog posts
11:56:14FromGitter<mratsim> As a highlight, I reach OpenBLAS speed with pure Nim: https://github.com/numforge/laser/blob/master/benchmarks/gemm/gemm_bench_float64.nim#L201-L255
11:56:19FromGitter<narimiran> (jitterland)
11:56:48FromGitter<mratsim> so in the future I will fuse neural net operations like relu/tanh/sigmoid into the matrix multiplication
11:57:14FromGitter<mratsim> to gain 1.5 to 2x speed by avoiding to loop again for NN activations
11:57:38serialdev[m]That will be super nice!, specially in constrained systems
11:57:50FromGitter<mratsim> and I will also fuse the im2col reshaping for convolution into the matrix multiplication call
11:58:12serialdev[m]so computer vision killer features for nim there
11:58:41FromGitter<mratsim> another good read by the way, fast convolutions research: https://github.com/numforge/laser/blob/master/research/convolution_optimisation_resources.md
11:59:56serialdev[m]How do you have the time! also this is amazing work :)
12:00:43FromGitter<mratsim> Good read as well, with Intel engineers participation: fast recurrent neural net: https://github.com/mratsim/Arraymancer/issues/228
12:01:45*rockcavera quit (Read error: Connection reset by peer)
12:04:11serialdev[m]optimizing rnns would be great , since most people are actively avoiding them atm due to the high cost
12:09:31FromGitter<mratsim> It can be fast but, it can’t compare to convolution speed I think. I’ll see when I revamp the backend.
12:09:45*rockcavera joined #nim
12:14:23serialdev[m]Thanks for the links btw :)
12:17:42FromGitter<narimiran> btw, for those of you who weren't here last night — great opportunity to practice your nim skills and have loads of fun: Advent of Code https://nim-lang.org/blog/2018/11/26/advent-of-nim.html
12:23:37FromGitter<mratsim> the fastest = someone who wakes up at 5am UTC :P
12:27:07FromGitter<narimiran> yup, been there done that. can recommend, will do it again :D
12:28:58FromGitter<narimiran> it would be really nice if some of us would make it to global leaderboard (it is highly competitive and people are very fast). it would be a prove that in nim, although it is compiled, you can be as fast as people in scripting languages
12:29:52FromGitter<alehander42> i think it's really just a matter of popularity: if you have 100x people in lang X, it's probable that more of them would be up there
12:31:00FromGitter<narimiran> so, let's increase number of nim people participating ;)
12:38:30*sknebel quit (Ping timeout: 250 seconds)
12:46:51*dom96_w quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
12:47:42*dom96_w joined #nim
12:50:48*Snircle joined #nim
12:53:07serialdev[m]mratsim you might find this library interesting, I was evaluating it before settling on arraymancer https://github.com/marcoheisig/Petalisp (it is in lisp) kinda covers the same ground and might have some interesting attributes.
13:01:46*sknebel joined #nim
13:39:21*PMunch joined #nim
13:43:06FromGitter<mratsim> you need to tag gitter people with @ before their name ;) and thanks
13:46:20FromGitter<mratsim> Yeah it’s exactly the same scope as Laser. Being a building block for high performance computing high level library similar to Numpy
13:46:57FromGitter<mratsim> I can’t really read lisp though :P
14:00:50*sknebel quit (Ping timeout: 250 seconds)
14:02:39*sknebel joined #nim
14:05:06*leorize joined #nim
14:26:39*vlad1777d quit (Ping timeout: 246 seconds)
14:44:15*leorize quit (Quit: WeeChat 2.2)
14:44:25*leorize joined #nim
14:44:37*Vladar quit (Remote host closed the connection)
14:57:11shashlickis it possible to have multiple of same argument with cligen - e.g. -I with gcc can be done multiple times
14:58:16Araqshashlick, read my remark on github?
15:01:48shashlickyes Araq - was going to respond but good to discuss here
15:02:13shashlickso first, nimgen runs most wrapper headers thru gcc -E to resolve all preprocessor declarations
15:02:49shashlickso running on the user's system is useful since all platform specific stuff gets done that way
15:03:14shashlicksecond, most nimgen wrappers compile in the source rather than using dynlib so i need more than just the headers
15:03:34shashlickso what nimgen does is clones the repo (sparse checkout usually of essential files) as the first step
15:04:00shashlickyou can now pin to a commit/tag so that can be fixed and updated when needed so I don't need to check in headers into the wrapper repo
15:05:21PMunchAraq, I have two issues with nimlsp. First, how should the user specify where the "nimsuggest.nim" file is? I guess this could be done with a -d:nimsuggestpath switch but does that work with Nimble? Second is that nimsuggest needs to find the Nim source files, which fails for nimlsp but works for nimsuggest (because of how findExe works, I think).
15:08:25Araqshashlick, well that's why c2nim translates #define, after the preprocessing step, you have already lost
15:09:08AraqPMunch, why does it matter where nimsuggest.nim is? nimlsp.exe should be next to nim.exe
15:10:19shashlickAraq: what's the value in us preprocessing which is mostly used for platform specific stuff when the compiler already does all that for us
15:10:31shashlickmakes the code specific to the user's system
15:10:39Araqbut that's wrong, it's usually not platform specific
15:10:52PMunchBecause with the PR I created nimlsp can now import nimsuggest.nim instead of calling it as a program on run-time
15:10:52Araq*after* the preprocessor it's platform specific
15:10:59PMunchSo it needs nimsuggest.nim to import it
15:11:32shashlickbut then you want c2nim or equivalent to go through all defines and only convert procs/structs that are relevant to the platform
15:11:42shashlickwhy do that when the preprocessor already does that for you
15:12:41Araqno, you want c2nim to translate an #ifdef to a 'when defined'
15:13:16shashlickjust try "gcc -E -dM -xc /dev/null" - there's over 350 default defines on my ubuntu system
15:13:26Araqthe preprocessor throws away precious information
15:13:56Araqand stuff in /usr/include is not representative for C libs that are supposed to be wrapped for other PLs
15:16:43leorizePMunch: you could use git submodule to fetch Nim's source
15:16:52PMunchHmm, I guess that's true..
15:17:03leorizeor you could just create a nimble package out of nimsuggest
15:17:10PMunchThat's probably the "rightest" way to compile it
15:17:32shashlickAraq: ok lets use an example - https://github.com/vurtun/nuklear/blob/master/src/nuklear.h
15:17:49shashlickhow much of that ifdef stuff is relevant to the nim wrapper that needs to be generated?
15:17:52PMunchWell it already is a nimble package
15:18:04PMunchBut it's old and only meant for issue tracking
15:18:11PMunchOld as in 0.15.2 old
15:18:22leorizewell, maybe you could now patch it up?
15:18:35leorizenimble package can be within a subfolder, no?
15:19:42shashlickAraq: your idea is interesting but that means we need to convert all this into Nim when most of it is irrelevant
15:20:24shashlickand the only benefit is to have a generic nim file that is done up front rather than generating a specific nim file on the user's machine
15:20:36leorizeisn't that the point of c2nim?
15:22:43shashlickleorize: related to https://github.com/nim-lang/Nim/issues/8327 - more about how to architect and maintain wrappers
15:23:46Araqshashlick, "the only benefit"? sorry, but that's how every wrapper was created without nimgen
15:23:59Araqand I argued before that it is a tremendous benefit
15:24:39Araqbecause then you have Nim code that is only subject to Nim compiler versioning problems
15:25:41*Araq is looking at nuklear.h
15:27:49shashlickhttp://nimgen.genotrance.com/nimnuklear/nuklear.nim.html is what nimgen+c2nim generate after running on Windows
15:27:59shashlickafter running it thru gcc -E
15:28:46shashlickthat is a snapshot which a user gets when he runs it on his machine - i don't have to maintain this output, i maintain the process
15:29:24shashlickand the tools are built in nim, no depending on anything so it is not like yak and cmake and such
15:29:25Araq"after running it thru gcc -E" <-- that is AGAINST c2nim's design
15:30:06Araqwhat you should do instead is to wrap the stuff you don't need in an #ifndef C2NIM #endif section, for example
15:31:34shashlickif you run c2nim on nuklear.h directly, it dies on line 37 - it does not handle preprocessor stuff today
15:31:43*Trustable joined #nim
15:34:54shashlicklet me ask this differently - and maybe I missed something you said earlier - while I agree it is more valuable to generate a generic Nim wrapper, it is tedious since neither c2nim nor nimterop understand the preprocessor and the crazy ways people have used it
15:35:13shashlicki already have to deal with the actual C/C++ code and ensure it is wrapped correctly and works with Nim
15:35:52shashlickwhat is the value in investing a lot of time on the preprocessor rather than leveraging existing tools and bringing more libraries into the Nim ecosystem
15:36:33shashlickeach user can still checkin the resulting wrapper
15:41:44shashlickjust like you would expect the user to download a dll or support files or whatever and maintain that as part of his build environment
15:43:26shashlickagain, i'm talking as a package maintainer, not a user of the package. I have my pragmatic hat on - the one that thinks having more libraries makes Nim easier to build with
15:44:01shashlickbut if this approach is unusable as a user then I'd like to know why
15:49:33Araqc2nim understands the preprocessor
15:49:54Araqbut you have to translate the #define into a #def
15:50:20Araqa #define for c2nim is always something it tries to convert into either a const or a template
15:50:34Araqa #def is a directive for c2nim's builtin preprocessor
15:50:50*narimiran joined #nim
15:54:17PMunchleorize, yeah I guess there could exist a nimsuggest.nimble alongside compiler.nimble in the root of the repo and that would install the nimsuggest folder..
15:54:52shashlickwell a file like nuklear.h has 120 lines with a #define in it, some of which should be changed to #def and others left alone
15:56:12shashlickthat works if i love nukear.h enough to sit and hand examine and edit it and then check in my copy and leave it like that until i'm motivated enough to do it again
15:56:34shashlickits hard to automate and keep compatible with nuklear as it evolves
15:58:37shashlickI understand that's how it has been with existing Nim wrappers that have been created since the user was doing it and found it worth it for their application
15:59:27shashlickit does not scale when I want to bring 200 wrappers into the ecosystem
15:59:42PMunchleorize, oops maybe not: Error: Only one .nimble file should be present in /home/peter/Projects/Nim
16:00:35shashlickbut again, my motivation is to build wrappers quickly and maintain indefinitely - if they are useless then I will obviously reassess 🙂
16:02:37shashlickfor example, I wrapped all of tree-sitter in a couple days and built nimterop on top of that - you should be able to nimble install https://.../nimterop and it will just work - i think there's value in that but i'm very open to adjusting if there's deficiencies
16:07:57Araqhttps://github.com/curl/curl/blob/master/include/curl/curl.h
16:08:04Araq#define CURL_HTTPPOST_PTRBUFFER (1<<5)
16:08:12Araq^ do you get this const with your approach?
16:09:32shashlickright now, I only look for "#define XXXY int|float|hex" but it is a poor method since it does not handle ifdefs
16:09:50shashlickbut @timotheecour found a way to get #defines in the preprocessor output
16:10:10shashlickusing "gcc -E -dD" you get hte preprocessed output as well as resulting #defines
16:10:21shashlicki'm investigating that to get better results than what nimgen does today
16:10:25Araqas I said, you don't have to re-do it all over again if a new header file version arrives
16:11:13Araq"gcc -E -dD" is just as wrong, it assumes e.g. LINUX is defined and expands the macros correspondingly
16:11:52shashlickit seems to work fine on Windows mingw so far but i'll look into it
16:12:47Araqwell I'm not saying your approach doesn't work, I'm saying it will bitrot much more quickly than the old approach
16:14:26Araqand if you lack a good solution for the #define problem, I don't see the benefits over c2nim
16:15:00shashlickso i have had 25 wrappers being tested daily and only one wrapper failed in the last year due to some file rearchitecting of fastText - was a 5 minute fix
16:15:22shashlickit is an ongoing effort so I hope to improve it
16:15:31shashlickwill approach -dD with reservations
16:16:25Araqyeah but that was by fixed by you
16:16:53Araqwould I have been able to fix a failing 'nimble install foo'?
16:17:05Araq(assuming nimble install does the nimgen dance)
16:18:38Araqand how do I install the win32 version of the foo wrapper when I'm on Linux?
16:19:43Araqgenerated platform specific wrappers are not good. In ordinary Nim code I can do 'nim check --os:linux --cpu:i386 stuff.nim' on a Windows machine
16:19:45shashlickyou are far smarter than me - I'll bet you could fix it!
16:20:02shashlickhttps://github.com/genotrance/nimfastText/commit/c1c7f6e61f2c233f0426ac120e32ba8ea37402cb is the commit change for nimfastText and half of it was cleanup - only new thing was meter.nim
16:20:10shashlickyes - nimble install does the nimgen dance
16:20:47shashlickyou cannot install the windows version on linux since I rely on the preprocessor so yes that use case isn't possible with this approach
16:21:12shashlickbut that's why I use Nim and not C/C++
16:22:01shashlickactually, if you install mingw on linux and use that preprocessor, it will generate Windows wrappers
16:22:10shashlicksince all defines will be windows specific
16:22:34Araqwell your changes to nimfastText.cfg are black magic to me
16:22:46Araqand yes, I'm sure I can figure it out. eventually.
16:23:03Araqbut here is what I usually do instead.
16:23:07Araq'brew install foobar'
16:23:16Araq--> shitty error message
16:23:37*Araq curses every package manager and uses something else.
16:24:13shashlickso i hope you try one of the nimgen wrappers and see how it works for you
16:24:20Araqok
16:24:24shashlickand perhaps use nimgen to wrap something in the future and see what you think
16:24:37Araqgive me 5 "nimble install" commands please
16:26:11shashlickhttps://docs.google.com/spreadsheets/d/1sM1TE6Cp8owVU1NxYyKd_0rhJCS2YbHorzhDVAoxxk4/edit#gid=0 - that's a list of all nimgen wrappers so far
16:26:46shashlickanything of interest to you, if it is in the nimble directory, you can `nimble install X` and it should work
16:26:57Araqnimble install nimfastText
16:27:03Araq Error: Package not found.
16:27:16shashlickits not in the nimble directory yet
16:27:32shashlicknimble install https://github.com/genotrance/nimfastText
16:27:39shashlickfingers crossed 😄
16:27:53shashlickyou had to pick the most painful one didn't you!
16:28:07shashlickhttps://github.com/genotrance/nimfastText/blob/master/tests/tnimfastText.nim is a test example
16:28:20Araqhttps://gist.github.com/Araq/876d2c5255eff71fe3dae53b3e323d19
16:30:06*ng0 quit (Quit: Alexa, when is the end of world?)
16:30:08AraqI'm adding comments to this gist
16:30:51shashlickweird - that's not my package - it is something else
16:31:38AraqSuccess: nimgen installed successfully.
16:31:39AraqDer Befehl "nimgen" ist entweder falsch geschrieben oder
16:31:39Araqkonnte nicht gefunden werden.
16:31:50Araqwhich means "could not find nimgen"
16:32:08shashlickis ~/.nimble/bin in your path?
16:33:48shashlicklooks like I never added nimfuzz to the nimble directory - probably because there was a name conflict and i forgot
16:33:57Araqit wasn't, so ok, error on my part.
16:34:08*Vladar joined #nim
16:34:43Araqnimfuzz remains broken though
16:35:15shashlickyes - like I said, nimfuzz is another package, not the one I am maintaining - will have to post on nimble with another name
16:35:23shashlicki've added it to my tracker
16:35:48Araqnimble install nim7z
16:35:51Araqworks.
16:36:14Araqso ... congrats.
16:36:19shashlickwhew!
16:36:21Araqfor now. :P
16:36:52Araqbut it's a complex setup so the chances are higher it will break
16:37:11shashlicki agree - that's why i have both appveyor and travis daily builds running
16:37:23shashlick0.18.0, 0.19.0 and devel across win/lin/osx
16:37:26Araqwhereas with the old appraoch you at least have some outdated Nim code you can fix
16:38:16Araqnevertheless it seems to work well enough that I'll shut up now.
16:38:28shashlickbut see nuklear-nim - it was built 2 years ago and no updates since
16:39:04shashlickand nuklear has evolved quite a bit since then
16:40:41shashlickanyway, is it safe to say that your primary concern is that the nimgen approach is more fragile than what you have had so far?
16:44:42shashlickI attempt to alleviate that with a wide CI test matrix
16:45:11shashlickother feedback I got was - #defines are still primitive/kludgy, needs re-eval
16:45:25leorizeis there any approach that doesn't end up with a CI matrix? :P
16:45:57shashlickreproducible builds - we don't have that yet, need to highlight that end user should save the generated wrapper
16:46:34shashlickcross compilation - have a solution for Windows on Linux, but not the other way around
16:46:42shashlickanything else, please let me know
16:47:14leorizereproducible builds is unlikely with nimgen I think
16:48:16shashlickwell - nimgen now allows locking upstream to a particular commit/tag so you can fix that
16:48:38leorizeif we could build upon libclang, it can be a decent C -> Nim compiler with all the customization that one could hope for
16:48:45shashlicki haven't done that for any wrappers so far though
16:48:50shashlicknext is to checkout a particular wrapper commit which is possible with nimble already
16:49:12shashlickbeyond that, user needs to fix their OS, C and Nim compilers
16:49:48shashlickleorize: I looked into libclang - it is a huge dependency so i've been building on tree-sitter instead
16:50:06shashlickif I could assume clang is always available, it would be altogether different
16:50:07leorizeyea, but libclang is rather reliable
16:50:17leorizeand capable of parsing any kind of C code you throw at it
16:51:18shashlickno doubt - if Nim standardizes on clang across all OS, would work for me
16:51:47leorizelibclang and friends can be installed quite easily on every os
16:52:41leorizethe bad thing about libclang is that the ABI breaks ever so frequent
16:53:22shashlickyes but compare a 250MB clang download vs. a 4mb binary that can be built and installed seamlessly via nimble
16:55:30shashlickworks better for the nimgen model anyway but as you can see, I'm quite committed to that idea 😄
16:56:01shashlickanyway, nimterop is still a POC but does look promising
16:56:47shashlickreality is when end users actually try it out its output
17:04:06*narimiran quit (Remote host closed the connection)
17:11:50*nsf quit (Quit: WeeChat 2.3)
17:20:44Araq> anyway, is it safe to say that your primary concern is that the nimgen approach is more fragile than what you have had so far?
17:21:05Araqmore fragile, yes. also... "more stuff to learn". for me at least.
17:22:04Araqplus I think Nim programmers enjoy reading Nim code moreso than a complex setup that produces Nim code.
17:22:28Araqwe did the same with Nimscript vs nim.cfg. "Nim all the things!"
17:24:06*dom96_w quit (Changing host)
17:24:06*dom96_w joined #nim
17:25:35dom96_wI'm guessing this is possibly what you guys are discussing, but now that I'm seeing nimgen discussed here I wonder: why not just turn nimgen into a CLI app?
17:25:55leorizethen wouldn't it become c2nim?
17:26:14dom96_wIt would be a wrapper on top of c2nim
17:26:23shashlicknimgen already has a CLI interface
17:26:29shashlickit isn't as capable as the cfg file though
17:26:40dom96_wI do agree with Araq's sentiment that the wrapper Nim code should be committed
17:27:25dom96_was cool as it is that nimgen runs during a Nimble package's install process it makes me really nervous :)
17:27:44dom96_wAnd I don't see much upside over just running `nimgen` before committing.
17:28:22Araqdom96_w, you can't run nimgen before committing because you would need to run it for ever OS/CPU combination
17:28:29leorizenimgen currently generates tailored wrapper
17:28:42shashlickAraq: you don't need to look at the cfg file as a user - just the generated nim code which is documented
17:28:58dom96_wAraq: Why? Why doesn't nimgen generate `when` statements for that?
17:29:02shashlicke.g. http://nimgen.genotrance.com/nimnuklear/
17:29:12Araqdom96_w, it doesn't, that's what I've been complaining about
17:29:39shashlickdom96_w: it runs the preprocessor on the user's system during nimble install prior to generating the wrapper
17:29:49shashlickso you get user specific wrappers at install time
17:31:06dom96_wI think you should consider implementing support for generating a single wrapper for all platforms
17:31:42leorizethen it would be c2nim lol
17:32:32shashlickwhat is the benefit really? it is a huge effort to ensure that the preprocessor is correctly interpreted without the preprocessor
17:33:17leorizeif you play around with gcc specs, you can configure default definitions however you want
17:33:26shashlicknote that nimgen only automates the wrapper process - it still leverages c2nim thru and thru
17:34:01shashlickif you want c2nim to process the #defines correctly, you need to make a bunch of edits - change #defines to #def, use #ifndef C2NIM and stuff
17:34:24shashlickthe preprocessor already does all that for you, why recreate it when what you really want is reproducible builds
17:50:23*leorize quit (Quit: WeeChat 2.2)
17:54:14Araqbecause #define is also used to declare constants and macros that belong to the very API you're trying to wrap.
17:54:40Araqand c2nim has lots of options to mitigate this pain anyway. or more stuff could be added for it
17:56:19shashlicklike I mentioned, I do pull those out in nimgen today and am exploring using gcc -dD
17:57:13Araqwell I answered your question
17:58:27shashlickso net - if we can get reliable installs, relevant #defines converted to consts, along with reproducible builds, will we meet the basic requirements?
18:16:14Araqyes, I think that covers all concerns
18:16:38PMunchdom96_w, did you use a library when you created your snake game?
18:16:52dom96_wFor which part?
18:17:21shashlickAraq: okay thanks - will aspire for that goal. i think I have reliable installs mostly. other two are WIP
18:17:27PMunchFor the rendering and game loop stuff
18:17:36dom96_wI wrote my own
18:17:41dom96_wit's called gamelight
18:17:56Araqshashlick, yeah, I agree. No rush.
18:18:32Araqand I'll continue to add features to c2nim :P I like its code base. ;-)
18:18:51PMunchAh the classic optimism of a new project "[...] But hopefully with time it will be polished." - 2 years ago :P
18:19:03AraqI always thought we can do away with the #def vs #define distinction by some clever heuristic
18:19:09shashlickok thanks for all the feedback
18:19:11dom96_wPMunch: Did you answer my question in #nim-offtopic? (Can't check myself because I don't have access to my BNC)
18:19:25PMunchAbout where my talk was?
18:19:35*literal joined #nim
18:19:36PMunchI didn't quite understand what you asked :P
18:19:37shashlicki think -dD is working well in my limited tests
18:19:56shashlickAraq: what i do need is help with fixing all the c2nim issues I've opened over the years
18:20:00*theelous3_ joined #nim
18:20:03dom96_wPMunch: someone told me you're giving a lightning talk at FOSDEM
18:20:11dom96_w(About Nim)
18:20:36PMunchWell I've submitted it, but they aren't announcing the accepted talks until December 15th
18:20:49dom96_wAhh, which devroom?
18:21:00PMunchLighting talks don't need a dev room
18:21:08PMunchThey are held in the main auditorium
18:21:10dom96_woh.
18:21:24PMunch"Lighting talks are short — 15 minute long — talks on a wide variety of topics. Anyone who has something interesting to say about an open source or free software topic can apply. We particularly encourage topics that do not fit in any of the developer rooms."
18:21:32FromGitter<arnetheduck> Araq, shashlick, the first of you that generates a faithful conversion of llvm-c's headers wins a beer from me :)
18:21:46dom96_wCool. Pretty sure I've seen some devrooms that have lightning talks too
18:22:00PMunchOh yeah probably
18:22:12PMunchThe devrooms are pretty free to make their own program
18:23:17PMunchIt would be cool to organise a "Small languages" or just "Language design" devroom, that could feature some longer talks on a variety of languages
18:23:39PMunchWent to some of the Ada talks last year it was really interesting
18:23:54*dom96_w quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
18:24:31shashlick@arnetheduck: challenge accepted 😄
18:26:32shashlickat least the enums seem ordered!
18:35:22*nsf joined #nim
18:47:35FromGitter<arnetheduck> wait until you get to the part where they use `#defines` to generate function signatures.. like the ugliest code of nim macro maximalists: public function prototypes generated using macros - totally useless.. oh, and some enums are flags actually, with some enum values being combinations of previous enums (as Araq has pointed out, semantically c enum == nim int, or maybe nim `uint` - in `c` an enum is an integer
18:47:36FromGitter... unspecified size that varies between compilers)
18:50:28*PMunch quit (Remote host closed the connection)
18:53:56*bitstoppa[m] quit (Remote host closed the connection)
18:53:56*xomachine[m] quit (Remote host closed the connection)
18:53:58*petersjt014[m] quit (Remote host closed the connection)
18:53:59*notbot[m] quit (Remote host closed the connection)
18:54:01*MrAxilus[m] quit (Write error: Connection reset by peer)
18:54:02*Demos[m] quit (Write error: Connection reset by peer)
18:54:05*stisa[m] quit (Remote host closed the connection)
18:54:05*serialdev[m] quit (Read error: Connection reset by peer)
18:54:05*zielmicha[m] quit (Read error: Connection reset by peer)
18:54:05*hohlerde quit (Read error: Connection reset by peer)
18:54:05*zacharycarter[m] quit (Read error: Connection reset by peer)
18:54:06*yglukhov[m] quit (Read error: Connection reset by peer)
18:54:08*syntonym[m] quit (Remote host closed the connection)
18:54:08*toofly[m] quit (Remote host closed the connection)
18:54:08*spymasterd[m] quit (Remote host closed the connection)
18:54:08*blarghz[m] quit (Remote host closed the connection)
18:54:08*Miguelngel[m] quit (Remote host closed the connection)
18:54:10*GitterIntegratio quit (Read error: Connection reset by peer)
18:54:11*gh0st[m] quit (Remote host closed the connection)
18:54:11*jaens[m] quit (Remote host closed the connection)
18:54:11*narimiran[m] quit (Remote host closed the connection)
18:54:11*TheManiac[m] quit (Remote host closed the connection)
18:54:12*sg-james[m] quit (Read error: Connection reset by peer)
18:54:12*k0mpjut0r quit (Read error: Connection reset by peer)
18:54:12*dyce[m] quit (Read error: Connection reset by peer)
18:54:12*Connor[m] quit (Write error: Connection reset by peer)
18:54:13*sendell[m] quit (Remote host closed the connection)
18:54:14*xylef quit (Write error: Connection reset by peer)
18:54:14*pqflx3[m] quit (Remote host closed the connection)
18:54:14*exothermic[m] quit (Remote host closed the connection)
18:54:14*endes[m] quit (Remote host closed the connection)
18:54:15*monofuel[m] quit (Remote host closed the connection)
18:54:15*macsek1911[m] quit (Remote host closed the connection)
18:54:17*ArchieT[m] quit (Remote host closed the connection)
18:54:17*planetis[m] quit (Read error: Connection reset by peer)
18:54:18*codevance[m] quit (Remote host closed the connection)
18:54:18*TheKing[m] quit (Remote host closed the connection)
18:55:23shashlickGee thanks for bursting my bubble!
18:56:08shashlickYes I have nothing yet for #define function signatures
18:56:54shashlickBut am yet to try the distinct int method for enums
19:02:59*floppydh quit (Quit: WeeChat 2.3)
19:11:51*dyce[m] joined #nim
19:14:23*Trustable quit (Remote host closed the connection)
19:22:58*Connor[m] joined #nim
19:22:58*codevance[m] joined #nim
19:22:58*Demos[m] joined #nim
19:22:58*TheKing[m] joined #nim
19:22:58*xylef joined #nim
19:22:59*ArchieT[m] joined #nim
19:22:59*GitterIntegratio joined #nim
19:22:59*sendell[m] joined #nim
19:22:59*gh0st[m] joined #nim
19:23:00*serialdev[m] joined #nim
19:23:00*notbot[m] joined #nim
19:23:00*k0mpjut0r joined #nim
19:23:05*Miguelngel[m] joined #nim
19:23:05*zacharycarter[m] joined #nim
19:23:05*spymasterd[m] joined #nim
19:23:05*TheManiac[m] joined #nim
19:23:05*macsek1911[m] joined #nim
19:23:05*exothermic[m] joined #nim
19:23:05*blarghz[m] joined #nim
19:23:06*bitstoppa[m] joined #nim
19:23:06*endes[m] joined #nim
19:23:06*narimiran[m] joined #nim
19:23:06*syntonym[m] joined #nim
19:23:06*xomachine[m] joined #nim
19:23:06*sg-james[m] joined #nim
19:23:06*pqflx3[m] joined #nim
19:23:06*toofly[m] joined #nim
19:23:06*yglukhov[m] joined #nim
19:23:06*petersjt014[m] joined #nim
19:23:06*zielmicha[m] joined #nim
19:23:07*stisa[m] joined #nim
19:23:07*hohlerde joined #nim
19:23:07*jaens[m] joined #nim
19:23:07*MrAxilus[m] joined #nim
19:23:07*planetis[m] joined #nim
19:23:07*monofuel[m] joined #nim
19:24:47*jjido joined #nim
19:25:56*narimiran joined #nim
19:37:34*kungtotte quit (Remote host closed the connection)
19:46:14*jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
19:56:32*syntonym[m] quit (Remote host closed the connection)
19:56:32*spymasterd[m] quit (Remote host closed the connection)
19:56:33*hohlerde quit (Remote host closed the connection)
19:56:33*bitstoppa[m] quit (Remote host closed the connection)
19:56:33*endes[m] quit (Remote host closed the connection)
19:56:33*stisa[m] quit (Remote host closed the connection)
19:56:33*notbot[m] quit (Remote host closed the connection)
19:56:33*ArchieT[m] quit (Read error: Connection reset by peer)
19:56:37*xomachine[m] quit (Remote host closed the connection)
19:56:37*macsek1911[m] quit (Remote host closed the connection)
19:56:37*exothermic[m] quit (Remote host closed the connection)
19:56:38*codevance[m] quit (Remote host closed the connection)
19:56:40*narimiran[m] quit (Remote host closed the connection)
19:56:40*Miguelngel[m] quit (Remote host closed the connection)
19:56:40*zacharycarter[m] quit (Remote host closed the connection)
19:56:42*dyce[m] quit (Remote host closed the connection)
19:56:44*jaens[m] quit (Remote host closed the connection)
19:56:46*Connor[m] quit (Write error: Connection reset by peer)
19:56:47*yglukhov[m] quit (Remote host closed the connection)
19:56:47*petersjt014[m] quit (Remote host closed the connection)
19:56:47*serialdev[m] quit (Remote host closed the connection)
19:56:48*Demos[m] quit (Read error: Connection reset by peer)
19:56:49*monofuel[m] quit (Read error: Connection reset by peer)
19:56:52*k0mpjut0r quit (Read error: Connection reset by peer)
19:56:52*gh0st[m] quit (Read error: Connection reset by peer)
19:56:52*sendell[m] quit (Remote host closed the connection)
19:56:52*sg-james[m] quit (Remote host closed the connection)
19:56:52*TheManiac[m] quit (Remote host closed the connection)
19:56:52*MrAxilus[m] quit (Remote host closed the connection)
19:56:52*pqflx3[m] quit (Remote host closed the connection)
19:56:53*xylef quit (Remote host closed the connection)
19:56:54*toofly[m] quit (Remote host closed the connection)
19:56:54*planetis[m] quit (Remote host closed the connection)
19:56:54*blarghz[m] quit (Remote host closed the connection)
19:56:54*zielmicha[m] quit (Remote host closed the connection)
19:56:55*GitterIntegratio quit (Read error: Connection reset by peer)
19:56:56*TheKing[m] quit (Read error: Connection reset by peer)
19:59:01*krux02 quit (Remote host closed the connection)
20:03:16*dyce[m] joined #nim
20:10:01*kungtotte joined #nim
20:13:11FromGitter<alehander42> @arnetheduck in principle, there is nothing wrong with using ast macros to generate public functions: you can easily generate sensible documentation too in many cases
20:14:21*Connor[m] joined #nim
20:14:21*TheKing[m] joined #nim
20:14:21*Demos[m] joined #nim
20:14:21*codevance[m] joined #nim
20:14:22*xylef joined #nim
20:14:22*ArchieT[m] joined #nim
20:14:23*GitterIntegratio joined #nim
20:14:23*gh0st[m] joined #nim
20:14:23*sendell[m] joined #nim
20:14:23*serialdev[m] joined #nim
20:14:23*notbot[m] joined #nim
20:14:23*k0mpjut0r joined #nim
20:14:28*endes[m] joined #nim
20:14:28*bitstoppa[m] joined #nim
20:14:28*zacharycarter[m] joined #nim
20:14:29*hohlerde joined #nim
20:14:29*TheManiac[m] joined #nim
20:14:29*xomachine[m] joined #nim
20:14:29*pqflx3[m] joined #nim
20:14:29*exothermic[m] joined #nim
20:14:29*narimiran[m] joined #nim
20:14:29*yglukhov[m] joined #nim
20:14:29*zielmicha[m] joined #nim
20:14:29*Miguelngel[m] joined #nim
20:14:29*spymasterd[m] joined #nim
20:14:29*macsek1911[m] joined #nim
20:14:30*petersjt014[m] joined #nim
20:14:30*blarghz[m] joined #nim
20:14:30*syntonym[m] joined #nim
20:14:30*planetis[m] joined #nim
20:14:30*jaens[m] joined #nim
20:14:30*toofly[m] joined #nim
20:14:30*sg-james[m] joined #nim
20:14:31*monofuel[m] joined #nim
20:14:31*MrAxilus[m] joined #nim
20:14:31*stisa[m] joined #nim
20:14:45*zachk joined #nim
20:15:41*zachk quit (Changing host)
20:15:41*zachk joined #nim
20:25:36FromGitter<yyyc514> Is this is a repository so I can submit fixes? `https://nim-lang.org/docs/backends.html` ?
20:26:47FromGitter<GULPF> That file is in the docs folder in the main repo
20:27:13FromGitter<yyyc514> found it
20:27:20FromGitter<yyyc514> what is rst hmmm
20:28:12FromGitter<GULPF> It's restructuredtext. It's like markdown, but harder to remember
20:31:15FromGitter<alehander42> it's easy to get used to
20:35:50FromGitter<arnetheduck> @alehander42 "in theory, theory and practice are the same, in practice..."
20:41:04FromGitter<alehander42> there is a right way and a wrong way to do most things
20:44:58*jjido joined #nim
20:46:06*Vladar quit (Remote host closed the connection)
20:48:20*jjido quit (Client Quit)
20:54:17*zakora quit (Quit: WeeChat 2.2)
20:55:09*jjido joined #nim
21:01:26FromGitter<alehander42> but yeah i get your worry :D
21:06:06*Snircle quit (Quit: Textual IRC Client: www.textualapp.com)
21:06:39*Snircle joined #nim
21:19:52*nsf quit (Quit: WeeChat 2.3)
21:25:15*Jesin quit (Quit: Leaving)
21:28:00*Jesin joined #nim
21:41:00*narimiran quit (Ping timeout: 252 seconds)
21:46:29*cozachk joined #nim
21:48:30*zachk quit (Ping timeout: 252 seconds)
21:50:16FromGitter<arnetheduck> say I have an array[42] and a seq with at least so many items.. what's the idiomatic way to set the first 40 elements of the array to the first 40 elements in the seq?
21:54:54FromGitter<kayabaNerve> ```for i in 0 ..< myArr.len: ⏎ myArr[i] = mySeq[i]``` ⏎ ⏎ ? [https://gitter.im/nim-lang/Nim?at=5bfdbd2e8dffd748dbf1c62b]
22:02:27FromGitter<zetashift> can't you just use toSeq?
22:03:57FromGitter<kayabaNerve> @zetashift Seq -> Array
22:04:12FromGitter<kayabaNerve> With seq.len > arr.len
22:04:33def-Where's the fun in doing it idiomatically? myArr = cast[ptr type(myArr)](addr mySeq[0])[]
22:04:34*kapil____ quit (Quit: Connection closed for inactivity)
22:06:31def-or just copyMem(addr myArr[0], addr mySeq[0], sizeof(myArr))
22:07:13FromGitter<arnetheduck> was hoping for something more beautiful. oh well.
22:29:59*cozachk quit (Quit: Leaving)
22:31:28Araqdon't use cast, don't use copyMem, use the for loop.
22:31:49*stefanos82 quit (Quit: Quitting for now...)
22:32:03Araqit's 2018, compilers can optimize for loops into copyMem
22:34:09*zachk joined #nim
22:36:46dom96but... for loop is more code than a simple copyMme
22:36:47dom96*mem
22:37:35Araqit's one line that crashes at runtime if you get it wrong
22:37:57AraqcopyMem is one line that crashes eventually, somewhere else if you get it wrong
22:39:20Araqnot to mention that you need a copyMem'able type T for copyMem to work. And usually the "I know better than the stupid Nim compiler" type of programmer doesn't know Nim all that well
22:40:48*zachk quit (Changing host)
22:40:48*zachk joined #nim
22:44:03xaceDo we have an equivalent of godbolt.org for nim?
22:45:13dom96Don't think so
22:45:19dom96It would be really nice to add Nim support there
22:45:33xacehttps://github.com/mattgodbolt/compiler-explorer/issues/206 # seems like it has been brought up
22:45:55xaceimho, it would help with these types of things, for example copymem vs forloop etc... and other uncertanties a user might have
22:59:27*Jesin quit (Quit: Leaving)
23:07:25Araqhttps://github.com/isocpp/CppCoreGuidelines/blob/master/docs/Lifetime.pdf this is hot shit
23:08:23Araqit shows how we can compute useful properties about nim programs without fixpoint iterations
23:09:01Araqalehander42: this should be interesting for you
23:10:13Araqhow they treat loops is pure genius
23:15:53*jjido quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
23:24:17*kapil____ joined #nim
23:49:30*zachk quit (Read error: Connection reset by peer)
23:49:58*zachk joined #nim
23:50:32*zachk quit (Changing host)
23:50:32*zachk joined #nim