<< 07-05-2015 >>

---

00:10:02	*	solidsnack quit (Ping timeout: 246 seconds)
00:38:01	*	Jaood joined #nim
00:46:39	*	darkf joined #nim
00:55:59	*	xificurC quit (Read error: Connection reset by peer)
00:56:34	*	xificurC joined #nim
00:59:49	*	silven quit (Ping timeout: 245 seconds)
01:00:55	*	silven joined #nim
01:03:41	*	zzach quit ()
01:06:10	*	dddddd quit (Ping timeout: 272 seconds)
01:08:49	*	vikaton joined #nim
01:19:50	*	Jesin quit (Quit: Leaving)
01:22:27	*	kz joined #nim
01:37:19	*	Pisuke quit (Ping timeout: 255 seconds)
01:47:09	*	Pisuke joined #nim
01:47:31	*	kz quit (Ping timeout: 246 seconds)
01:54:31	Xe	is there a way to set the contents of language pragmas at runtime?
01:55:35	Xe	specifically i'm looking for a way to change the `dynlib` pragma of a function or reload the shared object, or should I wrap dlopen / dlsym / dlclose for that?
01:56:34	*	vendethiel joined #nim
02:01:37	*	ddl_smurf joined #nim
02:02:02	fowl	Xe those funcs are available in dynlib
02:02:24	Xe	ah cool
02:03:02	fowl	Araq has some late-loading mechanism used for opengl extension loading
02:03:30	fowl	Its near the top of opengl.nim
02:04:34	Xe	aha
02:04:47	Xe	so I would cast the resulting call to the type of function I expect
02:05:04	fowl	Xe maybe if you have a macro that converts proc x(...) Into var x: proc... And load/reload it later
02:05:22	Xe	Hmm
02:05:45	Xe	do I need to keep a handle to the .so open for the function to keep working?
02:05:56	Xe	imported function*
02:06:20	fowl	You could call x() and changing the func x points to will update other calls (because they now use pointer to function behind the scenes)
02:07:26	fowl	Xe I think the DL has to stay open otherwise the symbols in it won't exist
02:13:06	*	Jesin joined #nim
02:15:01	Xe	add5Func = cast[proc (a: int) int {.cdecl.}](lib1.symAddr("add5"))
02:15:05	Xe	seems to fail compilation with
02:15:21	Xe	type mismatch: got (proc (a: int){.closure.}) but expected 'proc (a: int): int{.closure.}'
02:15:29	Xe	or did I miss a colon
02:15:59	fowl	It means add5func is a closure (bare proc type is closure)
02:17:32	Xe	ah
02:19:06	fowl	Er
02:19:22	fowl	No you did miss a colon
02:19:46	Xe	https://gist.github.com/Xe/3e3d18cef5602036660e
02:19:55	Xe	yeah I did
02:20:29	Xe	it's now failing with a different compile reason
02:20:44	fowl	You're importing a closure function on purpose?
02:21:09	Xe	I'm trying to import a function from a C .so
02:21:26	Xe	it's defined as a nim function though
02:21:39	Xe	https://github.com/Xe/nim-shared-object-example/blob/master/add5.nim
02:21:39	*	kz joined #nim
02:21:46	*	rgv151 joined #nim
02:21:53	Xe	this gets built as https://github.com/Xe/nim-shared-object-example/blob/master/Makefile#L2
02:23:28	fowl	Change those closure pragmas to nimcall
02:24:39	Xe	ah
02:24:41	Xe	okay
02:24:43	Xe	Yes that works now
02:24:45	Xe	thank you so much fowl
02:25:26	fowl	no prob
02:29:00	Xe	fowl: is there a way to make nim quietly compile?
02:31:50	saml_	aporia cannot find NimSuggest in PATH
02:36:48	fowl	Xe, try verbosity:0
02:37:10	fowl	saml_, are you on windows
02:37:19	saml_	no, linux
02:37:36	saml_	looks like compiler/nimsuggest doesn't get built
02:37:49	saml_	using devel branch of Araq/Nim
02:37:55	onionhammer	wish it was built as part of nims install, but it isnt
02:38:51	saml_	so i built it manually . now aporia finds it. but Go To definitnio segfaults
02:42:03	Xe	okay so as far as I've gathered
02:42:08	Xe	options before filenames with nim c
02:49:48	fowl	--run will pass the arguments after the file name to the exe
02:55:12	*	jholland quit (Quit: Connection closed for inactivity)
02:57:53	*	Jaood left #nim (#nim)
03:00:26	*	endragor joined #nim
03:01:54	*	BitPuffin|osx quit (Ping timeout: 245 seconds)
03:05:34	*	kz quit (Ping timeout: 246 seconds)
03:15:17	*	Xe quit (Ping timeout: 265 seconds)
03:17:13	*	saml_ quit (Remote host closed the connection)
03:22:10	*	vendethiel quit (Ping timeout: 255 seconds)
03:22:17	*	jj2baile_ is now known as jj2baile
03:22:31	*	rgv151_ joined #nim
03:23:28	*	rgv151 quit (Ping timeout: 272 seconds)
03:26:06	*	aboisvert joined #nim
03:26:24	*	Xe joined #nim
03:29:44	*	yglukhov__ joined #nim
03:30:15	*	bougyman quit (Ping timeout: 256 seconds)
03:34:23	*	yglukhov__ quit (Ping timeout: 264 seconds)
03:45:51	*	gyeates joined #nim
03:47:20	*	Kingsquee joined #nim
03:49:30	*	TEttinger joined #nim
03:56:02	*	darkf quit (Quit: Leaving)
03:56:20	*	darkf joined #nim
03:59:44	*	aboisvert quit (Quit: aboisvert)
04:05:14	*	Demon_Fox quit (Ping timeout: 245 seconds)
04:06:00	*	Demon_Fox joined #nim
04:12:26	*	endragor_ joined #nim
04:15:47	*	endragor quit (Ping timeout: 264 seconds)
04:19:53	*	kz joined #nim
04:21:38	*	endragor_ quit (Remote host closed the connection)
04:26:04	*	kz quit (Ping timeout: 246 seconds)
04:26:12	*	gyeates quit (Ping timeout: 256 seconds)
04:27:21	*	endragor joined #nim
04:27:47	*	gyeates joined #nim
04:29:53	*	Sembei joined #nim
04:30:36	*	endragor_ joined #nim
04:31:57	*	endragor quit (Remote host closed the connection)
04:32:21	*	EulersBackup quit (Ping timeout: 250 seconds)
04:33:09	*	Pisuke quit (Ping timeout: 256 seconds)
04:36:13	*	endragor_ quit (Remote host closed the connection)
04:40:11	*	endragor joined #nim
04:44:29	*	endragor quit (Ping timeout: 256 seconds)
04:45:42	*	endragor joined #nim
04:47:23	*	endragor quit (Remote host closed the connection)
04:50:39	*	endragor joined #nim
04:52:33	*	endragor quit (Remote host closed the connection)
04:52:50	*	endragor joined #nim
04:53:16	*	aboisvert joined #nim
04:56:47	*	endragor quit (Remote host closed the connection)
05:04:40	*	endragor joined #nim
05:07:37	*	endragor quit (Remote host closed the connection)
05:12:08	*	ddl_smurf quit (Quit: ddl_smurf)
05:13:25	*	vikaton quit (Quit: Connection closed for inactivity)
05:14:57	*	endragor joined #nim
05:16:41	*	filwit joined #nim
05:27:43	*	aboisvert quit (Quit: aboisvert)
05:32:17	*	endragor quit (Remote host closed the connection)
05:34:09	*	endragor joined #nim
05:37:37	*	endragor quit (Remote host closed the connection)
05:38:06	*	endragor joined #nim
05:44:06	*	BlaXpirit joined #nim
05:46:36	*	gsingh93 quit (Ping timeout: 240 seconds)
05:49:46	*	phira joined #nim
05:54:12	*	Jesin quit (Quit: Leaving)
06:05:21	*	dtscode joined #nim
06:05:56	*	gokr_ quit (Ping timeout: 240 seconds)
06:09:02	*	Jesin joined #nim
06:30:25	*	nande quit (Remote host closed the connection)
06:58:00	*	TEttinger quit (Quit: Leaving)
07:07:36	*	TEttinger joined #nim
07:08:41	*	yglukhov__ joined #nim
07:11:04	*	bjz joined #nim
07:16:00	rgv151_	is there a way to make macro return value?
07:16:34	ldlework	rgv151_: macros have to return ast, that's the point
07:19:49	rgv151_	i c
07:20:47	*	bjz quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)
07:21:12	*	bjz joined #nim
07:21:23	*	bjz quit (Client Quit)
07:21:47	*	Ven joined #nim
07:27:29	*	BlaXpirit quit (Quit: Quit Konversation)
07:27:38	*	endragor_ joined #nim
07:31:50	*	endragor quit (Ping timeout: 276 seconds)
07:33:21	*	OnO joined #nim
07:36:20	*	OnO quit (Client Quit)
07:36:41	*	bjz joined #nim
07:38:14	*	OnO joined #nim
07:40:57	*	JinShil joined #nim
07:42:55	*	HakanD joined #nim
07:44:22	*	OnO quit (Quit: My iMac has gone to sleep. ZZZzzz…)
07:45:25	*	endragor_ quit (Remote host closed the connection)
07:49:32	*	bjz quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)
07:50:21	fowl	neat https://github.com/MasonMcGill/numerics
07:58:28	dtscode	neater http://eleks.github.io/js2js/#features
07:58:52	ldlework	lol
07:59:15	*	bjz joined #nim
08:00:07	ldlework	dtscode: https://github.com/jterrace/js.js/
08:00:54	*	Triplefox quit (Remote host closed the connection)
08:01:07	dtscode	i love everyhing about this
08:01:26	*	Triplefox joined #nim
08:01:28	*	Trustable joined #nim
08:01:37	ldlework	oh my god all I want to do is go to sleep
08:01:44	ldlework	whhyyyy saltstack. whyyyyyyyy
08:02:08	dtscode	i want to sleep but compilers :/
08:03:08	r-ku	who happens to know how to check if type is ref type (in macros)?
08:03:41	*	bjz quit (Client Quit)
08:05:39	filwit	r-ku: if T is ref: ...
08:05:50	filwit	r-ku: oh, macros, sorry
08:05:57	filwit	r-ku: use 'getType()'
08:06:02	filwit	in macros
08:06:23	*	JinShil quit (Quit: Konversation terminated!)
08:06:34	r-ku	and check returned NimNode == nnkRefTy?
08:06:43	someplace	what's argv in nim?
08:07:00	filwit	r-ku: yes, basically.
08:07:03	r-ku	thanks
08:07:59	yglukhov__	someplace: os.commandLineParams
08:08:07	someplace	cheers
08:08:45	filwit	r-ku: another hint: if you're using getType() with a typeDesc, you might need to call it twice.. the first call will give you just the 'typedesc' AST, so you call it again on the second child to get the real type AST.. if you're working with an instance though you only need it once.
08:09:53	r-ku	hmm somehow i cant pull it off. i get type identifier from AST, i suppose thats not what getType really wants
08:11:08	filwit	r-ku: if you gist an example of what you're going for I can help you future.
08:11:14	filwit	further*
08:11:45	r-ku	https://gist.github.com/r-ku/79454071e33c89ed3485
08:12:16	r-ku	im exploring Jehans named constructor idea by rewriting proc to allocate and initialize type. saves one keyword
08:12:21	filwit	r-ku: you need to pass by 'typed' i think. getType doesn't work with 'untyped'
08:13:07	r-ku	still getting "node has no type"
08:13:29	filwit	r-ku: k one sec.
08:14:56	*	Jehan_ joined #nim
08:15:18	filwit	r-ku: i'm not even seeing getType() here, but maybe I'm just tired?
08:15:40	r-ku	https://gist.github.com/r-ku/79454071e33c89ed3485#file-gistfile1-nim-L14
08:15:54	filwit	r-ku: i'm not sure exactly what you're going for with this, (also, Jehan_ just signed on, if you're trying something he suggested, maybe he'd be better help)
08:16:06	Jehan_	Umm. what's the matter?
08:16:31	r-ku	im exploring your named constructors idea by rewriting proc instead of having make "keyword"
08:16:45	r-ku	just cant figure out how to check if type is ref or not inside a macro
08:16:47	ldlework	that's crazy that you can reimplement proc
08:16:49	fowl	Gettype(n).typekind== ntyRef
08:17:55	Jehan_	I'm not sure if you can if it's an immediate macro.
08:18:23	fowl	I was about to say that - it can't run on a proc macro like that
08:18:33	filwit	r-ku: here's how I do this: use a template + 'when T is ref' and call a macro based on that
08:18:41	fowl	The proc is passed before it is instantiated
08:18:55	r-ku	oh thats a good idea filwit, its probably most appropriate way
08:18:55	Jehan_	Immediate macros basically postprocess ASTs after parsing, but before semantic analysis.
08:18:59	*	MyMind joined #nim
08:19:12	*	EulersBackup joined #nim
08:19:17	fowl	5 examples of macroshttps://gist.github.com/fowlmouth/3c74409c30533bb0a5b4
08:19:25	Jehan_	filwit: You'd have to embed that in the macro output.
08:19:27	fowl	Macros.Gettype *
08:20:09	fowl	Gn
08:20:43	filwit	Jehan_: macro output? I'm not sure what you mean. Let me make a example for r-ku real quick (i'm doing this exact thing in my own macros)
08:20:54	r-ku	hey filwit but wait, for template to work proc will have to be analuzed/defined already. then i wont be able to rewrite it?
08:20:56	Jehan_	macro output = the code that the macro returns.
08:21:39	*	yglukhov___ joined #nim
08:21:40	filwit	oh, i see, cause you're using this as a {.pragma.} macro
08:21:53	r-ku	yup, cause i want to rewrite proc itself
08:23:06	filwit	r-ku: what I do is have an 'impl' macro which accepts a type and a body of procs
08:23:08	filwit	impl Foo:
08:23:19	filwit	proc init = echo me.name
08:23:22	filwit	type of thing
08:23:25	gokr	fowl: So... getType gives you the type of something in the AST, so its after semantic analysis?
08:23:30	r-ku	but its terribly ugly
08:23:43	gokr	Oh, he went to bed
08:23:46	fowl	Yea gokr
08:23:54	gokr	Now...
08:23:58	fowl	Irc from bed ;)
08:24:10	gokr	fowl: So if we want to go even further...
08:24:11	filwit	r-ku: i actually really like it for many other reasons, but what you're going for might be impossible ATM
08:24:44	r-ku	well that check is not critical, but it would take away ability to tell when one is shooting himself in the foot because macro expects non-ref type
08:24:49	gokr	fowl: I am thinking of even more reflection - but ... all the code hasn't been parsed of course, only in the current module or?
08:25:05	*	yglukhov__ quit (Ping timeout: 265 seconds)
08:25:25	fowl	gokr: in the current scope
08:25:46	gokr	fowl: Right, so any thoughts on having macros that can operate even later in the compilation process?
08:25:47	filwit	r-ku: let me try something real quick and get back to you
08:25:59	r-ku	thanks for help
08:26:30	fowl	gokr: sure lots of times I'm using macros to build some compile time state and another macro that acts on it later
08:27:11	gokr	Just curious on what the "next step" is in this field.
08:27:47	fowl	Remind me tomorrow I'll make an example
08:28:01	gokr	fowl: Sure, cool stuff
08:28:30	gokr	I can't help contrasting it with the full reflection of Smalltalk and trying to see where it can go.
08:29:02	*	endragor joined #nim
08:34:47	ekarlso	any cool things lately in nim ?
08:38:29	filwit	r-ku: yeah i'm pretty sure this is impossible. You have to pass the Type as 'typed', but you're trying to pick it up from the proc which you have to pass as 'untyped'
08:39:07	*	Trustable quit (Remote host closed the connection)
08:39:25	filwit	r-ku: it's unfortunate really. You could do it like this: proc init {.ctor:Foo.} = ...
08:39:36	r-ku	im no longer even sure if i can cover both ref and non-ref with one constructor declaration
08:39:43	*	Trustable joined #nim
08:40:17	gokr	ekarlso: Did you see nimpdf?
08:40:32	ekarlso	gokr: nope :p
08:41:00	gokr	https://github.com/jangko/nimpdf
08:41:03	gokr	Quite impressive
08:41:16	filwit	r-ku: if you're going for named constructors, why not simply overload by type, eg: proc init(T:typedesc[Foo]) = ...; let f = Foo.init()
08:41:38	ekarlso	gokr: meh, I would like to see more network / services libraries in nim
08:41:46	ekarlso	like http stuff, protobuf etc
08:41:59	r-ku	filwit: its pretty much what i got now, or close to it
08:42:03	filwit	r-ku: you could make a {.ctor.} macro which basically just turns 'proc init(f:var Foo)' into 'proc init(f:typedesc[Foo])'
08:42:11	r-ku	init(T: type Foo, ...)
08:42:25	gokr	ekarlso: That doesn't make nimpdf less cool ;)
08:42:26	ekarlso	gokr: it's where stufl ike Go has gotten tons of attention ;p
08:42:33	r-ku	type Foo same as typedesc[Foo]?
08:42:38	filwit	yes
08:42:48	r-ku	yeah so thats what i got
08:43:04	r-ku	but i wanted basically to define single constructor that later worked for (ref Foo) too
08:43:19	gokr	ekarlso: Sure
08:44:21	*	Demon_Fox quit (Quit: Leaving)
08:44:49	r-ku	i be damned, it can be pulled off by doing so: proc init[T: Foo|(ref Foo)](self: var T) {.ctor.} =
08:45:31	filwit	r-ku: the problem is ref-types need to be allocated
08:45:50	r-ku	for some reason Foo() work in both cases somehow
08:45:57	filwit	though technically you could just make the 'ctor' macro inject 'self = T()' as the first line
08:46:08	r-ku	ah wait not in both heh
08:46:20	r-ku	yes thats what im doing already too
08:46:52	filwit	well then that should work, since Nim's default constructors allocate ref types
08:47:03	r-ku	actually non-ref type misbehaves, field does not get set
08:47:20	*	Trustable quit (Remote host closed the connection)
08:47:20	filwit	though the real solution here is build this pattern into Nim like I was planning on doing..
08:47:32	filwit	r-ku: ah.. right
08:47:47	r-ku	this is what i got so far if interested https://gist.github.com/r-ku/79454071e33c89ed3485
08:47:59	*	Trustable joined #nim
08:48:12	fowl	We have constructors that you manually write already. In the std lib they start with init or new
08:48:41	filwit	those are not constructors
08:48:42	*	Trustable quit (Remote host closed the connection)
08:48:51	filwit	those are allocators
08:49:02	fowl	Heh oh
08:49:22	fowl	And type.init isn't an allocator. It's a constructor?
08:49:27	*	bluenote joined #nim
08:49:28	*	Trustable joined #nim
08:49:33	*	Trustable quit (Remote host closed the connection)
08:50:04	r-ku	fowl: those constructors are abomination
08:50:26	filwit	fowl: i don't see an 'init' in system, where are you finding this?
08:50:45	r-ku	filwit: i believe initTable() has that
08:50:53	*	Trustable joined #nim
08:50:55	fowl	I said "start with"
08:51:03	filwit	ah, okay
08:51:47	fowl	Ref-agnostic "allocator" is possible
08:52:10	filwit	there are many reasons I don't like this and think Nim should have a real constructor pattern (one that triggers on it's built-in constructors).. I need to make that now I think.
08:53:29	Jehan_	What's a "real constructor pattern"?
08:53:36	r-ku	filwit: on thread i posted Jehan_ pointed out that T(x) is problematic because it is already used for converting to other type. then it becomes ambiguous, is it a conversion or ctor w/ one param. maybe you have ideas for fixing that?
08:54:30	filwit	actually that's not really a problem, in fact constructors with a single parameter would just work as a way to overload the conversion
08:54:48	gokr	Jehan_: Your little "make" macro that enables the "make MyType.someProc(x,y,z)" style, was that mainly to demonstrate techniques or a serious idea?
08:55:18	Jehan_	gokr: It was basically an adaptation of Eiffel's constructors.
08:56:05	gokr	Jehan_: Because... I can't really see its benefit. Ok, so it is visually standing out, and its separately specifying the type - but I mean... I don't see the exact benefits.
08:56:45	Jehan_	gokr: I'm personally fine with `newT` and `initT` myself.
08:56:48	filwit	Jehan_: we've talked about this on the forums already. Basically we have a few issues with just using named function to allocate things (there's too many to list here, and I can't remember them all of the top of my head right now).
08:57:37	Jehan_	filwit: Yes, and I think the issues are imaginary.
08:58:10	gokr	Personally if we a) added ability to restrict access to the current object literal "constructor" and b) added some kind of pragma to mark procs as being constructors - then life seems fine to me.
08:58:46	gokr	filwit: I would actually like to see such a list of issues in more detail.
08:59:10	filwit	Jehan_: yes i know your position on it, but I disagree (however you could, of course, be right). Either way, I think it wouldn't hurt for me to implement it and then debate about how it works instead of just talking theory.
08:59:32	Jehan_	The primary issue seems to be that Nim has a different look and feel from C++/C#/Java.
09:00:01	filwit	no that's not the issue at all, at least, not any issue I was ever talking about
09:00:41	r-ku	my main issue with constructors is that 1. they are not tied to type in any way and 2. there can be initSomething|cookSomething|makeSomething|newSomething|blahblahblaSomething
09:00:48	Jehan_	For example, while Scala has constructors (if only for Java compatibility), they are typically only used internally in a package and the companion object has Nim-like constructor fucntions.
09:01:05	r-ku	now what Jehan_'s named constructors fix is Type.<ide here can return list of constructors>
09:01:13	r-ku	this makes things cleaner A LOT
09:01:17	Jehan_	r-ku: Those are features, not bugs.
09:01:31	r-ku	feature to make as much mess as one wants?
09:01:34	Jehan_	You do not want to have constructors tied to types. That's a leaky abstraction.
09:01:49	filwit	gokr: the discussion about my constructor concept (and Jehan + other's arguments against it) starts on this page: http://forum.nim-lang.org/t/703/6
09:01:53	r-ku	i have no idea what that means Jehan_
09:02:09	gokr	r-ku: In Smallltalk which has "named constructors" there is no mess.
09:02:16	r-ku	but if you want two types use same constructor - just make overload and call other ctor
09:02:28	Jehan_	It means that you cannot have new T() return an instance of U, for example, where U is a different type.
09:03:06	r-ku	i do not understand why one would call "new T()" if he wants U, then call "new U()"
09:03:11	Jehan_	This can occur, for example, if you change the implementation of T and U is a proper subtype of T.
09:03:11	filwit	but you use named constructors for those, because they're not 'constructors' but 'factories'
09:03:28	Jehan_	r-ku: Because of information hiding.
09:03:31	gokr	In Smalltalk one could however argue that the class is visible at the "call site" - but... if we marked "constructor procs" then an IDE can easily show us all these places too. And Nim is statically typed so the information is there AFAIK.
09:03:51	Jehan_	filwit: Yes, and once you've made something a "constructor", you cannot change that without breaking the API.
09:04:11	r-ku	Jehan_: didnt you mean U being proper supertype of T?
09:04:37	r-ku	actually sorry
09:04:39	Jehan_	r-ku: No, I meant subtype. A specialized implementation of T.
09:04:40	r-ku	my logic got messed
09:04:45	r-ku	indeed
09:05:06	Jehan_	filwit: Most people in favor of constructors are talking aesthetics. I'm talking about the software engineering aspect.
09:05:15	r-ku	but this very well reminds me c++ implementation hiding with virtual header stuff
09:05:27	r-ku	does this even apply in nim?
09:06:05	gokr	Nim not having constructors á la Java/C++ etc is a huge virtue IMHO. Of course, I come from Smalltalk, so its natural to me.
09:06:33	filwit	Jehan_: well i'm not fully understanding the problem you'd addressing really. I don't see how 'breaking the API' is any more a problem then changing the parameters of a proc is 'breaking the API' which will happen all the time. But again, I don't see what you're referring too exactly
09:07:00	Jehan_	gokr: Yeah, it's a matter of background, I think. I come from using languages like Eiffel and ML, so I don't get the obsession with using C++-style constructors.
09:07:15	gokr	I just ask people that haven't used a lang with named constructors - to not just say "its a mess". Because its not.
09:07:21	filwit	Jehan_: and I was never really talking about the aesthetics. Personally I was always happy with Type.new(...) overloading
09:07:31	Jehan_	filwit: I'm not changing the parameters or the result type of the proc in my example.
09:07:45	filwit	but then it's an invalid comparison?
09:07:45	Jehan_	filwit: Type.new has the exact same problem.
09:08:05	Jehan_	filwit: No, because I'm not using constructors. You are.
09:08:12	Jehan_	Hence my API doesn't break and yours does.
09:09:09	filwit	I don't see how this is actually a problem though. You's saying that if you change a fundamental type in your API, that all the constructors of it will break? Perhaps you could give an code example for clarity.
09:09:45	Jehan_	filwit: No.
09:10:11	Jehan_	I can write makeMeATLikeThing and have it return an instance of either T or U.
09:10:56	Jehan_	When you write `new T`, you're stuck with an actual instance of T.
09:11:07	Jehan_	Because new T hardcodes the allocation of an instance of T.
09:11:34	Jehan_	This is also why C++ has all these other hacks like placement new.
09:12:03	gokr	One very important thing you often do in Smalltalk - is to find all instantiations of a class. There are basically two ways in Smalltalk - either look at all references to the class (catches slightly more than all instantiation sites) or look at all callers of the specific named constructor (if there is one). In Nim Araq mentioned we can probably identify "constructor procs" with heuristics, or else I can live with a pragma marking them.
09:12:12	filwit	yeah sure, but now lets talk about how often you want a constructor to return T or U vs the places you just want a convenient way to construct an object.. and nothing about constructors prevents you from making named constructors which do what you're suggesting.
09:13:07	ldlework	gokr: why would I ever need to do that in Nim?
09:13:08	Jehan_	filwit: Constructing an object isn't hard.
09:13:25	gokr	ldlework: Find instantiation sites?
09:13:38	filwit	Jehan_: i agree, but it could be improved. That's my point.
09:13:42	ldlework	or all instances of a class or any other craziness you just mentioned
09:14:02	gokr	ldlework: Are you serious?
09:14:14	ldlework	gokr: I've been writing code for a long time
09:14:25	ldlework	I simply have never needed to do either of those things
09:14:31	ldlework	I'm asking you why you would
09:14:37	*	Ven quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
09:14:46	Jehan_	ldlework: Debugging?
09:15:00	gokr	There are tons of reasons. But if you haven't used advanced IDEs on a very large codebase - then fine.
09:15:01	Jehan_	ldlework: Right now I'm trying to hunt down a bug in the GC.
09:15:05	ldlework	Jehan_: tools do that, so okay maybe gokr is writing a debugger?
09:15:11	filwit	Jehan_: one of the biggest ways it could be improved that has nothing to do with aesthetics, is by making it impossible to use the constructor syntax T() where you want explicitly set variables T(1, 2), that need to be constructed (eg, they can't just be marked 'not nil' or are obj type)
09:15:29	ldlework	gokr: are you writing an IDE/debugger?
09:15:45	Jehan_	For which it would be very helpful to be able to enumerate all instances of a type. (Though, in fairness, the problem is here that I'm doing this in a part of the code where I can't do memory allocations.)
09:15:46	ldlework	gokr: you said it was 'very important' but that seems more like 'extremely narrow'
09:15:54	ldlework	I was just thrown off
09:15:56	gokr	ldlework: No, its not narrow
09:16:05	ldlework	gokr: what other usecases are there for that?
09:16:10	r-ku	ehm either of you would happen to know how i could make variable alias? result = T(); var self = result; makes copy of self into result. i want self to be alias of result. possible? (non-ref type)
09:16:30	gokr	ldlework: There are literally tons of reasons for me to want to find all places a specific class is instantiated or otherwise used.
09:16:42	ldlework	I'm asking what some of them are.
09:16:56	Jehan_	filwit: Not sure I understand what you mean there?
09:17:10	gokr	Ok, so let's say I want to change class X, I want to find all its usages.
09:17:12	Jehan_	If it's about exposing the type without the constructor, yeah, that's in need of a clean-up.
09:17:23	ldlework	gokr: we already covered "features of your ide so it can refactor for you"
09:17:28	gokr	Or let's say I am learning about class X - I want to look at all the code using class X to see sample code
09:17:39	ldlework	gokr: okay so "IDE refactor features"
09:17:47	gokr	ldlework: The whole Smalltalk env is written in itself.
09:17:54	ldlework	gokr: okay?
09:18:31	gokr	ldlework: You lost me. Let me go back a few steps here
09:18:43	ldlework	gokr: you said this was 'very important'
09:18:58	gokr	I am talking about the ability for tools/compilers to find all instantiation sites.
09:18:59	ldlework	but it doesn't seem very important outside of 'ide reflection features'
09:19:08	ldlework	gotcha
09:19:17	gokr	And tools can't do that easily unless there is something to go on.
09:19:24	ldlework	Yeah. That's been established.
09:19:28	gokr	Right now we don't have constructors.
09:19:31	ldlework	I guess that's the opposite of narrow to you.
09:19:32	gokr	We have procs.
09:19:59	gokr	So in Nim - if I want to see all places where we instantiate class X - what do I do?
09:20:07	gokr	Or rather, what does the tool do?
09:20:15	ldlework	I understand, what you want, and what you need.
09:20:30	ldlework	I was puzzled by how this is very important, since I have never needed this from a language for any software I personally have written.
09:20:37	ldlework	I hope you can understand my small momentary confusion
09:20:51	ldlework	And that it is not about what features are needed to implement IDE features.
09:20:53	gokr	ldlework: I am trying to understand it - but ... I am slightly perplexed.
09:21:04	Varriount	gokr: Look at every possible code path that calls a procedure that instantiates or creates a new instance of type X?
09:21:11	filwit	Jehan_: yes, basically. But my proposal did that + makes it easy to overload the attractive constructor syntax (which has the least amount of noise, so people will always want to use it IMO). So I don't see how we loose anything with that. We still have named constructors for the places you want, we get the ability to enforce a specific constructor is called (beyond manually allocating and setting memory), and we get to extend the sexy
09:21:11	filwit	constructor syntax that is basically useless for anything more complex than a Vector type.
09:21:23	ldlework	gokr: K
09:21:30	gokr	Varriount: Yeah, basically.
09:21:45	gokr	In Smalltalk we use a lot of cross referencing tools.
09:21:46	Jehan_	gokr: Good programming style should have all constructors for a type embedded in a module that declares that type.
09:21:49	Varriount	gokr: Which isn't an easy task.
09:21:58	gokr	Its very common to pull up a browser on all methods that reference class X.
09:21:59	filwit	basically useless.. **right now**
09:22:16	Jehan_	This generally makes it easy to find them without tools.
09:22:20	gokr	Jehan_: Yes, but that module has lots of procs.
09:22:49	gokr	Then ldlework asked why on earth I would want to find all instantiation sites.
09:23:05	Jehan_	filwit: So, back to aesthetics because it's "the attractive constructor syntax"? (That I personally don't find attractive at all, but that's subjective.)
09:23:16	ldlework	gokr: *other than implementing an ide feature*
09:23:22	gokr	I presume anyone who have used say Eclipse in a large Java codebase - have often pulled up "all references to class X"
09:23:36	gokr	Which is slightly wider net of course.
09:23:45	Jehan_	gokr: You have the same problem with a class that implements tons of constructors.
09:23:59	gokr	But in Eclipse one can also show all calls to a specific constructor.
09:24:04	filwit	Jehan_: it's not just about aesthetic appeal, it's about extending what people will gravitate towards (because of aesthetics) to be better
09:24:05	gokr	Jehan_: Yes.
09:24:45	gokr	But my original point was that I would like for "Nim" to be able to distinguish constructor procs. Either by pragma marking them, or as Araq says - it can be done with heuristics.
09:24:54	Jehan_	gokr: Here's a question: Is diagonalMatrix(n: float) = identityMatrix() * n a constructor?
09:25:11	Jehan_	I.e. if a constructor is implemented purely in terms of another constructor, is it still a constructor?
09:25:47	gokr	Jehan_: I agree that Nim has many more "mechanisms" that ... the term constructor is not as easily defined as say in Java or Smalltalk.
09:26:22	gokr	ldlework: Haven't you ever done a "show me all places where this constructor is called" in say Eclipse?
09:26:29	Jehan_	gokr: It's more that, for example, a purely functional language cannot really have constructors in the OO sense.
09:26:32	RegisterOn	http://forum.nim-lang.org/t/703/2
09:26:56	Jehan_	You instantiate all variables at once because they can't be mutated later on.
09:27:22	Jehan_	Advanced "constructors" then have to call upon the basic data constructors.
09:27:54	ldlework	gokr: thousands of times. you keep just ignoring my adjustment to the conversation in that, you said very important, this struck me as odd and I wanted to know more, you said language features, I said okay that seems narrow for something that's very imporatnt anything else, and you are now just repeating ide features over. I don't know what else to say. I wasn't attacking you I was just curious what you meant.
09:28:25	ldlework	I agree that reflection is both awesome and useful for ide features. No disagreement at all.
09:28:35	Jehan_	Anyhow, I have to tune out for now (work and finding the GC bug).
09:29:04	gokr	ldlework: Btw, I didn't say "all instances of a class" - I said instantiation sites. All instances is something else.
09:29:34	ldlework	gokr: which smalltalk must also do
09:29:38	ldlework	since it can reload instances
09:29:42	ldlework	(which is crazy)
09:29:48	Jehan_	gokr: Eh, that's easy. All calls of T(...) and all new(x) calls where x is of type T.
09:29:55	gokr	ldlework: yes, but I wasn't talking about that. And no, its not crazy.
09:30:01	Jehan_	Anyhow, gotta go.
09:30:05	*	Jehan_ quit (Quit: Leaving)
09:30:08	ldlework	gokr: crazy like a fox
09:30:13	ldlework	(ie, crazy awesome)
09:30:20	gokr	ldlework: ah
09:30:38	gokr	ldlework: But I am actually trying to answer you. Let me reread what you wrote above
09:31:12	ldlework	gokr: I thought you were saying that finding the instantion sites of a class was 'very important' in that it is important day-to-day or even for a wide range of tasks in the language.
09:31:21	ldlework	I was like "what could you possibly use that for"
09:31:49	ldlework	you brought up IDE stuff, which yeah is obvious. I asked if you could think of anything else, but mentioned more ide features. That's fine. Maybe that's all its good for.
09:32:06	gokr	Ok, so I use it all the time in Smalltalk - because in a large codebase I need to find all these places.
09:32:20	ldlework	Yeah but you don't use that as a way to implement features in your program
09:32:22	ldlework	That's what I meant
09:32:43	gokr	Eh... no? Did I ever claim that?
09:32:45	ldlework	You don't use those features to implement normal applications, usually, unless that application is some ide tool
09:32:58	ldlework	gokr: nope, you didn't. You said it was 'very important'
09:33:02	gokr	I just said I want my tools to be able to do it.
09:33:16	ldlework	And *I* wondered how it could be important, like, from a day to day persepective
09:33:24	gokr	And they can't do it if the procs are "just like all other procs"
09:33:40	ldlework	I know! but you'll just have to trust me, because I'm gonna stop saying it now.
09:33:59	ldlework	gokr: pharo seriously needs just a basic updated walkthrough
09:34:24	gokr	I said "One very important thing you often do in Smalltalk - is to find all instantiations of a class"
09:34:30	ldlework	Like no one has to update the whole by example book, just a new tiny little beginner's article
09:34:39	gokr	Ok, so you thought I was describing "code" that does that.
09:34:50	ldlework	gokr: yes, I interpreted that to mean "this is something you literally d oall the time while writing smalltalk applications"
09:35:02	ldlework	rather than "this is an important feature of smalltalk to support is many awesome ide features"
09:35:22	gokr	But let me then rephrase: "One very important thing you often do when you work in Smalltalk - is to have your tools finding all methods that instantiate a class."
09:35:29	ldlework	haha yes
09:35:32	*	filwit quit (Quit: Leaving)
09:35:33	ldlework	that's what I get you to mean now
09:36:05	gokr	"while writing" - yes. "in code" - no.
09:36:32	gokr	ldlework: Did you try the "PharoTutorial go"
09:36:39	ldlework	yeah but it was too tedious
09:36:51	ldlework	pharo is very click, type, click type
09:36:57	ldlework	and omg, I hate clicking
09:37:00	ldlework	and using the arrow keys
09:37:02	ldlework	:(
09:37:14	gokr	I am actually a really bad person to ask for learning advice on Smalltalk - since I haven't read a book on Smalltalk in like 15 years.
09:37:19	ldlework	haha
09:37:33	*	filwit joined #nim
09:37:58	ldlework	gokr: what would be really cool to see, is a small opengl demo, with a commandline repl, that allowed me to change the scene in realtime
09:37:59	gokr	Btw, there are different Smalltalks - GNU Smalltalk is more Vim style.
09:38:14	ldlework	or something similar to that anyway
09:38:16	gokr	But why a command line repl?
09:38:21	ldlework	so I can interact with it
09:38:27	gokr	You can type and run code everywhere in Pharo.
09:38:28	ldlework	I mean, a full editor would be cool too
09:38:34	ldlework	I know, but it sucks, as an editor
09:38:42	fowl	Lol
09:38:42	gokr	Literally *any* place you can type code - you can run it.
09:38:47	ldlework	I know
09:38:56	ldlework	but as an interface for editing text, its really horrible
09:39:07	gokr	So you think a REPL is better?
09:39:10	ldlework	trust me, I'm amazed by the actual environment
09:39:14	ldlework	gokr: if it has readline? yes.
09:39:34	gokr	I never understood why people love typing in line by line in a single window that echoes something back.
09:39:43	ldlework	gokr: well, emacs integration would be ideal
09:39:51	ldlework	I just figured repl wasn't asking for too much
09:40:00	ldlework	like, I underbid, by saying repl
09:40:18	ldlework	anything that is decent at editing text, unlike pharo
09:40:20	gokr	When I instead can use lots of tree explorers, inspectors etc - and run code and look at stuff and drill - in many windows in parallel.
09:40:24	filwit	ldlework: https://www.youtube.com/watch?v=db-7J5OaSag
09:40:30	fowl	ldlework: pharo included sdl2 so I bet it uses opengl (whether its exposed for you or not, idk)
09:40:37	ldlework	gokr: all that stuff is nice assuming you can /type text/
09:40:45	ldlework	gokr: hehe, I'm only half teasing though
09:40:48	ldlework	it was kind of painful
09:40:49	*	antranigv1 joined #nim
09:41:04	ldlework	filwit: heh nice
09:41:09	gokr	ldlework: There is some Emacs/Vim thingymabob to give those key bindings
09:41:24	ldlework	gokr: oh that would be wonderful
09:41:29	gokr	fowl: Pharo can do OpenGL in many ways
09:41:43	filwit	ldlework: i know i thought that was pretty cool.. it's for Oculus, which I have.. but I haven't tried this one yet..
09:41:55	ldlework	filwit: if I could get this with multiplayer, my life would be complete
09:41:59	ldlework	I yearn for the metaverse so bad
09:42:09	ldlework	and I don't mean secondlife, I just mean empty room that we can both script at the same time
09:42:16	filwit	watching yoru program come to live before your eyes in VR would be absolutely awesome though
09:42:17	ldlework	seriously, I would never leave there.
09:42:23	ldlework	filwit: agreed
09:42:27	filwit	lol, true
09:42:33	ldlework	I would make little artificial life experiments
09:42:37	ldlework	and give them names...
09:42:49	ldlework	I'm squeeee'ing so hard at the thought
09:42:49	gokr	filwit: That looks similar to what we can do in our system.
09:42:55	filwit	the only problem is Oculus doesn't have enough resolution to make coding in it comfortable
09:43:07	filwit	gokr: yeah i was just thinking that
09:43:21	r-ku	im nearly done with my macro, check it out here: http://forum.nim-lang.org/t/1190/2#7420
09:43:26	fowl	Haha that video. "I'm in three dimensions, I need a two dimensional space to work on though"
09:44:19	filwit	fowl: actually they're making a 3D compositor for Wayland (at least someone started it).. looked pretty cool, and basically breaks down everything we know about how to right compositors so far (I mean.. what IS a 3D "window" anyways)
09:44:28	Varriount	https://www.youtube.com/watch?v=VzFpg271sm8
09:45:19	antranigv1	anyone thought of nim for pedagogical use?
09:45:42	antranigv1	for 15yo kids, say (: as a replacement for pascal/python in school
09:46:07	*	antranigv1 thinks to post the question in the forums
09:46:09	gokr	Here is a video showing OpenCobalt which is same system as ours: https://www.youtube.com/watch?v=1s9ldlqhVkM
09:46:22	*	antranigv1 is now known as antranigv
09:46:26	r-ku	antranigv1: i bet it could replace pascal, but python is still easier
09:46:44	ldlework	gokr: the probelm I have with systems like these, is that they are optimized to simulate offices and avatars and such
09:46:55	antranigv	as a replacement for pascal, I'd love to see Oberon in schools
09:47:02	r-ku	tbh pascal in schools is terrible terrible decision, better teach them language that is actually useful in industry nowdays
09:47:04	antranigv	nim and oberon are similar in very ways.
09:47:11	Varriount	r-ku, antranigv: Nim still has some really pointy bits that stick out and trip you up.
09:47:15	gokr	ldlework: Our system is not "optimized" for that.
09:47:44	antranigv	no, I disagree. pascal is very goo, they (the kids) should know what is programming, not what is pascal (a language)
09:47:46	fowl	Oberon and pascal? Where do you go to school that teaches those
09:47:59	antranigv	I learned pascal in school in Syria
09:48:01	gokr	ldlework: Its in fact a generic synchronized object memory - that models a 3D space.
09:48:24	ldlework	yes and then there is a ton of stuff built ontop of that foundation right?
09:48:27	antranigv	oberon was a relpacement, no nothing changed, cz of the war
09:48:31	ldlework	that costs cycles and memory?
09:49:08	gokr	ldlework: Well... cycles and memory... not sure what you mean.
09:49:11	ldlework	gokr: can I run your thing easily, without any assumptions? multiplayer black programming void?
09:49:41	ldlework	or do I need a landscape and an avatar and so on
09:50:08	filwit	actually Nim would make a great educational language. It's safe enough and easy enough for beginners, is a natural stepping stone from Python in some sense, and can be used to teach all the low-level stuff without having to deal with C
09:50:33	ldlework	gokr: woah lorenz attractor demo is kind of exactly what I'm imagining
09:50:54	gokr	ldlework: Our system is a product for collab, not a programming framework. But our "worlds" can be very custom, its not hardwired.
09:50:59	*	HakanD quit (Quit: Be back later ...)
09:51:19	gokr	Croquet was built as a generic platform for shared 3D worlds, in whatever shapes and forms.
09:51:26	ldlework	gokr: okay, so you and I can't jump into a space and code collaboratively on things that are simulated in the same space?
09:51:46	ldlework	to me, this is what the word 'metaverse' means
09:51:49	ldlework	not these virtual offices
09:51:57	fowl	I too am interested in multiplayer smalltall
09:52:02	gokr	ldlework: In our product we don't expose that. Croquet however does, and we build on the same codebase.
09:52:24	ldlework	gokr: I tried to look at Croquet before, and I failed to use it.
09:52:31	*	gokr going to nim-offtopic
09:53:38	*	antranigv joins gokr
09:58:48	*	Ven joined #nim
10:03:06	*	bjz joined #nim
10:14:56	*	Ven quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
10:20:36	*	bjz quit (Ping timeout: 240 seconds)
10:35:03	rgv151_	i c
10:35:08	*	rgv151_ quit (Quit: Leaving...)
10:36:54	*	vikaton joined #nim
10:42:22	*	yglukhov___ quit (Ping timeout: 272 seconds)
10:43:00	*	yglukhov___ joined #nim
10:44:09	*	bjz joined #nim
10:48:18	*	JinShil joined #nim
10:52:10	*	HakanD joined #nim
10:53:01	Senketsu	umm is there any reason why importing a wrapper (db_sqlite/sqlite3 for example) causes double free error and such with threading? any way to prevent that ?
10:55:30	Senketsu	actually importing only sqlite3 doesnt seem to cause that
10:56:29	*	HakanD quit (Ping timeout: 245 seconds)
11:15:23	*	arnetheduck joined #nim
11:24:09	*	wb joined #nim
11:28:57	*	BlaXpirit joined #nim
11:34:28	*	HakanD joined #nim
11:52:54	*	milosn quit (Quit: leaving)
11:55:52	someplace	http://hastebin.com/hoqaqajoye.avrasm <- any idea how I'd debug this?
11:56:02	someplace	using the nim-lang/irc module
11:58:16	*	Jehan_ joined #nim
12:01:50	*	saml_ joined #nim
12:05:52	Senketsu	someplace, would check all strings first, probably you are accessing non-initialized string ?
12:07:52	*	gyeates quit (Ping timeout: 272 seconds)
12:08:36	*	drewsrem joined #nim
12:09:45	someplace	hm.
12:10:14	def-	Looks more like a GC bug to me
12:10:22	def-	maybe there's an existing Issue about it
12:10:22	someplace	I don't see where it would be happening in this code, but I'm worried that it's a GC or async bug
12:10:24	someplace	hm
12:10:32	someplace	I'll dig around
12:11:55	Jehan_	I've just found a GC bug (well, not technically a bug in the GC, but a result of the interaction of setjmp() and some compiler optimizations).
12:12:00	Jehan_	Not sure if it's related.
12:12:05	someplace	https://github.com/Araq/Nim/issues/2650 < same lib, no GC in backtrace though
12:12:34	Jehan_	GC in backtrace often shows up as a side effect of memory corruption.
12:12:54	someplace	that's fun
12:12:56	Jehan_	For example, if you do a copyMem() that's overwriting stuff, then you will probably wreck reference counts.
12:13:19	Jehan_	But that's why you want to be careful with the low-level primitives.
12:14:27	someplace	it only happened once, overnight, hasn't happened again yet
12:21:12	*	kokozedman joined #nim
12:22:29	kokozedman	hey guys, I'm trying to generate C++ codes, but I also use mips for OpenWrt. Specifying the compiler is mips.linux.gcc.exe for the C compiler, but what is the config for the C++ compiler definition?
12:25:17	def-	mips.linux.gcc.cpp.exe maybe
12:30:28	kokozedman	def-, yeah, that worked... thanks
12:32:04	Jehan_	def-: I've found the bug (I think).
12:32:21	def-	Jehan_: great
12:32:46	Jehan_	Basically, jmp_buf may not always be word-aligned.
12:32:53	drewsrem	It seems to me that compiler/nimlexbase.nim is the exact same as lib/pure/lexbase.nim only that it uses llstream.nim instead of "normal" streams. - Is it deliberate that llstream is not "general" enough to put into the stlib or does the stream-interface not allow the efficiency that llstreams provide?
12:33:02	Jehan_	In which case scanning the registers will not find the pointers therein.
12:33:21	Jehan_	It can also with --opt:size result in no addresses on the stack being found at all.
12:34:12	def-	Weird that this didn't pop up before
12:34:40	*	Kingsquee quit (Quit: Konversation terminated!)
12:36:17	filwit	r-ku: replied to your forum post
12:40:12	Jehan_	def-: It is pretty rare. It becomes fairly reproducible with --opt:size on OS X.
12:40:29	Jehan_	But on OS X, clang is the default, which doesn't do what gcc does.
12:40:46	Jehan_	Basically, sizeof(jmp_buf) == 0x94 on OS X.
12:41:16	Jehan_	clang still aligns it properly; gcc can sometimes put it at an address that isn't word-aligned.
12:42:46	Jehan_	I'm more concerned at the moment that C structs on the stack with __attribute__((packed)) can throw things off.
12:44:01	*	Matthias247 joined #nim
12:44:21	Jehan_	If packed structs are being used, then the entire stack has to be scanned in byte rather than word increments.
12:44:39	*	banister quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
12:46:40	*	gyeates joined #nim
12:50:35	def-	Hm, sounds bad
12:53:55	Senketsu	hmm only impure libs cause the seqfaults on import .. weird
12:54:43	*	JinShil quit (Quit: Konversation terminated!)
12:56:20	gyeates	so is nimble the successor of babel?
12:56:31	gokr	yup
13:02:54	BlaXpirit	s/the successor of//
13:03:14	BlaXpirit	just a rename
13:03:25	*	vikaton quit (Quit: Connection closed for inactivity)
13:03:49	drewsrem	In what "metasyntax" is nims grammar written in? https://github.com/Araq/Nim/blob/devel/doc/grammar.txt i.e. how do I read it?
13:04:12	cryzed	probably BNF?
13:05:53	drewsrem	I don't think so
13:06:24	drewsrem	But I guess it's a mixture
13:09:44	drewsrem	There's stuff like ^* ^+ e.g.: exprList = expr ^+ comma
13:16:56	drewsrem	Ahhh stupid me, it's right in the beginning of the manual
13:19:50	Jehan_	def-: Yeah. At least it looks as though if you're just putting packed C structs on the stack that it won't affect the positioning of other variables. So, as long as you don't haverefs within one (which you shouldn't, according to the manual), then you'll be good.
13:25:45	*	saml_ quit (Ping timeout: 240 seconds)
13:29:42	*	BitPuffin|osx joined #nim
13:37:09	*	BitPuffin|osx quit (Ping timeout: 256 seconds)
13:42:51	someplace	oh
13:42:57	someplace	re: that crash I found this in the compile low
13:42:59	someplace	log, even
13:43:01	someplace	lib/nim/pure/asyncdispatch.nim(1196, 8) Warning: 'cb' is not GC-safe as it accesses
13:43:03	someplace	'nameIterVar' which is a global using GC'ed memory
13:43:08	someplace	is that known? an issue?
13:44:17	fowl	A warning
13:45:56	someplace	this is re: a gc crash with asyncdispatch in the traceback
13:46:15	Xe	How do you define a c struct in nim?
13:46:29	Xe	is it just an object with a cexport pragma?
13:46:54	someplace	you probably want the packed pragma too
13:47:03	fowl	To export it yea
13:47:57	fowl	Packed just adds __attribute(packed)
13:48:23	Xe	do you know of a code sample?
13:49:03	fowl	Xe: its 'exportc' are you having trouble?
13:49:35	Xe	I'm not quite sure I understand how to make a structure-like thing in nim
13:49:51	fowl	Type x {.exportc.} = object ..
13:51:23	Xe	does the type have to be exported for the exportc pragma to work?
13:51:28	Xe	er
13:51:37	Xe	type X* vs type X
13:55:30	fowl	aha unless you use the type it doesnt show up
13:55:55	fowl	and when it does show up, its in the module's .c, not some .h where it would be usable
13:57:42	Xe	hmm
14:04:37	*	darkf quit (Quit: Leaving)
14:04:56	*	TEttinger quit (Ping timeout: 244 seconds)
14:09:21	Xe	bah
14:09:23	Xe	wifi
14:10:17	Xe	http://ix.io/ilY -> http://ix.io/im0
14:11:00	*	gyeates quit (Ping timeout: 256 seconds)
14:12:07	fowl	Xe, looks like "object" is not a good module name
14:12:25	Xe	welp
14:12:51	fowl	i believe if you tried to import it you would get a syntax error
14:13:51	Xe	is there a if __name__ == "__main__" in nim?
14:14:10	fowl	there's a special symbol "isMainModule" that is true when the current module is the main module
14:14:16	fowl	when isMainModule:
14:17:03	kokozedman	c2nim documentation is 404: http://nim-lang.org/docs/c2nim.html
14:17:13	*	yglukhov___ quit (Ping timeout: 264 seconds)
14:17:51	*	yglukhov___ joined #nim
14:22:06	Xe	fowl: so I have an object C exported into a .so file, how can I use it?
14:22:32	fowl	Xe, datatypes dont end up in the end product (dll, exe)
14:23:09	Xe	hmm
14:23:25	fowl	can you move it to a header and importc it instead
14:23:47	Xe	can a C struct refer to nim types and vice versa?
14:24:38	fowl	c is less useful than nim
14:25:03	Xe	I'm more thinking about how I would pass data between loaded libraries
14:25:04	fowl	for ex a generic function can't be in the compiled exe
14:25:19	Xe	primitives work fine, but I'd like to be able to pass objects around
14:30:16	Xe	things like hash tables, arrays, etc
14:33:07	*	Ven joined #nim
14:34:51	*	aboisvert joined #nim
14:37:27	*	Jehan_ quit (Quit: Leaving)
14:38:12	*	gsingh93 joined #nim
14:39:34	*	gyeates joined #nim
14:47:58	*	jholland joined #nim
14:49:26	*	pregressive joined #nim
14:49:43	*	milosn joined #nim
14:52:01	*	gsingh93 quit (Ping timeout: 250 seconds)
14:58:37	*	HakanD quit (Quit: Be back later ...)
14:58:49	*	Jehan_ joined #nim
15:00:02	*	HakanD joined #nim
15:04:56	*	HakanD quit (Ping timeout: 240 seconds)
15:08:41	*	gyeates quit (Ping timeout: 265 seconds)
15:13:01	*	HakanD joined #nim
15:14:09	*	vendethiel joined #nim
15:14:35	*	banister joined #nim
15:32:55	*	Matthias247 quit (Read error: Connection reset by peer)
15:34:11	*	ingsoc joined #nim
15:37:25	*	vendethiel quit (Ping timeout: 244 seconds)
15:37:45	*	gsingh93 joined #nim
15:47:48	*	sick joined #nim
15:48:14	sick	Hey all!
15:48:52	def-	Hi sick
15:49:18	sick	I have a table that contains a seq value (Table[string, seq[string]]), but I cant seem to 'add' to the seq
15:49:51	sick	I tried: t["a"] = @[], then t["a"].add "b"
15:50:05	sick	but it complains that the seq is not a var type
15:50:06	def-	t.mget("a").add("b")
15:50:13	sick	ahh
15:50:35	sick	that worked!
15:50:40	def-	this is something we wanted to improve, so that ["a"] works, but there are a few problems with that: https://github.com/Araq/Nim/pull/2435
15:50:41	sick	thanks def-!!
15:51:13	sick	nice I'll take a look
15:53:02	*	vendethiel joined #nim
15:53:33	*	aboisvert quit (Quit: aboisvert)
15:54:28	def-	Basically we wanted to have mget and [] have the same semantics, but then all the code depending on the old semantics breaks
15:54:37	*	HakanD quit (Quit: Be back later ...)
15:55:38	*	Ven quit (Quit: My MacBook has gone to sleep. ZZZzzz…)
15:55:47	*	Varriount|Laptop joined #nim
15:56:08	sick	Yea just finished reading the PR comments
15:56:46	sick	it would certainly look cleaner using []
15:57:12	sick	but I'm pretty impressed with Nim so far :)
15:57:19	*	zahary quit (Read error: Connection reset by peer)
15:57:23	Varriount|Laptop	Meep
15:57:35	Varriount|Laptop	I'm free! All my exams are over!
15:59:59	*	Strikecarl joined #nim
16:00:10	Strikecarl	Easiest way to clear screen in Nim? without spamming Echo ofcourse.
16:00:19	Strikecarl	looking for something like the windows command "cls"
16:00:32	def-	Strikecarl: http://nim-lang.org/docs/terminal.html#eraseScreen,
16:01:24	*	aboisvert joined #nim
16:02:27	Strikecarl	Weird.
16:03:24	Varriount|Laptop	Someone needs to create self-aware documentation
16:06:02	Strikecarl	"execution of an external program failed" ugh
16:06:08	Strikecarl	This basicly sums up my life
16:06:11	Strikecarl	theres an erro.
16:06:15	Strikecarl	error*
16:06:16	def-	--parallelBuild:1
16:06:19	Strikecarl	But it won't tell me what to do.
16:06:22	Strikecarl	what does that do?
16:06:42	def-	it builds 1 file at a time and outputs errors at compilation
16:06:49	Strikecarl	Oh thanks :)
16:06:54	Strikecarl	great
16:06:58	Strikecarl	didn't do anything
16:07:00	def-	but i think you should just compile without "-r"
16:07:02	Strikecarl	"Process terminated with exit code 1"
16:07:12	def-	and run it by hand afterwards
16:07:17	Varriount|Laptop	Strikecarl: What's the console output?
16:07:20	Strikecarl	oh wait
16:07:22	Varriount|Laptop	Like, all of it.
16:07:24	Strikecarl	i am SO FUCKING RETARDED
16:07:26	Strikecarl	lmao
16:07:29	Strikecarl	I had the program open
16:07:34	Varriount|Laptop	-_-
16:07:34	Strikecarl	the .exe it was trying to compile...
16:07:39	Strikecarl	facedesk...
16:10:14	def-	Hm, how can I get a variable into a nim file from the compilation line? os.getEnv seems to work at compiletime, but --os:standalone destroys it
16:11:04	Varriount|Laptop	def-: What kind of variable?
16:11:08	def-	a filename
16:11:14	def-	I'm doing something really weird
16:11:24	Xe	def-: code generation?
16:11:54	Varriount|Laptop	def-: Maybe you could use the process running procedures to grab output from the system shell?
16:12:04	def-	oh right
16:12:07	def-	staticExec
16:12:12	def-	thanks!
16:12:22	*	bluenote quit (Ping timeout: 246 seconds)
16:12:28	Varriount|Laptop	Why does os:standalone not work?
16:13:26	def-	no more posix compatibility for the times module
16:13:28	def-	(or os module)
16:13:42	def-	but I'm os:linux at compiletime, then os:standalone at runtime, just no way to encod ethat
16:14:17	*	vendethiel quit (Ping timeout: 246 seconds)
16:18:50	def-	great, works
16:18:56	def-	Now I can make https://github.com/def-/nim-small-coreutils nicer
16:19:00	*	mal`` quit (Ping timeout: 276 seconds)
16:25:02	*	mal`` joined #nim
16:25:52	BlaXpirit	def-, don't you think you're taking this binary size thing too far?
16:26:40	*	endragor quit (Remote host closed the connection)
16:27:13	Varriount|Laptop	def-: Your tools don't handle all the wierd corner cases of disrupted writes and the other 100 corner cases.
16:27:21	*	Strikecarl quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
16:30:56	*	OnO joined #nim
16:36:43	Varriount|Laptop	def-: I just saw your PR for taking advantage of GCC and Clang's arithmetic with overflow functions
16:37:41	*	vendethiel joined #nim
16:38:19	Varriount|Laptop	You would think that the C standard would have arithmatic overflow operations already...
16:43:59	*	rgv151 joined #nim
16:47:53	*	yglukhov___ quit (Ping timeout: 246 seconds)
16:57:36	*	vikaton joined #nim
16:59:21	*	vendethiel quit (Ping timeout: 265 seconds)
17:03:16	*	brson joined #nim
17:03:20	Jehan_	Varriount|Laptop: About the GC pull request you were asking about on GitHub: Basically, that my program kept crashing with gcc and --opt:speed or --opt:size.
17:03:37	Varriount|Laptop	Jehan_: Ah, ok.
17:11:20	def-	BlaXpirit: sure i am, but it's fun, so who cares?
17:12:17	BlaXpirit	not exactly my definition of fun
17:18:52	Jehan_	BlaXpirit: Well, lots of people also do sudoku puzzles ...
17:20:49	*	drewsrem quit (Quit: Leaving)
17:20:55	*	vendethiel joined #nim
17:24:04	*	brson_ joined #nim
17:26:57	*	brson quit (Ping timeout: 244 seconds)
17:26:57	*	EulersBackup quit (Ping timeout: 244 seconds)
17:33:47	*	brson_ quit (Quit: leaving)
17:33:55	*	brson joined #nim
17:37:25	*	Varriount|Laptop quit (Ping timeout: 246 seconds)
17:42:44	someplace	http://hastebin.com/panazapexo.pas <- the crash again, but with useSysAssert and useGcAssert
17:42:57	someplace	apparently happened just after the bot sent a message
17:43:39	*	vendethiel quit (Ping timeout: 256 seconds)
17:45:54	*	yglukhov___ joined #nim
17:48:59	dom96	someplace: Please report it on Github
17:49:51	someplace	sure
17:50:42	*	gokr_ joined #nim
17:51:03	*	gyeates joined #nim
18:04:10	*	filcuc joined #nim
18:12:58	dalarmmst	Can I do something similar to this in nim? return rexW ? (ext ? R8 : RAX) : (ext ? R8D : EAX)
18:13:15	*	filcuc quit (Quit: Konversation terminated!)
18:17:36	def-	dalarmmst: that's not very good style I'd say, but you can do: return if rexW: (if ext: R8 else: RAX) else: (if ext: R8D else: EAX)
18:18:08	def-	ifs can be used as expressions in Nim
18:18:19	*	grncdr joined #nim
18:19:04	dalarmmst	def-: Thanks
18:19:28	dalarmmst	What would be a better style you think?
18:19:42	def-	want to post your full proc somewhere?
18:22:05	*	grncdr quit (Client Quit)
18:22:32	dalarmmst	def-: http://pastebin.com/eV21VPcH I haven't written a proc yet, trying to convert my original C++ method
18:24:54	*	jryan joined #nim
18:25:29	jryan	Hey guys. I'm trying to cross compile to windows from linux using mingw-w64 on arch linux.
18:25:41	jryan	And getting nimcache/system.o:system.c:(.text+0x1971b): undefined reference to `memcpy'
18:25:42	*	nande joined #nim
18:25:51	jryan	Anything jump out at anyone?
18:26:02	def-	jryan: can you cross compile C code?
18:26:10	def-	if that works, Nim should also work
18:27:23	*	dalarmmst quit (Ping timeout: 250 seconds)
18:27:49	kokozedman	hey guys, I'm trying to use a C libraries that use ... as parameter; is it possible to use it in nim with c2nim? if so, then how do I do it?
18:27:58	def-	kokozedman: {.varargs.}
18:28:11	kokozedman	oh ok... didn't know about that
18:28:14	kokozedman	thanks def-
18:28:29	def-	http://nim-lang.org/manual.html#varargs-pragma
18:29:12	jryan	def-: Yes I can
18:29:44	def-	jryan: you can use --verbosity:3 --parallelBuild:1 to see the exact gcc commands used
18:29:48	def-	maybe you'll notice what's wrong
18:30:00	*	filcuc joined #nim
18:30:13	jryan	def-: Yeah I did that, and I didn't see anything weird during the linking step.
18:30:34	def-	so you can use memcpy in C programs linked with the same command?
18:30:40	def-	maybe the include is missing?
18:31:12	*	gyeates quit (Ping timeout: 265 seconds)
18:31:34	jryan	def-: https://gist.github.com/jjryan/99a98d18b4222769a396
18:31:41	jryan	Is what it's outputting as the linker command.
18:32:23	def-	looks pretty normal
18:32:30	def-	but then I never cross-compiled for Windows
18:33:39	*	koalazen joined #nim
18:33:41	*	vendethiel joined #nim
18:34:07	jryan	Seems to be acting differently now.
18:34:13	jryan	Not sure why... /usr/bin/x86_64-w64-mingw32-ld: cannot find -ldl
18:34:18	koalazen	Hi, where am I supposed to add tests for a change in the standard library ? In my case it's math.nim
18:35:17	def-	koalazen: maybe tests/floats or tests/stdlib/tmath.nim
18:35:55	*	grncdr joined #nim
18:36:28	def-	jryan: you do set --os:windows, right?
18:36:32	*	nande quit (Remote host closed the connection)
18:36:50	*	fizzbooze joined #nim
18:37:19	jryan	def-: nim c -d:windows --cpu:amd64 --verbosity:3 --parallelBuild:1 frontcut.nim
18:37:46	def-	that should be --os:windows
18:38:02	koalazen	thanks def-
18:38:06	jryan	sorry that was an old command def-
18:38:12	jryan	nim c --os:windows --cpu:amd64 --verbosity:3 --parallelBuild:1 frontcut.nim
18:38:22	kokozedman	what's the Nim way of using the offsetof () C macro to get the offset of an object member? I suppose Nim object types are converted to struct in C... I'm trying to use a third-party C library that takes the offset of a struct member
18:39:42	*	dalarmmst joined #nim
18:39:50	*	Demon_Fox joined #nim
18:39:58	def-	dalarmmst: hm, don't see a better way to do that
18:40:15	def-	dalarmmst: except if you fill them into a table maybe
18:40:21	*	OnO quit (Quit: Textual IRC Client: www.textualapp.com)
18:40:36	dalarmmst	def-: Ok, thanks for looking at it, and thanks for the nim help
18:44:23	def-	kokozedman: you want to use the macro or reimplement it in Nim?
18:44:49	kokozedman	def-, just the macro if I can... whichever is "efficient"
18:45:26	kokozedman	or, whichever works with the 3rd party lib
18:46:22	*	filcuc quit (Quit: Konversation terminated!)
18:46:31	*	flaviu quit (Remote host closed the connection)
18:47:15	*	Mat4 joined #nim
18:47:21	Mat4	hello
18:48:04	*	nande joined #nim
18:53:29	def-	kokozedman: does this work? https://gist.github.com/def-/2ed28c3be686fb02fa2d
18:53:47	*	loz1 joined #nim
18:56:03	kokozedman	def-, man! that looks really nice! Can't test right now because I'm just writing a bunch of wrapper, but I think it should be already. Thanks for sharing :)
18:56:19	*	filcuc joined #nim
18:56:56	*	vendethiel quit (Ping timeout: 276 seconds)
19:00:16	jryan	def-: If I compile them all using gcc, and have gcc link it for me, it works.
19:00:39	jryan	So no -c just *.c and -o
19:00:49	def-	jryan: so if you do it by hand it's fine? but when nim calls the same commands it doesn't work?
19:02:11	*	grncdr quit (Quit: Leaving.)
19:04:36	RegisterOn	nim has a problem
19:04:55	RegisterOn	it's a little bit too awesome
19:05:17	*	grncdr joined #nim
19:05:34	def-	dalarmmst: i'm curious, what are you working on?
19:09:28	*	koalazen quit (Ping timeout: 246 seconds)
19:12:28	*	filcuc quit (Remote host closed the connection)
19:15:30	*	ingsoc quit (Quit: Leaving.)
19:19:08	*	kokozedman left #nim ("Leaving")
19:22:18	jryan	def-: When I do it by hand it works yes. But I don't manually call the linker, I have gcc do that for me.
19:22:24	jryan	def-: Which is why it's working
19:22:34	jryan	Just tested the exe and it works too.
19:22:52	def-	jryan: then I guess it would help to find out how gcc calls the linker and tell nim the same
19:23:12	def-	wait, shouldn't gcc use gcc as the linker?
19:23:13	jryan	Yeah. I don't really know how to figure how what the heck it's doing.
19:23:14	*	filcuc joined #nim
19:23:31	def-	the linkerexe should be gcc, not gcc-ld
19:23:35	jryan	def-: oh really
19:23:37	def-	yes
19:23:49	def-	see in config/nim.cfg
19:24:03	def-	under "example of how to setup a cross-compiler"
19:24:11	jryan	oh my god i'm dumb
19:24:28	jryan	I should of copy pasted I guess hahah
19:24:32	jryan	instead of guessing
19:24:47	jryan	that totally fixes it
19:24:50	def-	great
19:25:18	jryan	I haven't been in C land for a long time, so I was confused why vanilla LD wasn't working. Makes sense now.
19:25:34	jryan	There's tons of default crap being passed through there.
19:28:38	jryan	def-: thank you very much
19:28:42	*	skyfex_ quit (Quit: (null))
19:30:15	*	skyfex__ joined #nim
19:33:10	*	Mat4 is now known as Mat4-coding
19:33:59	*	vikaton quit ()
19:36:17	*	jryan quit (Quit: Page closed)
19:37:02	*	fizzbooze quit (Ping timeout: 252 seconds)
19:38:32	*	skyfex__ quit (Ping timeout: 256 seconds)
19:46:36	*	vikaton joined #nim
19:46:37	Xaseron	my parallel build fails. is my code wrong or did i found a bug? https://paste.xinu.at/Sds/nimrod
19:48:30	*	Jesin quit (Quit: Leaving)
19:49:02	RegisterOn	Xaseron: it fails in what way?
19:50:00	Xaseron	/home/xaseron/nim/qsort/nimcache/qsort.c:264:36: error: ‘HEX3Atmp_146207Len0’ undeclared (first use in this function)
19:50:02	Xaseron	quicksort_146085(HEX3Atmp_146207, HEX3Atmp_146207Len0, HEX3Atmp_146210, HEX3Atmp_146213);
19:50:18	Xaseron	gcc 4.9.2
19:54:34	*	xificurC quit (Remote host closed the connection)
19:54:47	*	MyMind quit (Ping timeout: 246 seconds)
19:55:09	*	xificurC joined #nim
19:56:43	*	ingsoc joined #nim
19:57:15	RegisterOn	Xaseron: same under clang. The second-last line is the problem, compiles without it
19:57:46	*	Jesin joined #nim
19:57:58	def-	Xaseron: we have a few bug reports that have very similar sounding errors
19:58:19	def-	it works with seq instead of openarray
19:59:04	def-	for example this one: https://github.com/Araq/Nim/issues/1834
19:59:17	RegisterOn	yeah, changing to seq works
19:59:29	def-	I guess you should report it
19:59:38	RegisterOn	index error at runtime, but compiles
19:59:50	def-	for me it runs fine
19:59:55	*	OnO joined #nim
20:00:21	RegisterOn	def-: clang?
20:00:24	def-	gcc
20:00:31	RegisterOn	weird
20:00:36	def-	clang fine to
20:00:40	def-	too*
20:00:42	RegisterOn	all i did was take his code and change to seq
20:00:47	RegisterOn	anything else?
20:01:03	def-	i changed both instances of openarray to seq
20:03:26	*	vendethiel joined #nim
20:03:37	Xaseron	works for me with gcc and clang
20:04:19	*	Matthias247 joined #nim
20:04:20	RegisterOn	well, probably because i'm on 0.10.2 because brew isn't updated
20:04:36	def-	Right, 0.10.2 is probably broken
20:04:47	RegisterOn	nope, it works some of time
20:04:51	RegisterOn	some race condition?
20:05:02	RegisterOn	run it lots of times, breaks some of the time
20:05:12	RegisterOn	Error: unhandled exception: index out of bounds [IndexError]
20:05:14	RegisterOn	@[-31, 0, 2, 2, 4, 65, 83, 99, 782]
20:05:29	*	fizzbooze joined #nim
20:05:55	RegisterOn	seems completely random
20:06:23	RegisterOn	SIGBUS: Illegal storage access. (Try to compile with -d:useSysAssert -d:useGcAssert for details.)
20:06:26	RegisterOn	some of time...
20:09:48	loz1	hi guys, i'm trying to swap two ref's, and i think this doesn't work:
20:09:48	loz1	let tmp = field
20:09:48	loz1	field = field2
20:09:51	loz1	field2 = tmp
20:09:59	loz1	am i right?
20:10:31	RegisterOn	Xaseron: does it fail sometimes if you run many times quickly?
20:11:01	Xaseron	RegisterOn: nope
20:11:11	RegisterOn	weird
20:11:35	Xaseron	hmm
20:11:45	RegisterOn	maybe if you increase the array size so it gets more work
20:11:50	RegisterOn	seems like a race somewhere
20:11:52	Xaseron	sorting 1000 elements takes forever in the debug build
20:12:22	BlaXpirit	loz1, u can probably use `swap field, field2`
20:12:49	BlaXpirit	but your code should work too
20:12:52	def-	loz1: swap their contents? then swap field[], field2[]
20:13:04	BlaXpirit	right..
20:13:20	Xaseron	RegisterOn: looks like a deadlock to me
20:13:21	loz1	def-: good question, the type of field is Field which is ref FieldObj
20:13:40	loz1	def-: and i want to swap pointers
20:13:40	def-	loz1: if you swap them like that, you just swap their addresses, not the contents
20:13:47	def-	oh, then that should be fine
20:13:59	*	gokr_ quit (Ping timeout: 245 seconds)
20:13:59	loz1	how can i print the addresses?
20:14:06	BlaXpirit	maybe repr
20:14:13	BlaXpirit	not sure though
20:14:24	BlaXpirit	where's Mimbus? :(
20:14:46	def-	cast[ByteAddress](field) if repr doesn'T like it
20:15:11	Xaseron	RegisterOn: SIGSEGV: Illegal storage access. (Attempt to read from nil?)
20:15:41	RegisterOn	right
20:15:59	RegisterOn	parallel seems rather buggy
20:16:15	Xaseron	qsort: ../nptl/pthread_mutex_lock.c:81: __pthread_mutex_lock: Assertion `mutex->__data.__owner == 0' failed.
20:16:17	Xaseron	SIGABRT: Abnormal termination.
20:16:21	BlaXpirit	D:
20:16:30	RegisterOn	and IndexError from time to time
20:16:37	RegisterOn	all sorts of random failures
20:16:53	*	gokr_ joined #nim
20:17:02	def-	RegisterOn: you really shouldn't test with 0.10.2, it's known to be buggy
20:17:15	def-	Xaseron: what system are you on, I can't reproduce any crashes
20:17:31	RegisterOn	def-: brew has 0.10.2 so that's what I am using
20:17:56	Xaseron	arch linux or what do you mean by system?
20:18:08	def-	Xaseron: hm, but it seems to deadlock
20:18:21	RegisterOn	Xaseron: i'm on OSX
20:18:34	RegisterOn	def-: ^
20:18:50	def-	I have to say i've never used parallel stuff in Nim
20:19:17	RegisterOn	it's the first thing that meet users on the home page, so it better work :)
20:19:32	Xaseron	^^
20:19:58	*	fizzbooze quit (Ping timeout: 272 seconds)
20:20:04	def-	I agree, no idea why it's there
20:20:35	RegisterOn	concurrency tends to be a study in why it doesn't work, and Nim is no different :) I just spawn tons of processes and communicate via queues, Erlang style
20:20:49	RegisterOn	chicken-shit approach, but boy, does it always work
20:20:50	*	HakanD joined #nim
20:22:24	loz1	oh ok, if i remove parallel stuff everything works
20:22:33	*	loz1 sad
20:25:26	*	vendethiel quit (Ping timeout: 252 seconds)
20:25:35	def-	Xaseron: I guess a bug report would be good
20:25:43	dalarmmst	def-: I'm building a disassembler just for fun
20:26:02	def-	dalarmmst: that sounds fun!
20:26:46	Xaseron	even if i limit the thread pool size to 1 it deadlocks
20:27:14	Xaseron	def-: about the parallel deadlock or the compile error with openarray?
20:27:15	dalarmmst	def-: Yeah it is :) I'm working as a compiler engineer so it helps me in my work as well
20:27:17	Xaseron	or both?
20:27:27	def-	Xaseron: both of course
20:28:40	RegisterOn	"both, please" increases the chance of it actually getting done
20:34:02	*	Jesin quit (Quit: Leaving)
20:46:01	*	rgv151 quit (Remote host closed the connection)
20:53:13	*	elbow_jason quit (Read error: Connection reset by peer)
20:53:23	*	elbow_jason joined #nim
21:01:21	*	pregressive quit (Remote host closed the connection)
21:09:01	OnO	Araq: sorry, but we need some quick fixup for csources submodule: https://github.com/Araq/Nim/pull/2675
21:11:24	*	ingsoc quit (Quit: Leaving.)
21:17:17	*	Jesin joined #nim
21:18:43	*	Kingsquee joined #nim
21:20:24	*	Mat4-coding left #nim ("http://quassel-irc.org - Chat comfortably. Anywhere.")
21:20:39	*	fizzbooze joined #nim
21:21:29	*	gokr_ quit (Ping timeout: 245 seconds)
21:25:19	*	gokr_ joined #nim
21:25:48	*	gsingh93 quit (Quit: WeeChat 1.1.1)
21:37:45	*	gyeates joined #nim
21:42:21	*	pregressive joined #nim
21:42:21	*	pregressive quit (Remote host closed the connection)
21:43:03	*	pregressive joined #nim
21:43:50	*	yglukhov___ quit (Quit: Be back later ...)
21:44:00	*	Trustable quit (Remote host closed the connection)
21:44:13	*	fizzbooze quit (Ping timeout: 255 seconds)
21:47:37	*	Mat4-coding joined #nim
21:47:44	*	Mat4-coding left #nim ("http://quassel-irc.org - Chat comfortably. Anywhere.")
21:47:52	*	coffeepot joined #nim
21:49:40	*	gyeates quit (Ping timeout: 256 seconds)
21:51:17	*	loz1 quit (Ping timeout: 256 seconds)
21:52:58	reactormonk	OnO, I just hope it works
21:53:23	OnO	test, see for youself
21:53:48	OnO	thnx
21:55:48	OnO	still would be nice to have *.o ignored in csources repo
21:55:53	OnO	*.obj too
21:56:34	OnO	see https://github.com/nim-lang/csources/pull/16
21:56:53	OnO	otherwise after build csources are shown by Git as dirty
21:58:30	gokr	def-: Don't we consider the lambda syntax in future to "be lambdas"?
22:00:08	def-	gokr: i'm not sure about terminology. i guess all anonymous procs are lambdas, but I expect them to be a bit nicer than they currently are in Nim to use them
22:00:24	gokr	But you tried the ones in future, right?
22:00:48	gokr	The new restriction that all params must be typed does make them slightly verbose again though.
22:02:32	def-	and that they don't always work
22:03:09	Jehan_	I'd honestly consider making them part of the compiler proper rather than macros.
22:03:14	reactormonk	lambdas are procs with {.closure.} or {.procvar.} from what I understand
22:03:20	*	vendethiel joined #nim
22:04:06	def-	gokr: https://github.com/Araq/Nim/issues/2179
22:04:19	Jehan_	reactormonk: They're just syntactic sugar for the proc(...): auto = ... notation.
22:04:44	gokr	def-: But... doesn't that already work?
22:04:55	def-	gokr: I wish
22:05:37	gokr	I think I did several lambdas like that in this article: http://goran.krampe.se/2014/12/03/nim-seq/
22:05:39	def-	we're rather going in the opposite direction, argument types are now mandatory
22:05:51	gokr	yes, that is a bit annoying I agree.
22:07:34	gokr	This works just fine: echo(@[1,2,3].map((x:int) => x * 2))
22:08:09	gokr	So I am unsure what 2179 means.
22:08:40	def-	it means I'd prefer @[1,2,3].map(x => x*2)
22:09:10	gokr	So x instead of (x:int) ?
22:09:15	Jehan_	I haven't looked at it, but type inference can become a bit tricky when you omit arguments. Especially when you have overloading or subtyping.
22:09:22	def-	yes, and no return type and for it to work more consistently
22:09:36	gokr	def-: I didn't use a return type
22:09:42	Jehan_	def-: Last I tried, you could omit the return type?
22:09:49	def-	gokr: not in this case, but in some it fails to compile
22:10:08	gokr	I agree that it seems to be a bit... varying.
22:10:17	gokr	I think my comments in my article also notes this.
22:10:22	Jehan_	I have to admit that I don't use the future stuff a whole lot since I consider it unstable (not in the sense of it breaking, but there being no guarantee that it'll remain part of the language).
22:11:37	gokr	I think getting lambdas working is an important part. And... I can buy that Araq wanted to get rid of the untyped parameters in general since people thought it was type inferencing - when in fact it was generics.
22:11:58	gokr	But for lambdas, it would be nice to get it working without types.
22:12:23	def-	hat we have all these procs ending with It in sequtils shows that lambdas don't work that well yet
22:12:43	gokr	def-: I think its fairly old code though.
22:12:58	gokr	And they aren't procs are they?
22:13:15	gokr	They are templates.
22:13:30	Jehan_	def-: The advantage of mapIt() and such is that they provide better performance.
22:14:04	gokr	Right
22:14:34	Jehan_	You could get the same performance out of lambdas, but that would require the compiler to do the necessary inlining analysis.
22:15:51	Jehan_	Which is not entirely trivial and part of the reason why Scala's for statement tends to have poor performance.
22:17:33	Jehan_	(Scala's for is basically syntactic sugar for a composition of foreach, map, flatMap, and filter operations working on lambdas.)
22:17:36	gokr	At least the precedence of => is tweaked now so that you can avoid the ( ) around the body.
22:18:24	Jehan_	That means even for (i <- 1 to 100) {s += a(i)} turns the `s += a(i)` part into a closure that the JVM then has to worry about inlining.
22:19:56	def-	gokr: Ah, here was the example which I tried with lambdas instead of *It: https://github.com/Araq/Nim/pull/2175
22:20:05	def-	and you needed the return type there
22:23:43	*	pregressive quit ()
22:25:43	reactormonk	anyone know where the compiler resolves symbols?
22:25:51	*	vendethiel quit (Ping timeout: 256 seconds)
22:27:59	gokr	def-: I seemed to get away without return types in that code.
22:28:13	gokr	This works: us.filter((x:User) => x.active).sortedByIt(it.id).map((x:User) => x.lastName)
22:28:25	gokr	But there is no sortedBy :)
22:29:30	def-	it's just called sorted now
22:29:48	gokr	Ehm... no...
22:29:53	gokr	That takes a compare proc
22:30:19	gokr	I was thinking something like sortedBy((x:User) => x.id)
22:33:06	fowl	reactormonk: sem*
22:33:26	*	vikaton quit (Quit: Connection closed for inactivity)
22:33:56	*	dtscode quit (Remote host closed the connection)
22:40:46	*	bjz quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)
22:42:37	sick	does anyone know why spawn keeps complaining about the passed in proc not returning a value?
22:43:13	*	filcuc quit (Quit: Konversation terminated!)
22:45:05	sick	using the example in the guide (http://nim-lang.org/docs/manual.html#parallel-spawn-spawn-statement), gives me this error: spawn.nim(8, 20) Error: expression 'processLine(x)' has no type (or is ambiguous)
22:45:25	sick	no clue what to do :(
22:45:59	ldlework	sick: don't worry someone will know what the problem is
22:47:33	reactormonk	fowl, ok
22:47:47	reactormonk	fowl, I'm looking for global variables.
22:49:27	*	grncdr quit (Quit: Leaving.)
22:52:05	Jehan_	An alternative way to do filter and map with macros: https://gist.github.com/rbehrends/c7127c8e61344bb1dadc
22:52:42	def-	Jehan_: nice
22:52:43	Jehan_	Wondering if this can be regularized without having all the boilerplate all the time.
22:53:46	reactormonk	Jehan_, neat. do you get useful error messages?
22:54:14	reactormonk	with types etc.
22:54:17	*	OnO quit (Quit: My iMac has gone to sleep. ZZZzzz…)
22:54:42	Jehan_	reactormonk: No idea. Just a quick hack.
22:55:05	Jehan_	Macro-related error messages can be anywhere from very good to "what the hell does that mean"?
22:55:20	Jehan_	Unavoidably, really, since the error reporting mechanism has only so much context to work with.
22:56:57	*	coffeepot quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
23:02:02	*	saml_ joined #nim
23:02:12	*	aboisvert quit (Quit: aboisvert)
23:09:32	*	Jehan_ quit (Quit: Leaving)
23:14:37	*	vendethiel joined #nim
23:18:47	*	HakanD quit (Quit: Be back later ...)
23:19:35	*	Matthias247 quit (Read error: Connection reset by peer)
23:21:27	reactormonk	fowl,
23:21:30	reactormonk	ehh
23:21:41	reactormonk	semVarOrLet is executed once per var/let statement group?
23:22:46	fowl	I'm not a good person to ask about the compiler >.>
23:23:21	reactormonk	>:)
23:24:11	reactormonk	I'm still chasing my double-global bug... suggestionResultHook_335823: ["sfExported", "sfGlobal"], [], ["compiler/suggest.nim", 32, 2], skVar and suggestionResultHook_456030: ["sfExported", "sfGlobal"], [], ["compiler/suggest.nim", 32, 2], skVar
23:24:14	reactormonk	should be the same
23:24:48	reactormonk	one of them gets added to the sigmatch module... *scratches head*
23:24:57	reactormonk	I don't even mention that one in that module
23:30:14	reactormonk	-.-
23:30:23	reactormonk	there's an include somewhere in the compiler that screwed me up.
23:35:02	reactormonk	I guess I should read some more... # included from sigmatch.nim
23:35:59	*	BlaXpirit_ joined #nim
23:35:59	*	BlaXpirit is now known as Guest66546
23:35:59	*	Guest66546 quit (Killed (orwell.freenode.net (Nickname regained by services)))
23:35:59	*	BlaXpirit_ is now known as BlaXpirit
23:37:10	*	vendethiel quit (Ping timeout: 255 seconds)
23:37:11	*	Siecje joined #nim
23:37:15	*	Siecje quit (Max SendQ exceeded)
23:38:04	*	Siecje joined #nim
23:38:29	*	Siecje quit (Max SendQ exceeded)
23:39:14	filwit	gokr: any thoughts on how I should modify 'marshal' to have hooks? Marshal doesn't quite work for me, so instead of just remaking it I'll contribute back, but I've barely used these things in other languages, so I don't know 'best practice' here. My idea is to have basically this syntax: foo.serialize(shallow=false, hooks={"int":handleInt, "string":handleString})
23:39:19	*	Siecje joined #nim
23:40:23	filwit	where 'handleInt' is a `proc(i:int): string`, etc
23:40:51	filwit	err, make that `proc handleInt(i:int): JObject`
23:41:03	filwit	anyone have opinions?
23:41:10	gokr	I have only skimmed that code
23:42:29	filwit	okay, it just sounded like you where familiar with other serialization libs, and you might have specific ideas on how best to improve the API
23:42:58	fowl	Does marshal even work with variant objects
23:43:25	filwit	idk..
23:43:31	*	sick quit (Ping timeout: 246 seconds)
23:43:41	filwit	but I don't need it for variant objects right now, so not going to worry about that
23:44:53	reactormonk	pigmej, ok, got epc running with nim. Wanna hook it into nim-mode?
23:45:45	filwit	fowl: wait, i misunderstood what you meant by that.. I'm not sure how marshal handles case-objects.. one sec, let me test
23:45:56	filwit	cause I do want to support those
23:49:18	filwit	fowl: yeah it handles variant objects fine
23:51:33	fowl	Ah ok