<< 05-01-2014 >>

---

00:17:53	*	io2 quit ()
00:21:11	*	Demos_ joined #nimrod
00:27:06	*	BitPuffin joined #nimrod
00:27:18	BitPuffin	dom96!!!!
00:27:21	BitPuffin	dom96???
00:27:28	BitPuffin	dom96!?!?!?!?!
00:37:28	Demos_	wow divide by zero error in the compile, well done microsoft
00:38:47	Demos_	actually this could be real invalid code
01:08:47	Demos_	anyone know a way to generate a pdb file without defeating microsoft's compiler?
01:12:01	Demos_	Varriount, I just tried building devel on windows, got the same range error as you did
01:47:16	*	shodan45_ quit (Read error: Connection reset by peer)
01:47:26	*	shodan45_ joined #nimrod
01:50:14	*	shodan45 joined #nimrod
01:50:14	*	shodan45_ quit (Read error: Connection reset by peer)
01:56:32	*	shodan45 quit (Quit: Konversation terminated!)
01:56:46	*	shodan45 joined #nimrod
02:01:31	*	shodan45 quit (Client Quit)
02:01:46	*	shodan45 joined #nimrod
02:03:33	*	shodan45 quit (Client Quit)
02:03:41	*	shodan45 joined #nimrod
02:09:41	*	shodan45 quit (Read error: Connection reset by peer)
02:09:49	*	shodan45 joined #nimrod
02:17:37	*	shodan45 quit (Read error: Connection reset by peer)
02:17:45	*	shodan45 joined #nimrod
02:26:05	*	shodan45 quit (Quit: Konversation terminated!)
02:26:10	*	shodan45_ joined #nimrod
02:28:53	*	shodan45_ quit (Read error: Connection reset by peer)
02:29:12	*	shodan45_ joined #nimrod
02:38:12	*	shodan45_ quit (Quit: Konversation terminated!)
02:38:23	*	shodan45_ joined #nimrod
02:39:42	*	Varriount|Mobile quit (Quit: AndroIRC - Android IRC Client ( http://www.androirc.com ))
02:52:37	*	brson quit (Ping timeout: 272 seconds)
02:54:16	*	brson joined #nimrod
03:15:02	*	PortablePuffin quit (Read error: Connection reset by peer)
03:15:32	*	shodan45_ quit (Quit: Konversation terminated!)
03:15:40	*	shodan45 joined #nimrod
03:18:30	*	PortablePuffin joined #nimrod
03:20:21	*	shodan45 quit (Client Quit)
03:20:31	*	shodan45 joined #nimrod
03:50:32	*	brson quit (Ping timeout: 260 seconds)
04:20:27	*	DvorakDude joined #nimrod
04:20:33	*	DvorakDude left #nimrod (#nimrod)
04:35:47	*	PortablePuffin quit (Ping timeout: 245 seconds)
04:38:45	*	BitPuffin quit (Ping timeout: 272 seconds)
04:40:21	*	PortablePuffin joined #nimrod
04:46:30	*	shodan45 quit (Quit: Konversation terminated!)
04:56:56	Demos_	if I have some type TThing and I have a ptr TThing and some functions that take a ref TThings is there anything I can do to call those functions with my TThings?
05:01:23	*	psquid joined #nimrod
05:09:42	wan	An empty subscript [] notation can be used to derefer a reference
05:10:09	Demos_	yes, I know
05:10:20	Demos_	but it can not go from a ptr to a ref
05:10:29	wan	oh, ref
05:10:48	wan	overload?
05:11:08	Demos_	I can't change the functions in question
05:11:30	Demos_	I can reimplement them, but that is less than ideal
05:22:37	wan	var my_little_ref = cast[ref TThing](my_little_pointer)
05:22:58	wan	It should do the trick
05:23:41	*	Mordecai joined #nimrod
05:23:49	Demos_	yeah, except then the gc would try and free it when the function returned
05:24:13	wan	Would it? It belongs to the upper scope.
05:24:21	Demos_	the ptr does
05:24:31	Demos_	no it would free it when the caller returned
05:25:27	Demos_	I was just using refs for everything but stopped as I needed more control, I think I may have been loosing track of a ref
05:26:00	Demos_	I am passing these pointer to c# and letting the .net gc call a free function in the finalizers
05:26:14	wan	Yeah, but refs in args mean you 'borrow' from the upper scope so that you don't have to copy a whole TThing, no ? I don't think the function you call with a ref would free that ref.
05:27:15	*	psquid quit (Ping timeout: 272 seconds)
05:30:17	Demos_	no, but once the calling function returned that ref WOULD be free. And the nimrod compiler can decide to take things by pointer if you just pass the value type
05:30:44	Demos_	one of the resaons why proc(a: int) = a + 1 is invalid code
05:39:21	fowl	wan, ref in nimrod is a garbage-collected pointer, pass-by-ref is called "var"
05:41:47	wan	Is there a difference for the function to call?
05:43:27	*	[1]Endy joined #nimrod
05:44:00	*	Demos__ joined #nimrod
05:44:28	wan	Does nimrod then triggers a registration with the GC if you cast a ptr as a ref and then pass it around?
05:44:52	*	ddl_smurf quit (Quit: ddl_smurf)
05:47:31	*	Demos_ quit (Ping timeout: 272 seconds)
05:54:14	Demos__	it may, and if it does the bug could be quite hard to detect
05:54:44	Demos__	and fowl isn't nimrod's "pass by val" actually pass by const val or by const ref
06:03:18	fowl	maybe
06:04:23	Demos__	I think the compiler has that freedom, but I have no idea if it actually uses it
06:32:08	*	girvo joined #nimrod
06:46:50	*	girvo quit (Quit: My iMac has gone to sleep. ZZZzzz…)
06:57:43	*	girvo joined #nimrod
07:01:53	*	girvo quit (Client Quit)
07:08:12	*	girvo joined #nimrod
07:30:19	*	[2]Endy joined #nimrod
07:33:50	*	[1]Endy quit (Ping timeout: 264 seconds)
07:55:07	*	girvo quit (Quit: My iMac has gone to sleep. ZZZzzz…)
08:54:21	*	Demos__ quit (Ping timeout: 272 seconds)
09:11:46	*	Demos__ joined #nimrod
09:35:36	zahary	Varriount: my faulty commit looks rather innocent, but I'll try booting on windows today
09:38:12	*	Demos__ quit (Ping timeout: 245 seconds)
09:57:59	zahary	Araq, you requested the not types when I first implemented tfAll and tfAny - I have the IRC logs to prove it :P
09:59:09	zahary	anyway, I added them for completeness, because they were quite easy to add to the current infrastructure - maybe someone could make use of types like "not Serializable"
10:11:59	*	zielmicha1 joined #nimrod
10:14:36	*	zielmicha quit (Ping timeout: 260 seconds)
10:15:03	*	zielmicha8 quit (Ping timeout: 272 seconds)
10:17:57	*	exhu joined #nimrod
10:18:22	exhu	hi
10:19:10	exhu	who is in charge of https://github.com/nimrod-code/cairo/issues/1 ? Please commit this line.
10:20:00	*	gradha joined #nimrod
10:20:14	exhu	and maybe someone can help with http://forum.nimrod-lang.org/t/333
10:33:51	*	fntzr joined #nimrod
10:34:27	exhu	hi Araq
10:35:08	gradha	exhu: replied http://forum.nimrod-lang.org/t/333/1#1734
10:36:00	Araq	hi exhu, gradha's answer is correct
10:36:10	exhu	Thanks guys!
10:36:12	Araq	hi fntzr welcome
10:37:08	Araq	exhu: I think the cario wrapper gets rid of the cairo_ prefix
10:37:19	Araq	and maybe also of the surface_
10:37:43	Araq	so, make a consistent pull request please
10:37:55	exhu	Araq, too bad, it makes harder to refer to the original cairo docs/tutorials
10:38:15	Araq	not really
10:38:47	Araq	you simply say somewhere "nimrod has modules and as such namespaces and so we got rid of primitive prefixes"
10:39:05	Araq	and the cario docs even suggest this for wrappers iirc
10:39:50	exhu	Araq, do only module names get prefixed in binary form? i.e. importing private/mymodule and lib/mymodule can cause name collisions?
10:40:25	Araq	yeah
10:40:28	Araq	well
10:40:35	Araq	the rules are rather adhoc
10:40:46	Araq	the compiler tries to guess the "package" name
10:41:01	Araq	but there is no notion of package in the compiler/language
10:41:48	exhu	so it's very important to have module names unique
10:42:01	Araq	it uses the parent directory unless that's some crap like src, private, etc. :P
10:42:44	Araq	no it's not very important to have unique module names, it's important to have a sensible directory structure
10:43:16	exhu	Araq, how can i declare an alias for a type, i.e. like typedef in C works?
10:43:29	Araq	type Alias = SomeType
10:43:52	Araq	type sections give us that feature for free
10:44:39	gradha	exhu: your own forum example uses a type alias
10:44:44	exhu	ah, remembered the distinct keyword.
10:44:54	exhu	it was int32 and int that are distinct
10:45:17	exhu	that's why i got confused -)
10:50:26	exhu	i feel lack of package namespace and visibility rule, i.e. i want a private api between package modules to see each other and have a public api which does not include low-level stuff
10:51:13	exhu	an asterisk makes package-private functions and types visible to package users
10:51:44	gradha	wrap your group of interconnected modules through a public single module
10:52:13	exhu	gradha, you mean via 'include' ?
10:52:39	gradha	if you have modules P1, P2, P3, create API and import P1, P2, P3 from API
10:53:01	gradha	people importing API will see only what you want, though I guess they can still import manually Px
10:53:52	gradha	but IMHO people concerned about visibility of procs in modules are lousy documentation writers, programmers shouldn't dig into modules source code to figure out stuff
10:54:14	gradha	your documentation should be the API, not a .nim file
10:54:16	exhu	can i simply declare an alias for a proc? like API.sompublicproc = myrpivatemodule.MyProc ?
10:55:16	Araq	no use 'export' instead
10:55:18	gradha	never tried that, but you can export the sumbols you want
10:56:16	Araq	and btw nimrod's module system has already more features than most if not all other languages out there
10:56:37	Araq	so it's really annoying people still complain about it
10:58:25	Araq	the package stuff is about to distinguish jester/utils from nimgame/utils, not to distinguish private/utils from public/utils
10:58:38	Araq	use different module names in the same package for fuck's sake
10:58:46	Araq	how hard can it be?
10:59:16	Araq	</rant>
10:59:27	*	gradha ponders creating a foo babel package depending on bar
10:59:28	exhu	Araq, i don't mean using same package names in a single package, i mean collisions in different packages
11:01:57	exhu	Araq, like generating two object files with the same public symbols, e.g. module jester/utils exports "someUtilProc" and module nimgame/utils exports "someUtilProc" and then they get linked...
11:07:46	Araq	that's not a problem
11:09:21	Araq	how many 'add' procs do you think we have? ;-) we have a good enough name mangling to deal with this problem
11:10:01	fntzr	Araq, hi )
11:24:15	Araq	bbl
11:32:06	*	girvo joined #nimrod
11:40:31	*	Mordecai quit (Read error: Operation timed out)
11:41:16	*	Mordecai joined #nimrod
11:48:32	*	girvo quit (Quit: Textual IRC Client: www.textualapp.com)
11:58:56	*	exhu quit (Quit: Ex-Chat)
12:30:50	*	darkf quit (Quit: Leaving)
12:37:42	Varriount	gradha, I must thank you for informing me about the existance of the 'using' statement via the recent issue you posted.
12:37:56	Araq	hey, the using statement is new
12:38:07	Araq	and I'm not sure we should keep it :P
12:38:46	gradha	kill it with fire
12:38:57	Varriount	I think it'd be better in block form
12:40:34	gradha	block: using x
12:40:45	Varriount	Of course, I'm neutral as to whether it should stay or go. I haven't had much call to 'use' it yet.
12:45:31	gradha	Varriount: https://gist.github.com/gradha/8267759
12:47:00	dom96	OrionPK: Why not just use nginx's gzip capabilities?
12:47:18	dom96	I think adding gzip to httpserver would be appropriate, but i'm not sure.
12:47:27	gradha	Varriount: I just updated the gist, if you modify it to use the statement keyword instead the compiler complains with a fun message
12:47:39	gradha	expression '(s|a)' has no type (or is ambiguous)
12:47:58	gradha	I was going to add that pitfall to the docs soon
12:48:12	gradha	maybe I should add my awesome macro too
12:56:36	dom96	'using' does make more sense in block form IMO. I said that when it was introduced, right now I can't remember what the argument was against my opinion though heh
12:57:31	gradha	the argument against that was likely you could not disambiguate conflicting module imports
13:01:12	dom96	I don't see how not using it in block form helps with that?
13:01:43	gradha	if you have a block form, for each "using" you have to indent your whole source one level
13:02:17	gradha	the manual example makes it quite clear:
13:02:18	gradha	import windows, sdl
13:02:18	gradha	using sdl.SetTimer
13:06:30	dom96	oh, i see.
13:07:03	dom96	Well, in that case it should be a hybrid. You should be able to use it in a block form and as a single statement.
13:13:50	*	PortablePuffin quit (Read error: Connection reset by peer)
13:17:52	*	PortablePuffin joined #nimrod
13:22:44	*	Varriount|Mobile joined #nimrod
13:44:59	Araq	I don't get the example
13:45:07	Araq	"using sdl.SetTimer" wtf?
13:45:25	Araq	we already have from sdl import SetTimer
13:45:54	dom96	hrm, true.
13:45:59	gradha	we will have to ask zahary
13:48:19	*	BitPuffin joined #nimrod
13:56:43	*	Varriount|Mobile quit (Ping timeout: 272 seconds)
13:58:38	*	Varriount|Mobile joined #nimrod
13:58:54	*	Varriount|Mobile quit (Client Quit)
13:59:16	*	Varriount|Mobile joined #nimrod
14:00:43	Varriount|Mobile	How would one go about printing the value in a particular register in nimrod?
14:01:19	Araq	inline assembler
14:06:22	*	Varriount|Mobile quit (Remote host closed the connection)
14:14:35	*	bstrie quit (Ping timeout: 245 seconds)
14:15:35	*	bstrie joined #nimrod
14:59:36	*	faassen joined #nimrod
15:02:00	BitPuffin	dom96: ?!?!?!
15:02:22	dom96	BitPuffin: yes?
15:03:10	BitPuffin	dom96: dota nao?
15:03:46	Araq	BitPuffin: let's make a deal
15:04:02	BitPuffin	Araq: deal?
15:04:20	Araq	I send you the plan, you implement the runtime parts of the concurrency model :P
15:05:01	BitPuffin	lol ;)
15:05:08	BitPuffin	like I have the skillz!
15:06:13	BitPuffin	And if it was fairly trivial then I'd imagine you would have already done it yourself hehe
15:06:27	BitPuffin	dom96: alright too late now I'm gonna play xD
15:06:33	dom96	bah
15:06:45	dom96	I should study anyway...
15:06:58	Araq	bbl
15:07:27	BitPuffin	dom96: or if you get going I'll practice with bots while you prepare
15:07:41	BitPuffin	dom96: no time to think
15:07:44	BitPuffin	yes or nein
15:07:46	BitPuffin	ja or nein
15:07:58	BitPuffin	ja eller nej
15:08:01	dom96	nah
15:08:04	BitPuffin	nub
15:08:06	dom96	Maybe i'll play later
15:28:49	*	ddl_smurf joined #nimrod
15:34:14	*	aftersha_ joined #nimrod
15:34:42	*	faassen quit (Quit: faassen)
15:57:48	*	aftersha_ quit (Quit: Computer has gone to sleep.)
16:02:11	Araq	hmm is it just me or have {.pragmas.} stopped being ugly?
16:02:36	Araq	I removed the dots to see how the code then looks
16:02:53	BitPuffin	Araq: I think there is some beauty to it
16:02:53	Araq	and for some reason {.pragmas.} are sexy now :P
16:03:07	BitPuffin	dom96: I won ;)
16:03:10	BitPuffin	I played Chen
16:03:20	BitPuffin	the ultimate micro manage hero
16:03:23	BitPuffin	really annoying
16:03:26	BitPuffin	but it went quite well
16:04:00	BitPuffin	dom96: another window for you to join in
16:04:03	BitPuffin	going in 10
16:04:09	BitPuffin	5
16:04:11	BitPuffin	3
16:04:13	BitPuffin	2
16:04:15	BitPuffin	1
16:04:17	BitPuffin	0
16:04:25	BitPuffin	okay I'll play another one by myself then
16:04:35	Araq	can't you use PM for your gaming arragements?
16:05:01	Araq	it's becoming annoying
16:05:08	*	PortablePuffin quit (Remote host closed the connection)
16:15:08	*	Mordecai is now known as psquid
16:35:03	*	aftersha_ joined #nimrod
16:46:22	Varriount	Araq, good morning!
16:47:59	*	aftersha_ quit (Quit: Computer has gone to sleep.)
16:49:54	Varriount	Araq, theoretically, is it possible to implement 'using' as a block macro?
16:54:25	*	fntzr quit (Quit: Leaving)
17:08:03	*	Demos__ joined #nimrod
17:09:36	OrionPK	you mean like a C# style using?
17:10:47	Araq	Varriount: no otherwise zahary would have implemented it as a macro
17:11:22	OrionPK	are you talking about using or with?
17:11:40	Araq	the question was about "using"
17:12:19	OrionPK	I thought zahary was implementing a with statement
17:12:34	Araq	yeah but he called it "using"
17:13:14	OrionPK	so then my question is did varriount mean C# style using or javascript style with :P
17:13:32	Araq	our "using" is "with"-like
17:13:41	OrionPK	yeah
17:13:54	Araq	and btw JS got it from Pascal
17:14:06	OrionPK	im sure C# got it from somewhere
17:14:25	Araq	and I've seen it produce horrible code in pascal hence I'm against it
17:14:41	OrionPK	its the same story with JS
17:14:53	Araq	but then "can be misused" is not really a criterion we care about ...
17:15:21	Araq	"can be misused" is almost like "can be confusing"
17:18:25	Araq	so ... after 3 iterations the new thread stuff is so incompatible with the old API that we can simply deprecate everything and move on :-)
17:18:56	OrionPK	out with the old, in with the new
17:20:09	Araq	I still hope it works in practice
17:22:34	Araq	looks *really* good though :-)
17:23:48	OrionPK	cant wait to see it
17:24:15	Araq	I should be done by tonight
17:24:21	Araq	I know I've said this before :P
17:25:29	dom96	with the design or the implementation? ;P
17:25:34	Araq	the design
17:25:56	Araq	but I kept the design implementable
17:28:23	Demos__	with seems like a bad idea...
17:29:35	Araq	Demos__: to answer your question, the compiler already makes use of the fact it's allowed to pass by "const&"
17:30:00	Demos__	wow, kinda an old question but thanks
17:30:07	Varriount	Although 'using'/'with' seem like nice things, I'm afraid that they probably won't be needed very often (see: the 'do' statement)
17:30:52	Araq	zahary has lots of meta programming use cases that require 'using' under the hood
17:31:18	Araq	so we'll have some hard time to argue against it
17:31:56	Varriount	Araq, the problem with all this meta-programming is that half of it is broken or limited, or verbose
17:32:19	Araq	hey
17:32:51	Araq	saying that is like blasphemy here
17:33:03	Varriount	Araq, I meant the type-class metaprogramming stuff
17:33:21	Varriount	Not the 'with' statement, macros, etc
17:33:21	Araq	well he's working on it
17:33:29	Varriount	True. *sigh*
17:33:57	Varriount	I'm just grumpy because I can't build nimrod
17:34:27	Araq	yeah I can feel your pain
17:34:34	Varriount	In other news, Araq, it might actually be possible to implement that file id thing you mentioned.
17:34:44	Araq	cool
17:35:18	Demos__	Varriount, I had the same problem as you building devel
17:35:30	Varriount	The only.. caveat is that how 'unique' an id is depends on the os and fs
17:35:56	Araq	er ... what?
17:36:22	Varriount	Operating System and File System
17:36:39	Araq	er ... what?
17:36:59	Varriount	Depending on the two, a file copy or move operation could change file's ID
17:37:23	Araq	that's not an issue
17:38:22	Araq	oh sorry
17:38:44	Araq	time machine stroke again
17:38:51	Varriount	Howso?
17:38:57	Araq	os.sameFile already exists
17:39:48	Varriount	True. But if we have a working stat() like call, we could just replace sameFile's machinery to use the the proc/type
17:42:25	Varriount	Demos__, yeah, I read your message last night - apparently some wierd code zahary introduced is getting data types muddled.
17:44:12	Araq	well you can always roll back
17:47:07	Araq	samefile is the wrong interface though
17:47:22	Araq	an id can be hashed etc
17:49:46	gradha	what's the equivalent of "\n" for a character other than using the hex code?
17:50:23	Araq	\L
17:52:38	gradha	nice, zero hits in the docs
17:53:39	Araq	and now you know what you have to do
17:53:47	gradha	same for \T
17:53:52	Araq	and now you know what you have to do
17:54:06	gradha	does \L mean different line ending on different platforms?
17:54:12	Araq	no
17:54:23	Araq	it's always LF aka binary 10
17:54:39	Araq	\C is always CR aka binary 13
17:54:56	Araq	\R is the same as \C iirc
17:55:06	Araq	but \F is not the same as \L
17:56:34	Varriount	You know, it amazes me how similar Windows API's are to parts of the Posix standard
17:57:02	gradha	careful there
17:57:06	Araq	it amazes me how often windows' API is better :P
17:57:24	Araq	but I'm biased
17:57:36	Varriount	It's like the designers said "Hey, we should use these parts of posix, but modify them in such a way as to make them difficult to use in cross platform programs"
17:57:51	OrionPK	lol
17:58:27	Araq	nah ... I take it back
17:58:34	Araq	windows' API is awful
17:59:11	Araq	but at least they cared about binary compatibility and defined the numeric values of constants
17:59:39	EXetoC	Varriount: That would almost be surprising
17:59:46	Araq	instead of saying "yeah there is this opaque constant EFOO here, so you can only use it from C"
18:00:36	Araq	"Or you can write a C program to determine its value"
18:01:22	Araq	"good luck with that approach for the things we defined as C macros though :P "
18:02:50	Varriount	For example, most of the information returned by stat() on posix systems can be gathered by collating the data returned by the GetFileInformationByHandle and GetSecurityInfo
18:03:45	Araq	Varriount: windows is based on DOS though, not on Unix, so why should they care about Posix? Posix only sets in stone the design mistakes on Unix anyway
18:04:24	Varriount	Araq, you mean, like horrible naming conventions?
18:04:59	Araq	it's called prgmmng wtht vwls
18:05:07	Araq	("programming without vowels")
18:05:16	OrionPK	there's a \L ?
18:05:18	OrionPK	shit
18:05:20	OrionPK	that would be nice to know
18:06:21	Varriount	Araq: Well of course, everyone knows that vowels are an endagered species.
18:07:34	Varriount	Araq: I"m curious though, what do you feel are the design mistakes of posix/unix? Being around mostly Linux programmers, I've never really met anyone who gave it much criticism
18:07:53	Araq	uh oh not again
18:08:07	Varriount	Huh?
18:08:09	Araq	fork() lead to memory overcommitting for instance
18:08:55	Araq	redirection of stdin/stderr/stdout requires some heavy drinking to get right
18:09:22	Araq	even though it's claimed to be the essence of that OS
18:09:40	Araq	the sockets APIs are archaic crap
18:09:57	BitPuffin	Araq: ping me when you publish. I am eager to read
18:10:00	EXetoC	so that's just an issue with certain shells?
18:10:07	EXetoC	*not just
18:10:10	gradha	Varriount: you already asked the question in previous irc sessions, seems you didn't have enough
18:10:10	BitPuffin	it's gonna be intersting I am sure
18:10:51	Araq	the signal mechanism is ill specified and can be used for subtle attacks
18:11:50	Varriount	gradha: I like listening to people complain about topics that I have no opinion on. For one thing, it helps prepare me for upcoming problems.
18:12:17	Varriount	Araq: How do you feel about the Plan 9 os?
18:12:27	gradha	Varriount: that only works if you go now to #unix and ask why the windows API sucks
18:12:55	Varriount	gradha: I already know why the Windows API sucks.
18:13:05	Araq	the security model is primitive at best and simply doesn't work for desktops but then so does windows's model
18:14:58	gradha	Varriount: opinions are like pokemon, kind of different, and you never seem to collect all
18:15:39	Varriount	gradha: But each one is special, in it's own unique way.
18:15:43	EXetoC	Araq: the various shells out there could simplify redirection. or is the problem bigger than that? because those are the only specifics you've discussed IIRC
18:15:51	Varriount	Besides, I like pokemon.
18:16:08	gradha	meh, digimon beats pokemon any day
18:16:13	Araq	EXetoC: just check out osproc.nim (which we still got wrong ...)
18:16:27	Araq	has nothing to do with shells
18:16:37	Araq	but these would be my next complains
18:16:55	Araq	strange quoting rules, strange interaction of features
18:17:06	Araq	or how about ANSI escape codes?
18:17:18	Araq	ever tried to redirect something that produces colored output?
18:17:58	Araq	and the escape codes are in a way very typical
18:18:09	gradha	talking about colors, there's still no way to deal with color text and c++ people want a standard 2d drawing library
18:18:24	Araq	because Unix has a fundamental problem with *APIs*
18:18:41	Araq	so they come up with lots of other "protocols" instead
18:18:41	Varriount	^*Posix
18:19:32	gradha	see http://lists.cairographics.org/archives/cairo/2013-December/024858.html and facepalm yourself
18:19:56	gradha	there's still no standard way to put a pixel on the screen
18:20:19	Araq	gradha: yes because that would require an API
18:20:33	Araq	instead of dumping crap onto stdout
18:20:37	Varriount	Ugh. C++ api
18:20:43	joelmo	im reading in the tutorial about templates, how do i use pattern operators https://gist.github.com/joelmo/69aade091324cecbaee6
18:20:46	Araq	or is it stderr? who knows
18:20:49	gradha	well, there's no standard way to put text on the screen either
18:21:05	gradha	C really sucks for input/output other than "streams"
18:21:08	EXetoC	Araq: got it
18:22:01	Araq	Varriount: and finally package managers really are about patching over the braindead /usr/bin file hierarchy
18:22:06	Varriount	joelmo: What is the expected output?
18:22:14	joelmo	Varriount: 2
18:22:20	Araq	yes they are also about the dependencies
18:22:24	Varriount	joelmo: And what do you get?
18:22:45	joelmo	1
18:23:16	joelmo	its in the gist :P
18:23:44	Araq	joelmo: bug report please
18:24:08	Araq	though it's a strange example and not guaranteed to work
18:24:19	Araq	I'm curious why it doesn't work anymore
18:24:51	joelmo	I can post a bug report and link to the gist and the manual
18:25:21	Araq	I know it's from the manual
18:25:42	EXetoC	Araq: how would one deal with the upgrading of all packages then? and only those that aren't up-to-date already
18:26:11	Varriount	gradha: Why would a sane proprietor of standards, seeking something like a standard 2d library, want it to be in C++?
18:26:32	Araq	EXetoC: a package manager is useful but it's still also a patch over the broken directory hierarchy
18:26:48	gradha	Varriount: maybe they enjoy pain, or inflicting it unto others
18:26:52	EXetoC	right
18:27:27	Varriount	Isn't c++, like, anathema to standardization efforts? (or at least, interfacing efforts)
18:27:45	Demos__	well if c++ had proper modules it would be a pretty nice language (unless you like writing tools)
18:28:49	Varriount	Demos__: It would also be nice if there was some sort of api/abi standard.
18:28:55	*	gradha PARSING ERROR: nice and c++ found
18:28:57	Demos__	Varriount, well standardization was so successful that there are like 12 implementations, this causes problems for interfacing. But quite honestly if you want to be easy to interface with you need to be C
18:29:18	Demos__	well there is a standard API, called the c++ standard library
18:29:29	Demos__	I don't think a standard ABI is a good idea
18:29:40	Araq	Demos__: I disagree with you
18:29:43	EXetoC	gradha: should there really be a standard way for stuff like that? it's C we're talking about
18:29:53	Araq	there is plenty of stuff in C that doesn't interface well with other languages
18:30:02	Araq	you have to pick of subset of C for that to work
18:30:09	Araq	same for C++ really
18:30:24	Demos__	yeah, that was my point. No matter what you do the subset ends up being more or less the same
18:30:32	Araq	ok the subset is larger for C then for C++ but only because C is smaller language
18:30:50	Araq	*than
18:31:39	Araq	and partly the linking technology straight from the 60ies is to blame for that
18:31:51	Demos__	I partly defend c++ because I have spend so much time in it and I want to feel like that time was not wasted
18:31:53	gradha	EXetoC: standards are good if the market itself can't reach a defacto standard, I don't think a 2d api needs that
18:32:15	Demos__	gradha, I am not so sure
18:33:14	gradha	consider the opengl standard with regards to hardware, it was good when card vendors didn't agree what features should be supported at all
18:33:35	gradha	of course, everybody at some point catched up with opengl, and… opengl lagged behind, which is when standards suck
18:33:50	Demos__	the trouble with graphics is that you end up with group A that is interested in just SCREEN 7 PUT GREEN and grop B that wants as much programmability as possible
18:33:50	gradha	two sides from the same coin
18:34:03	EXetoC	is it not back on track now?
18:34:20	gradha	it's far behind DirectX according to people in the know
18:35:00	Demos__	I was under the impression that GL and DX are pretty much the same feature wise. DX is much less painful though
18:35:00	gradha	OpenGLES may play the same role, unifying the API of mobile devices, then preventing them from improving
18:35:45	Demos__	I think that graphics APIs that know something about the language they are being used from are a good thing
18:35:55	gradha	Demos__: I've been told the GL is only equals to DX if you use each card's propietary GL extension, which is another way to say "fuck you standard"
18:36:00	Demos__	not true
18:36:12	Demos__	GL4.4 has most everything core
18:36:45	Demos__	maybe some cutting edge Direct3D 11.2 stuff is missing but that is /really/ cutting edge
18:37:15	Demos__	now actually getting driver support for GL4.4 is another story
18:37:39	Demos__	if you use GL4.4 you are pretty much going to only run on windows
18:38:17	EXetoC	gradha: they didn't agree? are you referring to the fact that there are numerous extensions?
18:38:43	EXetoC	because GL was indeed first IIRC, and then that highly platform-specific API came along
18:40:00	Demos__	one thing that I like about direct3d is that MS makes zero effort to maintain compatibility between versions, so the whole thing is less ... wierd .. than openGL.
18:40:02	gradha	EXetoC: that's what I said, when everybody catched up with GL, it stopped being useful and lagged in progress of features
18:44:30	Demos__	anyway, back to c++... I think c++ is largely well designed if you ignore the syntax. Considering the goals of the language I think it does pretty well. wrt ABI compat I fear that trying to pin down an ABI just gets you stuff like COM and GObject
18:45:33	Araq	Demos__: may be surprising for you, but I agree
18:45:39	Demos__	oh, does anyone have an answer to my question about refs and ptrs?
18:46:04	Araq	you can try and cast ptr to 'ref' if the code never invokes the write barrier
18:46:14	EXetoC	gradha: yeah, and the only competitor now is OS-specific. I strongly dislike that approach
18:47:13	Araq	you can compile with -d:useSysAssert and -d:useGcAssert and then you get some nice checking for that
18:47:54	Demos__	so my actual situation is that I have an object with an "Open" function that returns a ref, and I would rather like to use pointers there, I think I am screwed though. Are there docs for those options?
18:49:15	Araq	why not use GC_ref/unref on the 'ref' immediately?
18:49:45	Araq	it's the same situation as returning a string as cstring and copying it immediately
18:51:02	Demos__	that is what I am doing now and my memory is getting corrupted, maybe the problem is elsewhere though
18:51:24	Araq	you can compile with -d:useSysAssert and -d:useGcAssert and then you get some nice checking then
18:51:57	Demos__	if I return a ref TFoo through a C function and treat it as a void* on the other side does that mess up the GC?
18:52:22	EXetoC	gcAssert is for debugging the GC?
18:52:26	Araq	no, what does mess up the GC is likely the fact that C# runs your stuff on a different thread
18:52:39	Araq	and so the GC tracks the wrong stack
18:52:47	Araq	that's what I would guess anyway
18:52:58	Araq	try GC_disable/enable
18:53:01	Araq	EXetoC: yes
18:53:10	Demos__	well I am never passing structures to/from c#
18:53:36	Araq	structures seem to be safer than pointers for your use cases though
18:55:14	Demos__	so c# just keeps an Int32 that nimrod gives it and passes that to nimrod which derefs it and gives me a member. Maybe structures would be better, but some of the structures in question contain stuff like strings and refs
18:55:32	Araq	Demos__: called 'setStackBottom' ?
18:55:44	Araq	cause you really should
18:56:04	Araq	in your C# main
18:57:10	*	OrionPK quit (Remote host closed the connection)
18:58:04	Demos__	can not find docs for that function...
18:58:16	Araq	yeah ... :-/
18:59:14	Demos__	those assertion options moved me closer to the bug
18:59:47	Demos__	it would be really nice if c#'s AccessViolationException actually reported the memory location you tried to access
19:01:00	Araq	Demos__: it's exported as initStackBottom(void)
19:01:17	Araq	so try to call that please
19:01:38	Araq	btw does use Visual Studio threads? I'm sure it does ...
19:01:56	Araq	you need to call that in the thread that runs the nimrod code
19:03:02	*	OrionPK joined #nimrod
19:05:54	Demos__	I have no idea which thread runs the code. You invoke some OLE related incantations and your code gets called... This is why I kinda wanted to use manual memory
19:06:29	Demos__	and setStackBottom is apparently exported as nimrtl_setStackBottom according to dependency walker
19:08:28	Araq	quite possible
19:08:40	Araq	but you need initStackBottom
19:09:30	Araq	well frankly ... if you don't know which thread runs the code your best bet is to disable the GC and enable it after all is over
19:09:34	Araq	sad but true
19:10:34	Demos__	wont that leak memory?
19:11:07	Araq	yeah but it might be fine, not sure what you're doing
19:13:31	Demos__	yeah, that "fixed" my problem
19:14:17	Araq	well the aggressive GC that's picky about its main stack has its downsides :P
19:14:43	Araq	you can also try to link against --gc:boehm
19:15:15	Demos__	well presumably it is better than the one that is only picky sometimes. What about using allocShared and deallocShared and passing the pointer off to the c# gc?
19:15:34	Araq	that's a good idea
19:16:33	Araq	and that really should work :P
19:18:04	*	OrionPK quit (Remote host closed the connection)
19:18:56	Demos__	but I get back into those functions that operate on refs, I could reimplement them but that would just be copy-pasting and changing a ref to a var.
19:20:26	Araq	well if you disable the Gc, ref is fine for those
19:21:23	Demos__	should I reimplement the initialization functions (the ones that go new(result))?
19:24:00	Araq	if you get rid of 'ref' that's what you need to do
19:25:26	*	OrionPK joined #nimrod
19:27:28	Demos__	OK final question: if I allocate something with alloc or allocShared that contains a string or ref and the GC is disabled how do I deallocate it. And if I used allocShared where does the string or ref go?
19:29:06	Araq	the Gc doesn't track what's in the memory allocated by alloc, so you need to GC_ref/unref these
19:29:45	Araq	the string or the ref always go to the thread local GC'ed heap
19:29:51	Araq	always
19:30:12	*	zielmicha-cloud_ quit (Ping timeout: 246 seconds)
19:30:19	*	zielmicha-cloud_ joined #nimrod
19:30:28	Demos__	but if the GC is off than I have no way of ever deallocating the ref or string
19:31:00	Araq	well you can always enable the GC and make it perform a cleanup
19:31:50	Demos__	right, but then I have the same problem with threads
19:32:26	Araq	which problem exactly? disabling the GC solves it
19:33:38	Demos__	the problem of corruption due to not knowing which thread I am on. And I don't want to just disable the GC and leak memory
19:37:25	Araq	well that's a hard problem to solve but not impossible unless VS doesn't even run it always on the same thread
19:37:40	Araq	though thinking about it ...
19:37:55	Araq	what's wrong with calling initStackBottom?
19:40:08	*	OrionPK quit (Ping timeout: 246 seconds)
19:46:44	Demos__	where is initStackBottom it is not in the index and I cant find it in dependency walker
19:47:20	Araq	grep for it, exportc it
19:47:30	Araq	document it
19:48:27	Demos__	k
19:52:19	*	Mat3 joined #nimrod
19:52:22	Mat3	hi all
19:52:38	gradha	yay
19:53:31	*	Icefoz quit (*.net *.split)
19:53:32	*	wan quit (*.net *.split)
19:53:38	Mat3	hi gradha
19:58:22	Araq	hi Mat3
19:59:10	Mat3	hi Araq
19:59:29	Mat3	how about your documentation effort ?
19:59:47	Araq	I'm in the 3rd iteration now
20:00:01	Araq	and things are now really looking good
20:00:13	Araq	plan to finish it tonight
20:00:35	*	BitPuffin quit (Quit: WeeChat 0.4.2)
20:02:56	*	OrionPK joined #nimrod
20:03:34	Mat3	nice
20:07:15	*	tumak_ joined #nimrod
20:08:23	*	tumak quit (Read error: Operation timed out)
20:10:39	*	gradha wonders if reading about kpop hearing metal is kosher
20:12:56	*	Mat3_ joined #nimrod
20:18:49	*	Mat3 quit (*.net *.split)
20:18:57	Mat3_	is someone here interested in a soft-core for FPGA (my VM design) ?
20:21:59	OrionPK	i would prefer hardcore
20:23:21	*	aftersha_ joined #nimrod
20:23:42	*	OrionPK quit (Remote host closed the connection)
20:25:37	*	capisce_ joined #nimrod
20:25:38	*	tumak joined #nimrod
20:33:13	Demos__	what the fuck
20:36:47	*	betawaffle quit (Remote host closed the connection)
20:36:50	*	ddl_smurf quit (*.net *.split)
20:36:51	*	gradha quit (*.net *.split)
20:37:03	*	mal`` quit (Read error: Connection reset by peer)
20:37:03	*	alex_nx quit (Write error: Connection reset by peer)
20:37:05	*	aftershave quit (Ping timeout: 264 seconds)
20:37:11	*	eigenlicht quit (Ping timeout: 264 seconds)
20:37:11	*	tumak_ quit (Ping timeout: 264 seconds)
20:37:19	*	capisce quit (Ping timeout: 264 seconds)
20:37:20	*	ponce__ quit (Ping timeout: 264 seconds)
20:37:26	*	eigenlicht joined #nimrod
20:37:27	*	Mat3_ interesting insights for a public channel, ehm
20:37:27	Araq	Mat3_: please elaborate
20:37:29	Mat3_	an FPGA is as much hardcore as an ASIC of course *g*
20:37:36	*	capisce joined #nimrod
20:37:36	*	capisce_ quit (Ping timeout: 260 seconds)
20:37:51	*	alex_nx_ joined #nimrod
20:37:53	*	mal``` joined #nimrod
20:37:59	*	alex_nx_ is now known as alex_nx
20:38:17	*	alphawaffle joined #nimrod
20:38:17	*	alphawaffle is now known as betawaffle
20:39:10	*	ddl_smurf joined #nimrod
20:39:10	*	gradha joined #nimrod
20:39:43	*	EXetoC quit (Write error: Broken pipe)
20:47:24	Demos__	I am ddebugging a problem where visual studio crashes if you type a keyword followed by a lowercase l
20:47:43	Demos__	although it crashes in a whole lot of other circumstances as well, so who knows
20:47:44	Araq	visual studio is cool
20:53:40	*	OrionPK joined #nimrod
20:54:21	Mat3_	do not know Visual Studio, however working with an effective programming editor like Notepad++ may be an simple alternative for you
20:54:33	*	OrionPK quit (Remote host closed the connection)
20:54:58	*	zielmicha8 joined #nimrod
20:55:01	*	OrionPK joined #nimrod
20:58:09	Demos__	well I am using notepad c++ but I am debugging code running in VS
20:58:21	Demos__	VS is not buggy my code is buggy
21:02:00	*	zauberparacelsus joined #nimrod
21:04:06	zauberparacelsus	As far as web microframeworks for Nimrod go, are there any alternatives to Jester? I looked at it, and unfortunately it doesn't fill my needs. Wildcard URL's are a feature that I would need so that I can properly convert my wiki engine to Nimrod.
21:05:33	dom96	wildcard urls?
21:06:28	zauberparacelsus	yeah, matching one URL string to potentially many other URLs. Like, for example, if the URL is "/page/1" or "/page/2", then they would all match to "/page/*"
21:06:41	dom96	You can do that with jester
21:06:51	zauberparacelsus	really? The docs said that you can't
21:07:15	zauberparacelsus	though, since I now see you're the dev, I'll take your word for it, lol
21:07:37	dom96	hrm, well I assumed "/page/@page" would fit your needs heh
21:07:46	zauberparacelsus	lol
21:08:03	dom96	I just remember that that is one of things from sinatra that I did not implement
21:08:06	zauberparacelsus	now, suppose I want to just match all possible URLs to a single pattern?
21:08:18	OrionPK	cant you use a regex
21:08:21	OrionPK	getre
21:08:26	dom96	yeah, you can use regex
21:08:34	Mat3_	Demos_: Try fdbg
21:09:24	dom96	zauberparacelsus: What is that you want to do exactly?
21:10:00	dom96	*is it
21:10:01	zauberparacelsus	basically, my wiki engine, which is implemented in Python and Flask, would match all URLs down to a single route.
21:11:36	OrionPK	why wouldnt a standard pattern work for that
21:11:58	*	zielmicha8 quit (Remote host closed the connection)
21:12:06	*	zielmicha8 joined #nimrod
21:12:06	dom96	yeah, presumably all you need is: get "/@pagename": ...
21:12:12	zauberparacelsus	dom96: thank you
21:12:37	Mat3_	ciao
21:12:54	*	Mat3_ quit (Quit: http://quassel-irc.org - Chat comfortably. Anywhere.)
21:16:10	Demos__	ha! found initStackBottom() strange that windows search did not pick it up, but nimgrep did
21:16:43	*	[2]Endy quit (Ping timeout: 252 seconds)
21:17:11	OrionPK	windows search doesn't look inside files, does it?
21:17:36	Demos__	it sometimes does
21:17:55	Demos__	oh, I think it just ignores nim files
21:18:45	OrionPK	you can set up full text indexes
21:19:13	OrionPK	I don't think that it's going to search in files ootb tho
21:20:36	Demos__	yeah, I just added nim to the list of files it will fully index
21:29:55	Araq	why do you use windows search when you can use nimgrep? :P
21:30:09	Araq	nimgrep is the best tool I've ever used *cough*
21:31:57	*	EXetoC joined #nimrod
21:36:07	*	brson joined #nimrod
21:37:12	Demos__	OK so the way to "take control" of c# gc is with finalizers, they get run a bit before the gc frees memory for an object. It turns out that the thread that the finalizer runs on in unspecified...
21:40:03	OrionPK	you're in dangerous territory :P
21:40:24	zauberparacelsus	are there any YAML parsers for Nimrod?
21:40:27	Araq	finalizers in C# and Java are quite slow though
21:40:34	OrionPK	no zauberparacelsus
21:40:45	Araq	zauberparacelsus: not that I know of, there are JSON parsers though
21:40:50	zauberparacelsus	ok, thanks
21:40:57	Araq	and wrapping a C lib is not hard
21:41:05	zauberparacelsus	yeah
21:41:10	Araq	if you do that, please put it on babel
21:41:18	zauberparacelsus	aye
21:41:24	Araq	so the next time we can answer with "yes" :-)
21:41:31	zauberparacelsus	yep
21:43:24	Demos__	yeah, I know they are slow, ima cross that bridge when I get there. And I also know I am in dangerous territory. The fact is that having to manually manage memory from an unmanaged heap even though you have a GC just sitting there is dumb, hence finalizers
21:43:51	Araq	sure, I would use them too
21:44:33	Araq	just sayin' that the nice generational copying collection GC has its downsides
21:44:50	OrionPK	so anyone want to try out a windows build of irc familiar
21:45:03	Araq	because if you copy what's alive, you don't really know what's dead
21:45:15	Araq	and so finalizers start to get really expensive
21:46:14	zauberparacelsus	are there any guides for creating bindings to C libraries?
21:46:28	Demos__	well yes, but c# does not care. c# is about getting somewhat decent performance without much effort
21:46:45	Demos__	http://lucumr.pocoo.org/2013/8/18/beautiful-native-libraries/
21:48:43	OrionPK	https://dl.dropboxusercontent.com/u/417554/ircfamiliar.zip
21:53:48	gradha	http://cbloomrants.blogspot.com.es/2013/12/12-31-13-statically-linked-dll.html
21:56:17	OrionPK	jesus christ gradha
21:56:21	OrionPK	are you trying to get us all killed
21:57:28	Araq	oh god
21:57:31	Araq	" Because DLLs are also an annoying disaster on Windows (having to distribute multiple files, accidentally loading from an unexpected place, and what if you have multiple products that rely on different versions of the same DLL, etc.). "
21:57:48	Araq	does the guy even know how DLLs work on windows? o.O
21:58:12	Araq	it's dead simple to use multiple different DLL versions ...
21:58:27	Araq	from multiple products at least
21:58:49	Araq	heck, I can still run all my windows software everywhere including on wine
21:58:57	Araq	try that with a linux binary
22:06:14	Demos__	well that is only because everyone distributes the windows dlls with the application
22:06:27	Araq	*shrug* it works
22:07:00	Araq	and it works since decades
22:07:40	Demos__	but it is strictly worse than using static libs, since you get none of the benefits of dynamic libraries
22:08:15	Araq	perhaps but it *works* :P
22:08:44	Demos__	true
22:08:45	Araq	windows is the only OS which really cares about to keep running your software
22:10:14	Araq	and most DOS based games are more portable than the average JVM software thanks to dosbox
22:11:02	Araq	the morale of the story is: if your software is valuable, it's portable soon enough
22:11:13	gradha	does dosbox support now 32bit mode?
22:11:27	Varriount	OrionPK: Just downloaded ircfamilier. Chrome says that it is "a potentially dangerous file"
22:11:33	Araq	I think so, gradha
22:11:34	Varriount	:O
22:12:45	Varriount	I really wish you could restrict a program's file access in a fine-tuned way..
22:13:04	Araq	gradha: I used to run warcraft 1 on it which is a 32bit application
22:13:41	gradha	it depends on the mode, I remember not being able to run any djgpp binaries, but I guess they have had time to improve since
22:14:01	Araq	well I said *valuable software* :P
22:14:10	gradha	ah, right
22:14:54	Varriount	OrionPK: Heads up, ircfamilier depends on pcre.dll
22:14:59	Araq	we should port nimrod to DOS
22:15:11	Araq	Varriount: nimrod ships with pcre.dll
22:15:51	Varriount	Araq: Yes, but end users of a program don't typically have nimrod installed.
22:16:12	Araq	so copy pcre.dll into the same folder as your .exe
22:16:31	Araq	that's how "DLL hell" is avoided ...
22:16:41	Varriount	Well yes, I know that. I just didn't know if OrionPK knew.
22:16:42	NimBot	Araq/Nimrod devel d3c9003 Zahary Karadjov [+2 ±11 -0]: introduce tyFromExpr; fixes #618
22:16:42	NimBot	Araq/Nimrod devel d08a21a Zahary Karadjov [+0 ±9 -0]: progress towards fixing tgenericshardcases
22:16:42	NimBot	Araq/Nimrod devel b4bdd39 Zahary Karadjov [+0 ±1 -0]: fix bootstrapping on windows
22:16:51	Varriount	Yay!
22:16:59	Araq	arg ...
22:17:12	gradha	uh oh
22:17:17	Araq	zahary: what's tyFromExpr and do we really need it?
22:18:01	Varriount	gradha: You pull up the seats, I'll get the popcorn
22:18:07	zahary	this is something that is completely unknown until we instantiate a generic; things like calls to compileTime procs accepting the generic params or exprssions like T.hight
22:18:33	*	zauberparacelsus left #nimrod (#nimrod)
22:18:34	zahary	see tgenericshardcases for such tortures to the compiler
22:18:44	Araq	so that's for typing generics?
22:19:03	Araq	hmm ok
22:19:07	zahary	it should be clear if you look at the test case
22:19:16	Araq	I remember it :P
22:19:22	Araq	no need to look at it again
22:19:45	Araq	so tyExpr doesn't work for that?
22:20:43	zahary	well, I'm trying to reduce the overlapping of meanings now (reusing the same tyType for multiple purposes) - tyExpr is a macro parameter that doesn't require semantic checking
22:21:16	Araq	that's good
22:21:34	Araq	but promise me it's your last addition :P
22:21:45	Araq	we really have lots of ty's now
22:21:54	Varriount	I wonder if, among all this metaprogramming, it might not be prudent to simply turn the generic instanciation syntax ('genericProc[type]') into simply a partial function applicator
22:22:16	Araq	and you always need to keep them in mind in skipTypes et al
22:22:19	*	BitPuffin joined #nimrod
22:22:48	Araq	Varriount: you still haven't convinced us that's a good idea
22:22:49	*	familiar125 joined #nimrod
22:22:57	Araq	hi familiar125 welcome
22:23:09	BitPuffin	I don't think that's a new nick
22:23:14	BitPuffin	Araq you are losing it
22:23:24	Araq	familiar125 is new
22:23:28	BitPuffin	WHERE'S THE BLOG POST, I'M LIKE REFRESHING AND SHITZ
22:23:44	Araq	hey
22:24:02	zahary	one last addition that I plan is the introduction of what I call "mimic types", but it won't require new ty*s maybe; a mimic type is type that promises to behave exactly as another type; I promise that my_string behaves like string; then if I try to use my_string with a string accepting proc, the compiler treats the proc just if it was a generic - you'll get its body compiled for my_string
22:24:04	BitPuffin	sry
22:24:24	*	familiar125 quit (Remote host closed the connection)
22:24:31	BitPuffin	aw snap diddely
22:24:32	Araq	zahary: I had this idea too once
22:24:36	BitPuffin	what's the point of it zahary
22:24:56	Araq	I came to the conclusion it can't work iirc
22:25:22	zahary	well, nominal typing must die, but that's a long story (BitPuffin);
22:25:26	Araq	BitPuffin: it turns concretes into generics
22:25:38	zahary	hmm, what problems do you see with it. I should work
22:26:07	Araq	I called that "proxy" type back in the days iirc
22:26:12	zahary	yes, I know
22:26:16	Araq	"mimic type" is nicer
22:26:23	BitPuffin	how about
22:26:26	BitPuffin	proxic
22:26:35	BitPuffin	or mixy
22:26:49	Varriount	BitPuffin: Proxic sounds like toxic. Mixy sounds ok though
22:27:13	Araq	well one problem is that you don't really want to support everything that "string" does
22:27:36	zahary	and mimic types are a bit more precise - you can tell specifically what you "mimic" and you can even say things like seq[char] is JavascriptByteBuffer
22:27:51	Araq	because then you end up re-implementing string as it is today
22:27:53	zahary	precisely, you only need to support what is required by the procs that you end up using
22:28:04	zahary	just like normal generics
22:28:20	Araq	yeah but the procs promised to work on "string" not on "string-like" so it's fragile
22:28:55	Araq	one version of toUpper works, the next version uses 'newStringOfCap' (optimized) and breaks your code
22:28:59	*	VarriountFamilie joined #nimrod
22:29:12	VarriountFamilie	Hm. Interesting
22:29:26	Araq	but as an experimental feature it's fine
22:29:44	zahary	sure, but my goal is to just battle nominal typing when third party libraries don't cooperate in producing generic code
22:30:03	Araq	just mark it as experimental
22:30:08	Araq	and "using" too btw
22:30:31	Araq	it's nice to be able to try out new things
22:30:44	zahary	actually, I could even reuse using as a declaration method and say that everything about using is experiental
22:30:49	zahary	using my_string as string
22:32:03	*	VarriountFamilie quit (Read error: Connection reset by peer)
22:33:36	Varriount	Araq: Do you want me to write up a list of reasons?
22:33:52	Varriount	A-la Python's PEP's?
22:34:18	Araq	yes at least
22:35:07	Araq	but even then it has 2 problems:
22:35:11	Araq	1. NIH
22:35:38	Demos__	DEBUG SYMBOLS! I have won the lottery :D
22:35:54	Araq	2. the core devs are busy implementing/fxing what they care about
22:36:44	*	xenagi quit (Quit: Leaving)
22:39:06	Varriount	Araq: Ah well.
22:52:58	Araq	zahary: please don't do "using my_string as string", that's inconsistent and so we need to change it even if the feature will be kept
22:53:22	Araq	type my_string = mimic[string] is fine
22:56:50	OrionPK	thanks varriount
22:57:17	zahary	hmm, I thought about something like this, but one problem is that it will let the user do the following error:
22:57:17	zahary	type
22:57:18	zahary	my_string = object …
22:57:18	zahary	my_string = mimic string
22:57:19	OrionPK	familiar125 is probably a regular nimrod user araq
22:57:40	OrionPK	that's the default name for irc familiar
22:58:09	Araq	OrionPK: I great when I think it's somebody new, I don't keep a list or something
22:58:11	zahary	i.e. you have to have a second type section or some strange logic code in the type section processing code
22:58:14	Araq	*greet
22:58:27	OrionPK	:-)
22:58:42	OrionPK	it was probably varriount
22:58:47	Araq	type my_string = object {.mimic: string.} ... but ok
22:58:56	zahary	it's possible to do assume my_string is string (assume being a magic with AST constraints)
22:59:03	Araq	I see the problem
22:59:31	*	zielmicha8 quit (Remote host closed the connection)
23:00:10	Araq	"using my_string as string" is not good enough though
23:00:40	zahary	what about assume X is Y?
23:00:54	Araq	doesn't stand out enough
23:01:01	zahary	these have problems on their own - sometimes you want the declaration to be public and sometimes not
23:01:21	Araq	anything that doesn't use 'mimic' is not clear enough
23:01:50	Araq	{.mimic: X = y.} for a start, I guess
23:02:31	Araq	or perhaps
23:02:48	Araq	type my_string {.mimic: string.} = object
23:02:51	zahary	I have a bit more features than simple one-to-one mapping as I mentioned.
23:02:52	zahary	you can do assume seq[char] is JavascriptByteBuffer
23:02:52	zahary	or assume MyAutoMocker is any
23:03:07	zahary	the second one is the equivalent of your original tyProxy idea
23:04:05	Araq	so use {.mimic: x = y.} but it will be strange for the docgen
23:04:51	zahary	I could do mimic my_string = string (if I go the AST constrained magic way)
23:05:10	zahary	you prefer pragma to make it more "scary"?
23:05:15	Araq	yes
23:05:22	zahary	it's fine by me
23:05:47	Araq	though the average nimrod code contains so many pragmas now that's really absurd to claim they are scary
23:06:17	Araq	so I played with {} instead of {. .}
23:06:37	Araq	which can be easily supported in the contexts they matter most
23:07:07	Araq	but now I like them too much ...
23:07:17	zahary	:)
23:08:23	Araq	I still think we perhaps should use [. .] for generics when it's ambiguous otherwise
23:08:39	Araq	var a: foo[bar] # stays
23:08:58	Araq	f[.x, y.](a, b) # generic, can't be array access
23:09:35	zahary	[] is indeed quite ambiguous, but this is likely to break lots of code
23:09:41	OrionPK	lol
23:10:14	Araq	well it's easy to make the compiler warn about old [] usages
23:10:22	OrionPK	so is {} pragmas going to happen?
23:10:28	EXetoC	ugh
23:10:38	Araq	and then people can migrate their code easily
23:10:57	Araq	OrionPK: make a forum post and ask around
23:11:07	OrionPK	I don't like the generics one
23:11:16	OrionPK	psh
23:11:27	Araq	I think the {. .} should stay but I can be convinced to support {} too
23:11:53	Araq	OrionPK: the generics one I'm talking about are rarely used
23:12:21	OrionPK	the nim forum support polls?
23:12:39	Araq	no ... but if you're lucky some people answer
23:12:54	gradha	like the camelcase poll
23:13:24	OrionPK	heh
23:13:42	OrionPK	people don't get fired up about small things
23:13:43	Araq	zahary: in fact "nimrod pretty" works quite well, we can make it transform foo[a, b] into foo[.a, b.] automatically
23:14:08	zahary	yeah, I figured
23:14:09	OrionPK	but they might lean way or the other just not enough to say anything
23:15:03	zahary	I'm too sleepy now to contemplate the consequence of making this change within the compiler, but most people will probably not like it. is our goal to just make our life easier as compiler writers?
23:15:39	Araq	well a[b] has some problems when it comes to `[]` and auto-bind in templates
23:15:54	Araq	and [. .] mitigates this issue
23:16:33	Araq	but alright, lets keep it the way it is
23:16:44	Araq	we're optimizing for popularity these days
23:16:45	zahary	OTOH, I've advocated the user of typedesc params instead of explicitly supplied generic params, so my opinion is that [. .] should not be present anywhere in the code
23:16:53	zahary	advocated the use
23:16:55	Araq	I know
23:18:59	zahary	I'll be off to bed, good night
23:19:10	gradha	good night, honey badgers
23:19:18	*	gradha quit (Quit: bbl, need to watch http://www.youtube.com/watch?v=dEf4PJZXBxA again)
23:19:27	Araq	alright good night
23:21:28	Demos__	is devel fixed on windows?
23:21:45	Araq	I think so
23:22:41	EXetoC	zahary: and it has no impact on either inference or default types?
23:23:37	EXetoC	the use of typedesc instead of []
23:23:45	EXetoC	BitPuffin: you don't like that, right?
23:24:11	Araq	EXetoC: sure it has some impact [] are optional for a start
23:24:33	Araq	he wants to use typedesc when it's mandatory
23:25:30	*	darkf joined #nimrod
23:34:50	Demos__	OK it built
23:34:59	Demos__	strangely it would not build using icc
23:40:53	*	familiar125 joined #nimrod
23:42:26	EXetoC	mandatory how? it sounded to me like he wanted only the typedesc approach to be used. if so, then I think this should be allowed: "proc f(x: var T, y: T, T: typedesc = int)". currently it's not because of the order
23:42:30	*	familiar125 quit (Remote host closed the connection)
23:42:53	Araq	no that's not what he meant afaict
23:42:54	*	dom_familiar joined #nimrod
23:43:08	Araq	hi dom_familiar
23:43:25	dom_familiar	welcome Araq
23:43:36	dom_familiar	Well... this doesn't work at all OrionPK
23:43:51	Araq	ha I knew it was you
23:43:57	Araq	do you know why?
23:44:06	Araq	cause your nick starts with 'dom' ...
23:45:36	*	dom_familiar quit (Remote host closed the connection)
23:45:47	EXetoC	alright nevermind
23:46:07	dom96	Araq: Well done.
23:51:51	Demos__	goddamn, I dont like c# as a language but the general "design for tooling" objective produced some amazing results
23:52:24	Demos__	I renamed a file and VS was like "ima just change all the references to what looks like the class that file is named after"
23:58:08	Araq	BitPuffin: still not sleeping?
23:59:43	OrionPK	whats up dom_familiar
23:59:49	OrionPK	what doesnt work