#nimrod logs for 11-11-2012

<<11-11-2012>>

01:05:10	*	XAMPP joined #nimrod
02:25:55	*	q66 quit (Quit: Quit)
02:41:30	Amrykid	Wasn't there an effort for nimrod to support 'compiler as a service' a while ago?
07:57:02	Araq	Amrykid: very basic support for compiler as a service exists, but it's not yet used from aporia
08:05:21	Araq	and until that is implemented I won't work on the feature as testing it via telnet is a pita ;-)
08:13:27	reactormonk	Araq, just write some tests
08:13:52	Araq	indeed I should ... looks like quite some work though :P
08:15:45	Araq	and I'm figuring out how to use my effect system to prevent deadlocks
08:15:45	Araq	looks like a simple problem ;-)
11:38:49	dom96	good morning
11:39:48	*	Vladar joined #nimrod
11:40:10	Vladar	Hi
11:40:20	dom96	hello Vladar
11:47:24	Vladar	So, is there any way to pass some comments from code to html-file generated by 'doc' or 'doc2'?
11:49:36	dom96	Yeah: ## starts a doc comment
11:50:19	Vladar	oh, probably missed that. Thanks!
11:50:31	dom96	Place it underneath a proc declaration
11:50:48	dom96	You can also document types like that.
11:51:07	Vladar	Will try it right now.
11:51:10	*	q66 joined #nimrod
11:51:23	dom96	for example take a look at the stdlib: https://github.com/Araq/Nimrod/blob/master/lib/pure/ftpclient.nim#L78
12:02:14	Araq	TEffect* = object of TObject ## base effect class; each effect should
12:02:14	Araq	## inherit from `TEffect` unless you know what
12:02:14	Araq	## you doing.
12:02:14	Araq	FTime* = object of TEffect ## Time effect.
12:02:14	Araq	FIO* = object of TEffect ## IO effect.
12:02:43	Araq	FRead* = object of FIO ## Effect describing a read IO operation.
12:02:44	Araq	FWrite* = object of FIO ## Effect describing a write IO operation.
12:03:32	dom96	Time effect?
12:03:34	Araq	FExec* = object of FIO ## Effect describing an executing IO operation.
12:03:34	Araq	that's what I've got so far ... any better names?
12:03:45	dom96	what is that?
12:03:50	Araq	dependent on some timing issue
12:04:01	Araq	for now I can only image "uses os.sleep"
12:04:14	Araq	*imagine
12:04:16	dom96	hrm, so wouldn't all of the times module be that? :P
12:04:39	Araq	'-' for TTime wouldn't :P
12:05:01	dom96	Does that mean that the output changes based on the current time, or does it mean that the output could be returned at an arbitrary time?
12:05:12	dom96	I guess both?
12:05:36	Araq	it means that a proc p uses os.sleep ;-)
12:06:05	Araq	or perhaps that it makes some time guarantees like "does not block for more than 3ms"
12:06:32	dom96	ok, so it does not apply to the times module
12:06:41	dom96	I guess the times module is FIO
12:06:55	Araq	no, the times module is FTime for the most part
12:07:04	dom96	how is it?
12:07:13	dom96	it doesn't use os.sleep
12:07:18	Araq	effects propagate so it's important that the built-ins are properly annotated
12:07:47	Araq	sure it doesn't use os.sleep, but it does deal with time
12:08:41	dom96	Yeah, but your definition of what FTime means does not suggest that it applies to the times module.
12:08:41	Araq	'-' for TTime is not FTime as it's a pure function
12:08:59	Araq	most other functions in times.nim are not pure
12:10:31	dom96	Why do we need a separate Time effect anyway?
12:10:53	dom96	sleep seems like an IO thing
12:11:02	dom96	and so does retrieving the current time
12:12:31	Araq	I want to avoid haskell's IO issues
12:12:49	dom96	what issues are those?
12:12:56	Araq	for instance, it's perfectly fine to do 'getenv' in a transaction and to retry it
12:13:10	Araq	yet haskell's type system would prevent it
12:13:24	dom96	what if the environment changes?
12:13:57	dom96	FTime suggests that the proc is timed, but the output will always be the same.
12:13:57	Araq	good point, but it could have changed before the first call to 'getenv' too
12:14:05	dom96	And in that case it does not apply to the times module :P
12:14:19	Araq	hm, a very good point
12:15:08	Araq	FTime is a very subtle effect
12:15:21	Araq	for instance it doesn't prevent evaluation at compile-time
12:15:32	Araq	and yet a call to os.sleep MUST not be optimized away
12:16:11	Araq	even if its argument is known at compile-time
12:16:53	dom96	why would you use os.sleep at compile-time?
12:17:37	Araq	the point is this should work:
12:17:43	Araq	proc p(): int =
12:17:49	Araq	os.sleep(15)
12:17:54	Araq	result = 12
12:17:57	Araq	const x = p()
12:18:22	Araq	though maybe that really shouldn't happen in practice
12:18:34	Araq	and it's bad code to begin with ... I dunno
12:20:05	dom96	Sorry, I still don't get why...
12:20:24	dom96	also I don't get what you mean about 'getenv'
12:20:56	Araq	well everything that's happening in a transaction needs to be able to be rolled back
12:20:58	dom96	'getenv' should have the FIO effect.
12:21:25	Araq	this means FWriteIO is not allowed in a transaction
12:21:39	Araq	but IMO FReadIO is fine in a transaction
12:21:52	Araq	getenv should be FReadIO
12:22:24	dom96	ahh.
12:22:29	dom96	That makes sense, yes.
12:22:36	dom96	Haskell doesn't separate those.
12:23:06	dom96	But I'm not sure it would be of benefit to it, if it did separate it.
12:28:43	Araq	well one of ghc's authors thinks so :P
12:29:55	Araq	but as usual I lost the link ;-)
12:31:24	Araq	it's however common sense: IO is simply a too wide field
12:31:56	Araq	for instance for my code the difference between "logging IO" and "writes to database" is big
12:33:52	Araq	in fact, for logging I might fake that it performs no IO ;-)
13:42:17	zahary	Araq, if you are still uncertain about the naming scheme, an alternative way is to organize it arround the verbs: TReads, TWrites
13:49:23	Araq	zahary: but that's confusing as 'reads' and 'writes' are further effect pragmas
13:49:57	Araq	but I have to go, see you later
15:13:25	*	ozkriff joined #nimrod
15:14:07	*	ozkriff_ joined #nimrod
15:14:55	*	ozkriff quit (Remote host closed the connection)
15:15:23	*	ozkriff_ quit (Client Quit)
15:17:53	*	ozkriff joined #nimrod
15:18:09	ozkriff	hi
15:22:36	Araq	welcome ozkriff
15:24:46	Araq	er zahary, what do we need quasiquote for? what's the difference to getast?
15:25:06	zahary	haven't we discussed this like gazillion times?
15:25:20	Araq	no?
15:25:57	zahary	are you sure? do you see the docs?
15:26:30	Araq	yes I do, but getast + a template seems to solve the very same problem?
15:27:10	zahary	yes, and this integrates it better into the language - we really have talked about AST quoting many times before
15:28:25	Araq	why did you start with getast then? :P
15:29:05	zahary	getAst is still useful if you want the template to be usable by itself
15:29:18	Araq	btw we also talked about using \ for getting back the operator
15:29:51	zahary	and about [. .] as quoting operator
15:29:51	Araq	and the lexing rules now reflect that
15:29:59	zahary	these details are easy to change
15:30:05	Araq	aha
15:30:11	zahary	in fact, the current grammar doesn't work very well with the feature
15:30:12	Araq	so it's no "final" patch
15:30:17	zahary	var @a = 10 # not possible
15:30:29	Araq	yeah I know
15:30:30	zahary	var x = `foo(10)` # not possible
15:35:52	*	ozkriff quit (Ping timeout: 245 seconds)
15:39:06	*	ozkriff joined #nimrod
15:40:42	ozkriff	I can't understand from docs how to use dynamic arrays
15:40:49	ozkriff	How can i set array length in runtime?
15:40:55	ozkriff	Only adding elements one by one to sequence?
15:41:14	ozkriff	sorry for my english(
15:42:26	Araq	ozkriff: there is also 'setLen' and 'newSeq'
15:43:19	Araq	zahary: that's one reason why I don't like 'quote'; I don't find it particularly useful with the current syntax
15:43:55	Araq	the next logical step would be to come up with a decent macro API; currently we only have the minimum that's necessary
15:43:57	zahary	well, I was going to talk to you about changing the grammar a little bit
15:44:20	Araq	problem is: I don't like how it reads in Lisp either ;-)
15:44:45	zahary	and getAst is better? just open my unittest module
15:44:49	Araq	the declarative approach only goes so far ... and then you want to construct it imperatively
15:45:39	zahary	no, no - I almost never want to construct it imperatively - I may want to glue some quoted bits together or to edit some subnode after quoting
15:46:05	zahary	which are both possible with getAst and quote
15:46:13	Araq	I'm thinking of loops
15:47:12	zahary	IMO the unittest module is a very good example for what's needed and what's useful
15:47:24	Araq	I'm looking at it (again)
15:48:14	Araq	macro check uses 2 accumulators
15:48:28	Araq	'argsAsgns' and 'argsPrintOuts'
15:48:46	zahary	but notice how I build everything out of smaller quoted pieces of code like asgn and print
15:49:04	zahary	if I had to build this with newNimNode:
15:49:05	zahary	when compiles(string($value)):
15:49:05	zahary	checkpoint(name & " was " & $value)
15:49:07	Araq	yeah but I'm arguing that's an API problem
15:52:22	Araq	I can't follow, what do you mean?
15:52:43	Araq	oh
15:52:53	zahary	I'm missing your point too - getAst/quote is exactly about mixing imperative code with "declarative" bits
15:53:55	Araq	argsPrintOuts.add getAst(print(argStr, arg))
15:54:16	ozkriff	Araq: thanks! And one little question more: what is nimrod-way for: sscanf(buffer, "v %f %f %f", &v->x, &v->y, &v->z)
15:54:16	Araq	ok, how exactly does the new built-in improve that?
15:55:34	Araq	ozkriff: that's much harder ;-)
15:55:50	zahary	it's like asking how closures improve on global function with a userData cookie param
15:56:22	zahary	it makes the code smaller, removes some boilerplate, puts related things closer in the code, etc
15:56:32	Araq	it's not; my point is: I can easily parse it
15:56:49	Araq	the way it is now especially because there are no quoting rules to consider
15:57:08	Araq	I get that you want the 'print' definition inline
15:57:41	Araq	and indeed I would prefer that too, but I dislike quoting
15:58:13	zahary	well, what's the alternative?
15:58:23	Araq	an anon template?
15:58:36	Araq	not sure yet ;-)
16:00:31	zahary	it's possible to make even more like closures - identifiers from the outer scope are automatically intercepted
16:00:37	Araq	the nice thing about templates is that you have named params which avoid the quoting
16:01:03	zahary	but then, you often need to interpolate a bit more compilated expression `foo(bar, baz)`
16:02:13	Araq	'quote' requires special semchecking which means it's actually more of a keyword
16:02:31	Araq	and indeed it builds an anon template and calls getast on it
16:02:42	zahary	not really - once we fix the lazy semantizing in overloading it will work fine
16:02:58	Araq	hrm that may be true
16:04:33	Araq	actually 'template' was supposed to be a quasiquote, at least within macros; but then I never got around it and template proved to be very useful for '!='
16:05:06	Araq	and it would be confusing if 'template' changed its semantics slightly in a macro body
16:09:06	Araq	ozkriff: you need to create some helper proc that calls strutils.parseFloat or parseutils.parseFloat
16:09:44	Araq	or if you're very lazy, you can easily wrap C's sscanf
16:15:52	*	ozkriff quit (Ping timeout: 245 seconds)
16:24:49	Araq	result = (template (Exp, lineInfoLit, expLit): stmt =
16:24:50	Araq	if not Exp:
16:24:52	Araq	checkpoint(lineInfoLit & ": Check failed: " & expLit)
16:24:54	Araq	fail())(checked, checked.lineinfo, checked.toStrLit)
16:25:21	Araq	hrm and you suggest closure-like capturing
16:25:28	*	Araq likes it
16:26:46	zahary	it's relatively easy to do, but I see 2 potential problems with it
16:27:41	zahary	1) as I said, you still need a way to interpolate more complex expressions (you either need quoting operator of some kind or you need to assign them to a named variable in the outer scope)
16:28:30	zahary	2) it will lead to silent errors - typos will be turned into nkIdents, mistyped variabled will be tyrned into nkIdents, etc
16:28:36	Araq	true but is (1) really common?
16:29:29	Araq	and how would 'quote' prevent that?
16:30:01	zahary	out of the 5 getAst calls in unittest, 4 use non-triavial exprsesions
16:30:11	zahary	that was about 1)
16:30:34	zahary	about 2), if you have quoting operator, the expression inside it must be a valid PNimrodNode - otherwise you get an error
16:31:12	Araq	I can't follow, every getAst call that I'm looking at does some easy substitutions
16:32:02	zahary	it's possible to interpolate things like ex.lineinfo too, but the rules become a bit more complicated
16:33:51	Araq	you can always pass ex.lineinfo to the anon template just like it's now
16:33:51	zahary	to clarify: how do you know I didn't mean `ex`.lineinfo?
16:34:32	zahary	ah, you are talking about your anon template alternative
16:34:43	zahary	well, it's more verbose for starters
16:34:55	Araq	is it?
16:36:37	zahary	well, I wrote a param list, had to think about names for the params, used some funny javascript-like parentheses
16:36:46	zahary	do you dislike quoting that much?
16:37:19	Araq	dunno have to see more real world examples ;-)
16:37:56	Araq	but I surely dislike an ever growing list of magics which need special semchecking
16:38:15	zahary	I did it only for performance reasons btw
16:38:23	zahary	it is possible to be defined as a normal macro
16:38:38	Araq	true
16:38:41	Araq	good point
16:38:42	zahary	and yet why did you implement static as a keyword :P
16:38:56	Araq	it felt right :P
16:39:33	Araq	I also want a 'static case'
16:39:44	Araq	and maybe 'when' should become 'static if'
16:39:54	zahary	if static(foo)
16:40:03	zahary	that's how it was supposed to be in my hypolang
16:40:41	zahary	if foo is const already, you can omit the static part
16:57:49	Araq	compile-time still seems an unnatural default :P
16:59:15	Araq	brb
17:36:05	*	ekselkiu joined #nimrod
17:52:13	Araq	back
18:02:25	*	XAMPP quit (Ping timeout: 246 seconds)
18:04:25	*	Guest67212 joined #nimrod
18:19:56	Amrykid	Araq, is there any documentation for the caas support?
18:21:39	Araq	'nimrod serve' starts it
18:21:58	Araq	and then you're supposed to give the socket a command line to tell nimrod what to do
18:22:16	*	Guest67212 is now known as XAMPP
18:22:23	*	XAMPP quit (Changing host)
18:22:23	*	XAMPP joined #nimrod
18:22:24	Araq	and no, there is no documentation
18:22:35	Amrykid	then I telnet into it, im assuming?
18:22:52	Araq	for testing, yes
18:23:22	Amrykid	ok, thank you
18:25:01	Araq	fyi: http://spin.atomicobject.com/2012/11/11/unifying-programming-and-math-the-dependent-type-revolution/#more-90305
18:44:12	Vladar	Nimgame v0.3 is released: http://forum.nimrod-code.org/t/107
18:47:08	Araq	Vladar: cool, does it run on linux?
18:48:07	Vladar	Araq: Of course, I making it on Linux.
19:20:06	q66	Vladar, got screenshots?
19:31:12	dom96	hrm, seems the latest compiler cannot compile nimgame. I get linker errors. "Multiple definition of ..." errors,
19:31:55	dom96	Vladar: Looking at the documentation it is looking quite nice and powerful. Nice job!
19:32:08	Vladar	q66: not much to look at: http://ge.tt/19mZfPR/v/0 you can see screen scaling and buttons.
19:33:54	Vladar	dom96: try 0.9.0 version. I'm also get those "Multiple definition" errors on latest builds. Don't know what changed in compiler lately to cause it.
19:34:53	*	Trix[a]r_za is now known as Trixar_za
19:38:46	dom96	Vladar: Awesome, it works with 0.9.0 :)
19:39:15	dom96	Buttons are a bit weird heh. Shouldn't pressing 'Exit' once quit?
19:39:41	Vladar	It should.
19:41:02	dom96	btw does it not support mp3s?
19:41:52	dom96	it kind of seems like the buttons are toggle-buttons.
19:42:38	Vladar	dom96: it depends on smpeg library. I tried windows version under wine and it played _some_ of mp3's with this smpeg.dll.
19:42:52	dom96	I see
19:43:02	Vladar	dom96: two of buttons are toggle ones.
19:43:18	Vladar	'shuffle' and 'next'.
19:43:58	q66	Vladar, interesting, how about OS X support :)
19:45:01	dom96	Vladar: When I start it and immediately press the 'play' button it seems to get stuck in the toggled position.
19:45:02	Trixar_za	Oh sweet
19:45:07	Vladar	q66: Depends on Nimrod and SDL =)
19:45:18	q66	SDL works just fine :P
19:45:38	q66	Nimrod should too
19:45:48	q66	so I guess there is no platform specific code in the thing?
19:46:18	Vladar	dom96: maybe it just freezes? It's win version, right?
19:46:29	dom96	Vladar: no, linux.
19:46:41	dom96	Vladar: That's quite possible.
19:47:06	dom96	However when I hover over the button it changes.
19:47:08	Vladar	q66: no, it is just frontend to SDL with some game specific classes.
19:47:47	Vladar	dom96: any console output errors? Also, might be some libs missing.
19:47:59	dom96	Vladar: No, nothing in the console.
19:48:05	dom96	Vladar: What libs are required?
19:48:57	dom96	Vladar: It seems I can reproduce it every time if I give focus to another application, then click on the play button.
19:49:15	dom96	Then clicking anywhere in the app makes the play button untoggle.
19:49:30	Vladar	oh, I will test it.
19:50:14	dom96	I like the songs, where did you get them? :)
19:50:15	Vladar	Thanks. Probably, something in button logic.
19:50:25	dom96	no problem.
19:51:48	dom96	btw the CPU usage of it seems to be pretty high, any ideas why?
19:53:30	Vladar	dom96: what songs? Isn't 'music' folder empty?
19:54:07	dom96	Vladar: nope, heh. Maybe you included them by mistake. I downloaded the source + demo.
19:54:40	Vladar	dom96: screen updates? Press 'F' and see how much FPS you get.
19:55:00	dom96	~200
19:55:30	Vladar	dom96: so here is your cpu usage, I guess.
19:56:28	Araq	Vladar: I'll look into the linker errors, should be easy to fix
19:56:48	Vladar	dom96: yeah, this is songs I tested it on. From "keygens music pack".
19:56:53	Araq	once I got the dreaded tags feature to work :-/
19:57:06	dom96	Vladar: haha. Nice.
19:57:18	Vladar	Araq: Thanks =)
19:58:29	dom96	Vladar: I'm guessing a sleep() in your infinite loop in ``start`` would lower the CPU usage.
19:58:58	dom96	But lower the FPS too :\
19:59:39	Vladar	I think, I'll add maxFPS option later.
20:01:20	Vladar	And come on, all games eating a lot of CPU =)
20:01:41	dom96	I guess, heh. Well great job :)
20:01:58	Araq	sleep is a good idea in a while true loop
20:25:36	Vladar	Bye.
20:25:51	*	Vladar quit (Quit: Leaving)
20:34:38	*	Trixar_za is now known as Trix[a]r_za
21:46:18	*	Trix[a]r_za is now known as Trixar_za
22:23:19	*	Boscop quit (Disconnected by services)
22:23:21	*	Boscop joined #nimrod
22:39:14	*	ekselkiu quit (Ping timeout: 260 seconds)
22:45:50	fowl	eh vladar comes back tell him to use TFPSmanager and framerateDelay() in sdl_gfx
22:47:18	Araq	tell him on the forum ;-)
23:41:09	fowl	i get a bunch of errors from gcc trying to compile his demo: https://gist.github.com/b03cbbfe67c20e0004cc
23:54:57	Araq	fowl: seems to be a recently introduces compiler bug
23:55:06	Araq	haven't looked at it
23:55:19	Araq	*introduced