<<25-08-2012>>

---

01:26:48	*	XAMPP_ quit (Quit: Leaving)
01:27:35	*	XAMPP joined #nimrod
01:45:55	*	q66 quit (Quit: Quit)
02:05:08	*	reactormonk joined #nimrod
04:36:27	*	XAMPP quit (Ping timeout: 276 seconds)
04:36:54	*	XAMPP joined #nimrod
05:24:33	*	XAMPP quit (Ping timeout: 276 seconds)
05:25:46	*	XAMPP joined #nimrod
05:25:46	*	XAMPP quit (Changing host)
05:25:46	*	XAMPP joined #nimrod
06:14:59	*	Trix[a]r_za is now known as Trixar_za
07:59:04	Araq	fowl: it's expected behaviour ;-)
07:59:28	Araq	you're on your own if you overwrite the terminating zero
08:05:36	*	Trixar_za is now known as Trix[a]r_za
08:51:54	dom96	hello
08:54:11	Araq	hi dom96
09:58:44	dom96	fowl: Looks like I will be using your sfml bindings for Ludum Dare ;)
10:03:42	fowl	dom96: cool, you might like my input helpers https://github.com/fowlmouth/gotterdammerungarschloch/blob/master/input.nim they handle most meaningful events
10:05:25	fowl	csfml is missing a renderwindow_getmousepos() so you have to track the position from the mousemove event
10:06:01	dom96	What license is that code? :P
10:07:02	fowl	smfl has an open license as long as you dont impersonate laurent gomila
10:07:16	dom96	Yeah, but I mean your code.
10:07:27	dom96	In particular that input.nim file.
10:08:47	fowl	hmm i dunno
10:09:28	fowl	i declare it, public domain
10:09:43	dom96	cool
10:09:49	dom96	thanks :D
10:10:14	dom96	The theme for this LD is Evolution btw
10:13:36	*	dom96 doesn't have a concrete idea yet
10:13:38	*	q66 joined #nimrod
10:15:35	fowl	i would never be able to decide on something in time to start coding it
10:21:57	dom96	hrm, I think i'll be fine without input.nim to be honest.
10:34:31	fowl	word
10:34:39	fowl	g'night
11:00:11	dom96	good night fowl
12:36:31	*	Trix[a]r_za is now known as Trixar_za
12:37:15	Trixar_za	dom96: When's the Ludum Dare?
12:37:25	dom96	It started already
12:37:35	Trixar_za	Er, crap?
12:37:46	dom96	Are you competing heh?
12:37:57	Trixar_za	http://eris.oppresses.us/tmp/idontremember.jpg
12:37:59	dom96	As of right now all I have is a window which displays the FPS :P
12:38:00	Trixar_za	I wanted to
12:39:25	dom96	you still should
12:39:27	dom96	Theme is evolution
12:41:09	Trixar_za	I've been looking for the one game for years. Could never remember it's name
12:42:09	Trixar_za	But it was very simple. You gave each shape 3 actions to deal with situations - as in turn left to avoid preditors, turn right towards possible mates, stuff like that
12:42:50	Trixar_za	That was my first experience of a teachable AI
12:42:54	Trixar_za	even if it was simplistic
12:43:03	Trixar_za	I wonder how hard it would be to write a cheap ass version of that
13:00:11	dom96	I have an idea, but it seems overfly complex as always
13:00:44	dom96	*ovely
13:00:47	dom96	*overly
13:13:36	Trixar_za	It's evolution right? So it can be anything
13:13:41	Trixar_za	Like Conway's Life
13:13:46	dom96	It can't be anything.
13:14:02	dom96	best that it has some sort of connection to 'evolution'
13:14:16	dom96	of course that could just be evolving your character by buying upgrades or whatever
13:14:16	Trixar_za	I meant in the form of one thing evolving to another
13:14:32	dom96	It seems most people are thinking of cells
13:14:46	dom96	And are making some sort of bacteria fighter game
13:15:06	Trixar_za	http://en.wikipedia.org/wiki/Conway%27s_Game_of_Life
13:15:11	Trixar_za	:P
13:15:29	Trixar_za	I'm always fascinated by them
13:15:55	Trixar_za	The only real 'good' use for math imho
13:16:19	dom96	Trixar_za: Are you competing then?
13:16:31	Trixar_za	I'll try
13:16:42	Trixar_za	I always wanted to write my own copy of Conway's Game of Life
13:17:16	dom96	What language will you use? :P
13:17:53	Trixar_za	I'll try Nimrod, but I'll probably fall back to Python :P
13:18:50	dom96	Don't you dare fall back :P
13:25:38	dom96	Araq: Did you fix Ctrl + /?
13:25:47	dom96	or is it still broken in Aporia?
13:26:10	Araq	still broken, but I don't mind
13:26:21	Araq	I always use 'when false' + indentation anyway
13:27:29	dom96	Araq: With Ctrl + / you can select the code you want to comment and it will comment out multiple lines for you.
13:27:36	dom96	(Using discard """ however)
13:27:48	Araq	I know, there is also a button for it
13:27:52	Araq	so I tried that :P
13:29:03	dom96	I see
13:31:01	Trixar_za	Damn you dom96
13:31:08	Trixar_za	Now you got me reading about Alife sims
13:31:21	dom96	hrm?
13:32:39	Trixar_za	Most of them are based off Core Wars or a mix of Conway's Game of Life and Core wars
13:37:33	Araq	ugh
13:38:26	Araq	I hate it when I make some feature cleaner only to discover that the old solution was easier and worked better
13:48:48	ccssnet	Araq: true, rarely happens to me though
13:49:02	ccssnet	probably because i do functional programming
14:07:39	*	Trixar_za is now known as Trix[a]r_za
15:12:34	dom96	argh
15:12:41	dom96	I think I'm going to give up.
15:13:21	fowl	http://www.youtube.com/watch?v=PW3RqGgEkHs
15:13:27	fowl	whys that
15:13:58	dom96	Every idea I have feels like too much work :P
15:25:37	fowl	hrm so i was going to write my events out in json like ["spawn-obj", ["Solar Mold", ["random-pos"]]], ["if-chance", [10, ["spawn-obj", ["Mold Sprite"]]]] then parse them and look up all the objects needed and store them for later but i have no idea how to store this type of thing :/
15:27:59	dom96	fowl: I'm curious how your game looks like, can you provide a screenshot?
15:28:18	dom96	fowl: Nimrod has a marshal module
15:28:25	dom96	I'm not sure if that's what you're looking for.
15:28:34	dom96	It actually uses JSON as the serialization format
15:29:35	fowl	dom96: http://minus.com/lytlUi08DbGft still have debug texts drawing and not much going on yet :p
15:29:54	dom96	lol
15:29:54	fowl	the lobby: http://minus.com/lPfr1Dd5YTdv5
15:29:56	dom96	'poo'
15:30:17	dom96	looks pretty cool
16:13:44	Araq	so, dom96, how's progress?
16:13:55	dom96	I gave up
16:14:05	Araq	:'-(
16:15:21	dom96	The more I think about what I want the game to be the more difficult it seems.
16:15:52	*	shevy quit (Ping timeout: 245 seconds)
16:20:01	dom96	Araq: So why doesn't Ctrl + / work for you?
16:20:19	Araq	because keys never work in gtk
16:20:31	Araq	they always work for you and not for me
16:22:22	fowl	works for me :)
16:22:58	dom96	Araq: I bet your WM messes with it.
16:23:09	Araq	don't worry about it
16:23:19	Araq	as I said, I don't care about that feature anyway
16:23:31	dom96	Yes, but it might not work for others.
16:28:57	*	shevy2 joined #nimrod
16:30:01	fowl	Hey I thought there was a clear(seq[T]) function
16:30:03	fowl	was i high
16:31:38	dom96	Just assign @[] to it?
16:39:11	*	q66 quit (Quit: Quit)
16:40:19	fowl	dom96: http://www.youtube.com/watch?v=wXzg0D-cKds :)
16:42:00	dom96	lol wtf
16:43:53	*	q66 joined #nimrod
16:49:58	Araq	fowl: use setlen(x, 0)
17:03:51	fowl	what do you think about this https://gist.github.com/3467925
17:09:19	Araq	fowl: well?
17:09:39	Araq	I fail to see your point; still a bug in stringstream?
17:10:23	fowl	yea
17:11:49	Araq	do you use the fixed system.clamp?
17:12:39	fowl	i didnt see a difference in it?
17:12:46	Araq	I fixed it :P
17:13:03	Araq	the < and > were the other way round
17:13:10	fowl	ah
17:13:19	fowl	i think it should be <= and >=
17:13:29	Araq	why?
17:14:07	fowl	<= and >= handle more edge cases
17:14:37	Araq	if x < a: return a
17:14:41	Araq	seems reasonable
17:14:50	fowl	in this case clamp(0, 0, -1) is being called and even though the max is -1 i assumed it would hit on the min 0
17:15:32	fowl	but i assume its like min(a, max(b, x)) its a bit different
17:15:43	Araq	hrm I dunno
17:15:53	Araq	maybe you're right
17:16:20	Araq	if b < a then the range is empty and 'a' should be returned ... perhaps
17:17:15	Araq	clamp(x, low(s), high(s)) should return -1 aka high(s) if 's' is empty
17:17:25	fowl	might as well use min(max()) it does everything that logic does
17:17:49	Araq	no, min(max()) is a cludge
17:18:06	Araq	you want to enforce it's in the interval [a,b]
17:18:25	Araq	btw I always get min and max wrong :P
17:19:15	Araq	'clamp' fits much more what I have in mind
17:21:01	fowl	then this: s.pos = clamp(pos, 0, s.data.high) should be clamp(pos, 0, s.data.len) ? you dont want to start writing on high(str) you want the next char (am i right in thinking this?)
17:22:06	Araq	you can either write to a stringstream or read from it
17:23:04	Araq	if you set the position it'd better be a valid position in the written string
17:24:10	Araq	ss.data.setlen 0 # plays with the internals
17:24:31	Araq	maybe 'pos' should be exported too so that you can do whatever you like with it?
17:24:42	fowl	i do that in my flush implementation
17:26:09	Araq	just fix it and make a pull request
17:26:21	Araq	extra points if the docs are improved too :P
17:27:27	*	shevy2 is now known as shevy
17:28:41	Araq	argh, I wish Zahary was here
17:29:05	fowl	Araq: here's a better example of it being weird; https://gist.github.com/3467925#file_test2.nim
17:29:18	Araq	my problem is easy to fix if only I would knew what the spec should look like ... :-/
17:30:08	Araq	fowl: as I said, you read and write from the stream that's why it doesn't make sense
17:30:21	Araq	it works if you only read or only write
17:30:49	Araq	as I said, improve it and make a pull request, I've better things to work out :P
17:31:02	fowl	Araq: o_O streams are ok to read and write from im showing the setposition weirdness
17:31:06	fowl	lol
17:31:08	fowl	ok
17:45:00	Araq	fowl: I thought setPosition only makes sense for reading :P
17:56:29	fowl	well yes it doesnt make sense to do that.. but i use setposition() in my flush function
20:52:52	Araq	ping Boscop
20:53:25	Boscop	pong Araq
20:53:43	Araq	I'm thinking about symbol binding rules again
20:53:47	Boscop	ah
20:54:21	Araq	and I figured that the compiler could easily do a partial type checking
20:54:44	Araq	so if we have a generic type T and an == that is concerned with T
20:55:22	Araq	it's an "open" symbol which means it will be bound at instantiation
20:55:43	Araq	and other occurances of == can be bound early
20:56:56	Araq	in fact, the compiler performs that partial type checking already for IDE support ...
20:58:53	Araq	I wonder why no programming language does it this way?
20:59:23	Araq	what am I missing?
21:10:31	Araq	Boscop: well? understood my idea?
21:11:04	Boscop	you mean == is a generic operator with type variable T?
21:12:47	Araq	no, == is overloaded for lots of types
21:13:16	Araq	but if we have current.key == key in a generic
21:13:43	Araq	this == should be "open", so that later definitions of == are considered in the instantiation process
21:14:14	Boscop	template overloading?
21:14:56	Araq	well in C++ terminology I'm talking about symbol binding rules in a template, yes
21:15:23	Boscop	D has template overloading
21:16:25	Araq	so D does it the way I'm proposing?
21:17:34	Boscop	which symbols would be bound early in your idea, and what do you mean by "and other occurances of == can be bound early"?
21:18:02	Araq	https://github.com/Araq/Nimrod/blob/master/lib/pure/collections/sets.nim
21:18:16	Araq	take a look at line 67
21:18:26	Araq	proc RawGet[A](s: TSet[A], key: A): int =
21:18:27	Araq	rawGetImpl()
21:18:48	Araq	which refers to line 52, unfortunately
21:18:54	Araq	so don't get distracted
21:19:08	Araq	line 55: if s.data[h].key == key and s.data[h].slot == seFilled:
21:19:46	Araq	the key comparison is an "open" occurance of ==
21:20:00	Boscop	so the == should be instantiated with A from RawGet?
21:20:12	Araq	yes
21:20:15	Boscop	ah
21:21:17	Araq	but 's.data[h].slot == seFilled' can be bound at definition time of the generic
21:21:21	Boscop	yeah, you can have mixin templates which are typechecked at instantiation time and place
21:21:46	Araq	yes well nimrod currently does it only at instantiation time
21:21:48	Boscop	that code would have to be treated symbolically
21:22:07	Araq	and I want it to be partially done at definition time
21:22:12	Boscop	ah
21:22:22	Boscop	so you could propagate the type variable A
21:22:30	Araq	exactly
21:22:34	Boscop	make rawGetImpl dependent on A
21:22:56	Araq	true, but just assume rawGetImpl() doesn't exist
21:23:04	Boscop	and then have the constraint (e.g. with type classes or similar) that == is defined for A
21:23:05	Araq	and RawGet has the body
21:23:33	Araq	that's true but I don't want that constraint to be explicitly written down
21:23:43	Araq	because it's redundant
21:23:56	Araq	and can get annoying
21:24:18	Araq	in this example, it would make the code clearer, I agree
21:24:44	Boscop	if you want typechecking to be done partially at def time, you'd need to know that == is defined for A, whatever A will be, otherwise you end up the same as C++ and have to typecheck the whole template body at each instantiation
21:25:01	Boscop	not to mention hard to read error messages
21:25:23	Boscop	of course you don't have to write the constraint down if the compiler can infer it
21:25:49	Boscop	but then the template has to know which types are variables
21:26:02	Araq	well yes, but I wonder why C++ doesn't do that partial type checking I'm talking about
21:26:03	Boscop	so it has to have a type variable where A will be plugged in
21:26:26	Boscop	because of template specialization
21:26:34	Boscop	it has to instantiate before it knows the body
21:26:45	Boscop	and it can`t infer constraints
21:27:36	Boscop	so it has to typecheck the body on each instantiation
21:27:48	Boscop	which is one reason for its slow compile-times
21:28:18	Araq	yeah, yeah, well I'm repeating C++'s mistakes and am aware of it ;-)
21:29:06	Boscop	yeah, I wouldn't do it the c++ way
21:29:33	Boscop	I would infer constraints, but for that the variable types in a template need to be known
21:29:55	Araq	the variable types in a template are known in nimrod
21:30:09	Boscop	because that's what the binding of the operators etc depends on
21:30:09	Araq	unless I'm missing something :-)
21:30:26	Araq	true
21:30:40	Boscop	hm, but in rawGetImpl the A is not known, but you want == to be chosen for the A in the caller scope
21:31:05	Araq	you got distracted by the rawGetImpl ;-)
21:31:23	Boscop	isn`t that what you want to be able to partially typecheck?
21:31:38	Araq	ignore that, just think its body is directly in RawGet[A] ...
21:32:08	Araq	(templates are resolved at definition time in generics already)
21:32:10	Boscop	ok, and what do you want to partially typecheck?
21:32:34	Araq	s.data[h].key == key # infer == for A
21:32:54	Araq	but I don't call it inferring because that's not what the compiler does ;-)
21:33:14	Araq	instead it would fail type checking and so the '==' would be marked as an "unbound" symbol
21:33:33	Araq	and bound at instantiation time
21:34:11	Araq	so yeah the generic would be checked at instantiation time like in C++
21:34:22	Boscop	so when RawGet is instantiated, all overloads of == in the instantiation scope are considered to choose the one for the A?
21:34:36	Araq	yes exactly
21:34:59	Boscop	and what is the problem?
21:35:14	Araq	well no problem for now
21:35:22	Araq	but I wonder why C++ doesn't do it this way
21:35:31	Araq	and if there is any PL that works this way
21:37:09	Araq	I'm not really trying to solve the implicit constraints problem
21:37:09	Boscop	c++'s lookup rules are pretty complicated
21:37:30	Araq	I'm trying to come up with symbol binding rules that are "natural"
21:38:28	Boscop	often there is no one true way
21:38:55	Boscop	language design is also experimentation to some degree
21:39:03	Araq	I know :-)
21:39:12	Araq	it helps to have production code around :D
21:39:30	Araq	and see how the rules would do for that
21:39:35	Boscop	one way to try would be: the union of the set of overloads from definition scope and instantiation scope should be considered for lookup
21:39:53	Araq	well that's another point
21:39:59	Boscop	so that custom definitions can override the ones at definition scope
21:40:00	Araq	and in fact Nimrod does that
21:40:06	Boscop	ah
21:40:31	Araq	the problem is when you have helpers which are private
21:40:48	Araq	you don't want client code to accidently override them
21:41:11	Araq	you can do that currently in nimrod with a "bind" statement, but that's the wrong default
21:41:32	Araq	default should be "bind", and 'open' symbols should be marked explicitely
21:41:48	Araq	which however is too ugly in practice
21:42:03	Araq	so I implemented the wrong solution :-)
21:43:09	Araq	and now I think I found a set rules to get the best of both worlds and wonder why nobody else does it this way ;-)
21:43:16	Araq	*a set of rules
21:43:16	Boscop	or you do it with implicit args, or explicit args, e.g. the == would be passed as arg to the template
21:43:34	Boscop	and it has a default value, the == at definition scope
21:43:45	Araq	yeah but that gets verbose too
21:43:50	Boscop	yeah
21:44:37	Boscop	another solution to preventing overrides would be local definitions
21:44:48	Boscop	so that they shadow outer ones
21:44:58	Araq	that's my "bind" statement
21:45:06	Araq	proc p[T] ...
21:45:12	Boscop	and you can use an import statement or 'using' to pull in something into the local scope
21:45:14	Araq	bind myHelper, myHelper2
21:45:19	Boscop	ah
21:45:37	Araq	but as I said, it's the wrong default ;-)
21:46:32	Araq	but thinking about it, my new idea doesn't work
21:46:39	Boscop	why?
21:46:42	Araq	proc p[T](x: T) =
21:46:56	Araq	helper(x) # I want this to be bound early
21:47:19	Araq	but 'x' is of type T, so will remain "open"
21:47:30	Araq	the compiler cannot tell the difference between:
21:47:32	Araq	x == y
21:47:34	Araq	and
21:47:36	Araq	helper(x)
21:49:22	Araq	ok, it can with an even more complicated set of rules
21:49:37	Araq	but ugh
21:49:44	Boscop	the set of overloads to be considered can be different for each symbol occurring in the template body
21:50:13	Boscop	so you do 'bind helper;" but don't bind ==
21:51:00	Araq	well sure and 'bind' works this way today already in Nimrod :-)
21:51:33	Boscop	so why doesn't your idea work then?
21:51:52	Araq	my idea was to come up with more complex rules and get rid of the 'bind'
21:52:15	Araq	but now I got it that the rules are not as simple as I imagined
21:52:47	Araq	it doesn't suffice to re-use my IDE features to solve that problem :D
21:53:39	Araq	now back to C++:
21:53:59	Araq	I'm not sure that checking only at instantiation time is really that much of a problem
21:54:19	Araq	it's a natural solution if you want compile time "duck typing"
21:55:17	Araq	and if you're gonna monomorphize generics
21:55:20	Boscop	but you get slow compile-times and unreadable errors
21:55:46	Araq	slow compile-times are also related to its non existing module concept
21:56:10	Araq	and the unreadable errors are due to the complex implementations
21:56:28	Araq	the STL for instance seems to be much more complex than it needs to be
21:56:44	Araq	but it has to 'struct' everything for better type checking
21:57:06	Araq	and C++ has a very weak type system
21:57:15	Boscop	yeah
21:57:19	Araq	with conversions between int and bool, etc.
21:57:50	Araq	so I hope these problems won't be as severe for Nimrod as they are for C++
21:59:02	Boscop	yeah, the problem is you can only instantiate the right specialization of a template if you know its arguments, which might be template arguments themselves
21:59:32	Araq	and a bad compiler error message is still better than any runtime error message IMHO
22:00:15	Boscop	yeah
22:01:20	Araq	btw you can have this in Nimrod:
22:01:41	Araq	proc `==` (a, b: myType, someFlag: bool = false)
22:02:10	Araq	and then the `==` should match a constraint that doesn't consider default parameters
22:02:54	Araq	this constraint stuff needs to be permissive IMO
22:03:03	Boscop	how do you call binary operators with more than two operands in nimrod?
22:03:17	Araq	prefix style then:
22:03:24	Araq	`==`(2,3, true)
22:03:54	Araq	you can still use the infix style if the default fits
22:04:07	Boscop	ok
22:06:47	Araq	when you monorphize (correct term?) generics, they are really macros, so you may as well want to get the expressitivity of macros in return ;-)
22:12:56	Boscop	but do you have some constraints for the template arguments, like concepts or type classes?
22:14:20	Araq	there is:
22:14:34	Araq	proc p[T: array|seq|object](x: T)
22:14:47	Araq	for now and a 'compiles' magic that you can use
22:15:03	Araq	which is like D's is(typeof(X)) crap
22:15:22	Araq	but readable ;-)
22:15:38	Araq	when not compiles(x + y):
22:15:52	Araq	{.error: "generic needs to support the + operator.}
22:20:33	Boscop	D also has __traits(compiles,expr)