<< 26-07-2013 >>

---

00:00:17	BitPuffin	EXetoC: huh?
00:00:29	*	mario-goulart left #nimrod ("ERC Version 5.3 (IRC client for Emacs)")
00:01:10	EXetoC	I think you did "type x = y", which is an alias
00:01:28	BitPuffin	hm
00:01:29	BitPuffin	well
00:01:35	BitPuffin	that's creating a type
00:01:43	BitPuffin	I'm trying to alias a proc
00:02:05	BitPuffin	alias '*.` = dot
00:02:06	BitPuffin	etc
00:06:31	EXetoC	I guess not
00:08:22	BitPuffin	hrm
00:10:17	EXetoC	like I said, you can use templates etc, but then you might as well just write a wrapper function
00:11:19	BitPuffin	EXetoC: pretty much what I'm doing
00:11:30	BitPuffin	really strange that you can't just create a symbol alia
00:11:31	BitPuffin	s
00:11:38	BitPuffin	gotta talk to Araq about that :P
00:13:52	BitPuffin	EXetoC: is `*x` a valid operator name, or does it has to be special characters
00:16:08	BitPuffin	has to be special oki
00:16:09	EXetoC	I think you need to include ` at the call site then
00:19:03	EXetoC	so it's not very useful in that case
00:19:07	*	EXetoC quit (Quit: WeeChat 0.4.1)
00:34:50	BitPuffin	I THINK, vectors are feature complete now
00:42:45	*	BitPuffin quit (Ping timeout: 264 seconds)
01:03:45	*	Trix[a]r_za joined #nimrod
01:19:30	*	DAddYE quit (Remote host closed the connection)
01:20:06	*	DAddYE joined #nimrod
01:24:21	*	DAddYE quit (Ping timeout: 248 seconds)
01:39:01	*	q66 quit (Quit: Leaving)
02:01:36	*	mario-goulart joined #nimrod
02:12:38	*	Associat0r joined #nimrod
02:21:15	*	DAddYE joined #nimrod
02:27:58	*	DAddYE quit (Ping timeout: 276 seconds)
03:24:18	*	DAddYE joined #nimrod
03:31:18	*	DAddYE quit (Ping timeout: 264 seconds)
04:13:21	*	OrionPK quit (Quit: Leaving)
04:27:36	*	DAddYE joined #nimrod
04:34:35	*	DAddYE quit (Ping timeout: 260 seconds)
05:26:25	*	DAddYE joined #nimrod
07:05:04	*	Endy joined #nimrod
08:16:11	*	DAddYE quit (Remote host closed the connection)
08:16:20	*	DAddYE joined #nimrod
08:16:26	*	DAddYE quit (Remote host closed the connection)
08:16:57	*	DAddYE joined #nimrod
08:21:23	*	DAddYE quit (Ping timeout: 260 seconds)
08:26:23	*	erlnoob joined #nimrod
09:17:38	*	DAddYE joined #nimrod
09:24:36	*	DAddYE quit (Ping timeout: 256 seconds)
09:46:57	*	BitPuffin joined #nimrod
09:48:06	BitPuffin	oihoi!
09:48:11	BitPuffin	sup errbody?
09:52:09	BitPuffin	Araq: is there a pretty way to create an alias?
09:59:50	Araq	BitPuffin: only via a template
10:00:13	*	Araq is no fan of aliases
10:00:47	BitPuffin	Araq: ah I see, well that's a shame
10:01:11	BitPuffin	Araq: it would be nice to just be able to do something like `alias bar = foo`
10:07:26	BitPuffin	Araq: when doing it with a template, do I still need to have type parameters?
10:08:02	Araq	I think so unless it's an immediate template
10:08:17	BitPuffin	okay
10:08:34	BitPuffin	well, I think I basically only need to change the keywords from proc to template then
10:08:47	BitPuffin	unless the args has to be of type expr
10:13:18	BitPuffin	Araq: do you think it would be necessary to put parenthesis around stuff like: a + (-b) in a template?
10:13:33	Araq	lol no
10:13:43	Araq	templates are not C's shitty #defines
10:13:54	BitPuffin	Araq: thank jesus
10:17:01	*	EXetoC joined #nimrod
10:17:58	NimBot	Araq/Nimrod master ce71477 Araq [+0 ±2 -0]: new vm: can execute simple programs
10:18:06	Araq	bye
10:20:14	BitPuffin	see you!
10:20:24	BitPuffin	o/ EXetoC
10:21:00	*	DAddYE joined #nimrod
10:27:07	*	DAddYE quit (Ping timeout: 246 seconds)
10:29:56	EXetoC	lo
10:30:30	BitPuffin	EXetoC: can you think of any operation that I haven't added to the vectors yet?
10:30:56	BitPuffin	https://github.com/BitPuffin/linagl/blob/master/src/linagl/vector.nim
10:33:31	BitPuffin	hmm
10:33:40	BitPuffin	I guess - could be made more efficient
10:34:00	BitPuffin	it currently goes through the vector twice :/
10:34:17	*	q66 joined #nimrod
10:35:24	BitPuffin	although I'm just assuming
10:35:45	EXetoC	twice?
10:35:55	EXetoC	those are the essentials I guess, but I'm no expert
10:36:19	EXetoC	I usually look at OGRE's implementation, which has some other stuff as well
10:36:43	*	[1]Endy joined #nimrod
10:36:45	BitPuffin	EXetoC: well I made it so that - is a + (-b)
10:37:02	BitPuffin	EXetoC: + goes through the vector once, and -b goes through the vector once too, to negate it
10:37:54	EXetoC	right
10:38:35	BitPuffin	so I changed that now
10:38:40	BitPuffin	hrm
10:38:56	BitPuffin	does one really need all the squaredMagnitude and all that jazz
10:40:37	*	Endy quit (Ping timeout: 248 seconds)
10:40:38	*	[1]Endy is now known as Endy
10:43:49	EXetoC	rarely
10:46:02	BitPuffin	I'll skip that for now
10:46:19	BitPuffin	now I wonder how I should do with matrices
10:46:24	BitPuffin	opengl stores them column major
10:47:11	BitPuffin	maybe I should just make it so that it's stored that way
10:53:16	BitPuffin	http://pastebin.com/GQ0hTzcs
10:53:36	BitPuffin	However it would be nice if one could write literals the way they actually are
10:53:56	BitPuffin	maybe I can override the = operator for literals somehow?
10:54:12	BitPuffin	and transpose the matrix they enter
10:57:23	*	erlnoob left #nimrod (#nimrod)
11:07:14	EXetoC	I don't think you can
11:09:16	BitPuffin	nope
11:09:17	BitPuffin	hmm
11:09:23	BitPuffin	I wonder how I'm gonna do then
11:09:29	BitPuffin	maybe implement a constructor for it
11:10:15	EXetoC	yeah
11:19:12	BitPuffin	what's the most efficient way to build a string?
11:31:51	BitPuffin	like if you are going to do a lot of concatenation
11:43:34	BitPuffin	char sequence?
11:46:29	*	gradha joined #nimrod
11:47:08	gradha	BitPuffin: strings are mutable in nimrod, use 'var s = "foo"; s &= "bar"'
11:48:56	EXetoC	and they are basically just seq[char] internally, right?
11:49:49	BitPuffin	okay gradha :), was just worrying if there was potential overhead doing that
11:50:10	gradha	EXetoC: no idea, maybe seqs don't guarantee memory continuity
11:52:48	gradha	BitPuffin: I guess if you already know all your strings sizes it may be more efficient to preallocate the target's destination size to avoid reallocs, newString can be used for that
11:53:17	EXetoC	dealing with C libs would be very complicated then
11:54:00	gradha	EXetoC: in system there are setLen procs for seqs, so I guess the memory is guaranteed to be continuous, otherwise those wouldn't exist
11:56:01	BitPuffin	gradha: newString(numberOfChars)?
11:56:19	gradha	yes
11:56:25	BitPuffin	will &= still work?
11:56:30	EXetoC	the documentation is pretty clear in this case
11:56:39	EXetoC	yep, string as string
11:57:26	EXetoC	no wait
11:57:56	gradha	BitPuffin: I think you would have garbage then, since &= will likely use the length to append more
11:58:17	EXetoC	"proc newStringOfCap*(cap: int): string"
11:58:29	gradha	that's more like it if you want to use add/&=
11:59:07	BitPuffin	well then it's too inconvenient :D
11:59:41	BitPuffin	easier to just use &=
11:59:42	gradha	EXetoC: yeah, you would use the lists module for non linear space
11:59:53	BitPuffin	then I don't have to worry about indices
12:00:06	gradha	BitPuffin: I would first profile before doing anything at all
12:00:30	BitPuffin	gradha: is there a good profiling tool for nimrod? or do you just mean time ./program
12:01:19	gradha	BitPuffin: I think you would use nimprof for profiling, though I haven't needed it so far
12:03:04	BitPuffin	gradha: where can I grab nimprof?
12:03:38	BitPuffin	oh it's in the stdlib
12:03:55	gradha	it needs a compiler switch to work
12:06:04	BitPuffin	I can't seem to find the docs for it
12:09:10	Araq	BitPuffin: http://nimrod-code.org/estp.html
12:09:33	BitPuffin	the exact second I click on that link on the website lol Araq :)
12:10:53	BitPuffin	isn't it just better to time the execution of something?
12:10:54	BitPuffin	or well
12:10:55	BitPuffin	not better
12:11:13	Araq	depends on what you're doing
12:11:15	BitPuffin	oh we're profiling memory
12:11:24	BitPuffin	yeah then we can't time it lol :)
12:11:38	Araq	usually profiling is not "try approach A" vs "try approach B"
12:12:04	Araq	you're quite lucky if you're able to do that
12:12:32	Araq	usually profiling means "what the heck? where is the time spent?"
12:13:16	Araq	also ESTP is no memory profiler but it can be used for that too
12:16:58	BitPuffin	Yeah I saw that
12:17:03	BitPuffin	--profiler:off lol :)
12:17:08	BitPuffin	why is it off?
12:18:45	BitPuffin	apparently creating a string out of a vector 50 000 times didn't do much
12:20:21	BitPuffin	Araq: Great that you're here btw, I'm trying to implement swizzling, It is supposed to return a vector that is the same length as the string it takes in, how do I write that in to the signature, because I can't seem to come up with anything that works
12:21:28	BitPuffin	proc swizzle*[T, I](a: TVector[T, I]; s: string): TVector[T, I] =
12:21:40	BitPuffin	doesn't really make sense to return a TVector[T, ___I___]
12:23:04	EXetoC	you'd need a template
12:23:27	EXetoC	unless you want it to be dynamic too, in which case you'd have to return a seq or something, obviously
12:24:07	BitPuffin	EXetoC: a template? and no I shouldn't need to return a seq, the string length should be known at compile time shouldn't it?
12:24:13	EXetoC	that would be kinda useless though
12:24:33	EXetoC	not if it's a function parameter
12:24:44	BitPuffin	Well
12:25:12	BitPuffin	you kind of call it like swizzle("xyzrgbargbstq")
12:25:26	BitPuffin	never really by generating a string and then calling it
12:25:34	BitPuffin	and I can't return a seq
12:25:38	BitPuffin	that would destroy errthing
12:25:40	EXetoC	yes, but you need a template then
12:26:02	BitPuffin	how would I do it with a template?
12:27:39	BitPuffin	hmm
12:27:46	BitPuffin	is there a way to make an array from a sequence?
12:28:46	EXetoC	I don't know. "template x(a: string{lit})..."?
12:31:00	*	BitPuffin goes to read the template documentation and hopes that Araq will come and enlighten me
12:37:55	BitPuffin	How do I check if a string only contains a certain kind of characters?
12:38:13	BitPuffin	like string.onlyContains("rgba") or whatever
12:38:31	BitPuffin	doesn't mean that it contains all of those, just that no other character is in there
12:40:18	BitPuffin	sounds like a regular expression to me
12:41:18	*	Associat0r quit (Quit: Associat0r)
12:43:58	EXetoC	or "not contains", where the second arg is a char set
12:44:36	gradha	maybe you could try using some variant of parseutils.skipUntil to see if you find a non matching character
12:44:43	gradha	for regular expressions there is the re module
12:45:19	gradha	or strutils.allCharsInSet
12:46:23	BitPuffin	EXetoC: considered that too
12:46:44	BitPuffin	gradha: hmm, I think re sounds like the winner
12:47:01	EXetoC	let's see if it works at compile-time
12:48:24	BitPuffin	EXetoC: well, templates pretty much inline them selves don't they?
12:48:47	BitPuffin	whatever was in the template body is substituded
12:51:00	EXetoC	yes but you probably want the 're' stuff to be executed at compile-time. anyway, it seems a little overkill
12:51:40	EXetoC	"xyzwxywzyxywz".contains({char(0)..char(255} - {'x', 'y', 'z', 'w'})
12:53:34	EXetoC	Have I missed anything?
12:56:12	BitPuffin	EXetoC: well strings aren't always known at compile time are they
12:56:30	BitPuffin	but hmm
12:56:33	BitPuffin	maybe that works
12:57:35	BitPuffin	EXetoC: but it won't happen at compile time will it?
12:58:42	EXetoC	so have it take a literal
12:58:46	EXetoC	it might
12:58:58	EXetoC	many operations are supported at compile-time
12:59:15	BitPuffin	interesting
12:59:23	BitPuffin	how do I specify it's a string literal?
13:02:07	EXetoC	see the second block of code here http://nimrod-code.org/manual.html#typedesc_toc
13:02:15	EXetoC	actually, those anchors don't even work for me
13:02:43	EXetoC	search for string{lit}. do a search for enumfield if you want a list of available constraints
13:04:18	BitPuffin	yeah the anchors suck
13:05:45	BitPuffin	EXetoC: can you use that with templates too? the example uses macro
13:10:17	EXetoC	yeah
13:11:26	BitPuffin	EXetoC: so you think `when not s.contains({char(0)..char(255} - {'x', 'y', 'z', 'w'}): {.fatal.}` will work
13:12:46	EXetoC	I don't know if these operations don't have any side effects, or at least if the compiler can accurately decide that
13:14:30	EXetoC	I'll try to figure that out
13:14:57	BitPuffin	didn't even know you could do that {blabla} - {blabla} stuff, that's very expressive
13:15:51	EXetoC	yup. set[T], and .. is usually just proc `..` in the system module
13:16:11	EXetoC	`@` as well (@[1, 2, 3])
13:18:33	EXetoC	BitPuffin: https://gist.github.com/EXetoC/6088773
13:19:31	BitPuffin	EXetoC: it actually compiles in when so it seems to work
13:19:46	EXetoC	alright
13:19:56	BitPuffin	EXetoC: although I didn't import strutils :s
13:24:33	BitPuffin	EXetoC: weird, got an invalid characters for swizzling
13:25:04	BitPuffin	oh
13:25:06	BitPuffin	it shouldn't be not
13:25:56	BitPuffin	hmm
13:25:57	BitPuffin	vector.nim(86, 23) Error: undeclared identifier: 'T'
13:26:38	BitPuffin	But T is inferred from the vector passed to swizzle :s
13:29:06	EXetoC	BitPuffin: I guess the template doesn't have any type parameters then
13:30:05	BitPuffin	EXetoC: well others had
13:30:08	BitPuffin	and they worked
13:31:17	EXetoC	I don't know what you mean, but the template needs type parameters, just as your functions do
13:31:55	BitPuffin	Well it did have type parameters
13:32:44	BitPuffin	https://gist.github.com/BitPuffin/cbaa99837fa494e07cbc
13:32:48	BitPuffin	EXetoC: ^
13:33:06	BitPuffin	and yeah I haven't updated so it also captures uppercase
13:36:06	EXetoC	oh well. [T](x: T) worked
13:37:06	EXetoC	don't you mean 0..s.len - 1?
13:37:12	EXetoC	(0.. <s.len)
13:37:17	BitPuffin	yeah
13:37:39	BitPuffin	EXetoC: so I should remove the I param?
13:43:34	EXetoC	it wasn't related to that specifically, but that might work, just that it's not constrained
13:43:52	BitPuffin	ah
13:43:56	BitPuffin	var a = [1, 3, 37].TVec3i
13:43:56	BitPuffin	echo($(a.swizzle("")))
13:44:04	BitPuffin	don't see why it doesn't work
13:44:27	BitPuffin	EXetoC: well then I get undeclared I
13:44:49	EXetoC	you'd have to change the parameter type as well then, to just T
13:45:06	EXetoC	which would be a start, assuming that it works
13:45:29	EXetoC	what part of it doesn't? arrays cannot be converted to string yet. I think it was possible before, but the function was commented out for some silly reason IIRC
13:45:33	BitPuffin	don't think that would work
13:46:06	BitPuffin	EXetoC: the part - var result: TVector[T, 0..len(s)-1]
13:46:06	EXetoC	also, $ isn't strictly necessary with echo, or any function that does something like args: varargs[T, `$`]
13:46:23	BitPuffin	I've implemented $ so it works
13:46:35	EXetoC	oh right. if only you could do something like T.ElemType
13:47:00	BitPuffin	I also tried removing it and doing type(a) but that obviously didn't work
13:47:13	BitPuffin	because that doesn't return what I want
13:47:36	BitPuffin	could this be because of a compiler bug? >_>
13:47:44	EXetoC	vec[0].type?
13:48:30	BitPuffin	well
13:48:40	BitPuffin	that assumes that the vector isn't 0 dimensional
13:51:19	EXetoC	someone in an alternate universe might need it before that (bug?) is fixed :p
13:51:26	EXetoC	anyway, it should still work
13:54:30	BitPuffin	I'll try
13:55:44	EXetoC	ok, arrays can't be zero-sized
13:55:57	EXetoC	so there you go
13:56:56	EXetoC	let a=[range[0.. -1] #Error: range is empty
13:58:12	BitPuffin	EXetoC: yeah I noticed
13:58:30	BitPuffin	EXetoC: weird though, non of the cases are cought
13:59:26	EXetoC	and if you print the value?
13:59:47	*	mario-goulart left #nimrod ("ERC Version 5.3 (IRC client for Emacs)")
14:00:15	BitPuffin	EXetoC: the value?
14:00:20	BitPuffin	ooh
14:02:04	BitPuffin	EXetoC: apparently I can't return result that way either
14:02:08	BitPuffin	it has to be discarded :s
14:03:30	BitPuffin	I have to go for a while, I'll push in the mean time, if you find a solution, feel free to do a pull request :P
14:04:40	BitPuffin	https://github.com/BitPuffin/linagl
14:04:42	BitPuffin	cheerio
14:05:42	Araq	BitPuffin: does that template work? :O
14:06:00	Araq	template swizzle*(a: TVector; s: string{lit}) ?
14:07:55	EXetoC	I don't think he has actually been able to have it generate a value yet. why, is it experimental?
14:08:20	Araq	not really
14:08:52	EXetoC	maybe he needs to return expr or something. what's the default, stmt?
14:09:01	Araq	yeah
14:09:09	*	BitPuffin quit (Ping timeout: 264 seconds)
14:09:12	Araq	make it return 'expr'
14:09:45	EXetoC	NimBot: hello
14:10:13	gradha	NimBot: don't be shy
14:11:45	EXetoC	Araq: wasn't the whole "proc `$`*[T](a: openArray[T]): string" issue silly? :p
14:12:38	EXetoC	do you want separate functions for array and seq?
14:13:29	EXetoC	nvm, a compile-time branch will do (for adding @)
14:13:41	Araq	I don't but fowl wants the @ iirc
14:13:49	Araq	also again
14:14:05	Araq	earlier versions conflated "openArray" and "varargs"
14:14:22	Araq	varargs means anything gets a buggy $ operator
14:14:29	Araq	so it was disabled
14:14:40	EXetoC	oh yeah
14:25:54	*	DAddYE joined #nimrod
14:32:37	*	DAddYE quit (Ping timeout: 248 seconds)
14:41:35	*	zahary_ joined #nimrod
15:21:20	*	Sergio965 joined #nimrod
15:29:06	*	DAddYE joined #nimrod
15:35:33	*	DAddYE quit (Ping timeout: 248 seconds)
16:13:39	*	Mat2 joined #nimrod
16:13:43	Mat2	hi @ all
16:17:22	Araq	hi Mat2
16:17:33	Mat2	hi Araq
16:18:32	Araq	interleaving the instruction stream with proc bodies is evil ...
16:19:00	Araq	I generate 'jmp XXX; <proc body>; XXX: ... '
16:19:14	Araq	but debugging this is pita
16:19:23	Araq	any good ideas, Mat2?
16:19:27	Mat2	I have check-up a VM model build from your sources and implemented some optimizations
16:19:59	Araq	well since it doesn't work yet how do some optimizations help? :-)
16:20:08	Araq	but sure, I'm interested
16:21:32	Mat2	ehm, the mdoelled vm used your encoding and instruction-set but for speed I've recoded it in forth
16:21:36	Mat2	^modelled
16:22:17	Mat2	the results told me that dispatching can be reduced to 10-15 clock cycles per instruction
16:22:44	Mat2	(dependent of the cpu of course)
16:22:56	Mat2	that's not bad for a register-based vm
16:23:33	Mat2	the greatest gain in performance comes with static trace-compilation
16:24:50	Mat2	this eases also debugging because you only have a list of address pointers which can be traced beside execution (just modifies the next routine)
16:25:40	Mat2	I hold a precached buffer in memory for 32 decoded instructions
16:26:29	Mat2	and another for 64 immediate parameters
16:26:46	Araq	how can you accurately model software pipelining when you don't know the typical instruction stream?
16:29:16	Mat2	the buffer is fixed in size and because each instruction compiles into an address and each buffer-element maps directly to an element of the immediate-data cache you can implement software-pipelining just by preloading in the primitive-code
16:30:36	Mat2	it's only wise to reserve some register (or statc variables) for that
16:30:44	Mat2	^static
16:31:36	Mat2	in forth jargon this is called static-replication
16:35:44	Mat2	http://www.complang.tuwien.ac.at/projects/interpreters.html
16:35:54	Mat2	^here you find some papers about this stuff
16:39:16	Mat2	^^ the idea is that if its not possible to meassure instruction streams directly, that's possible for indirect branches and memory references
16:42:12	*	DAddYE joined #nimrod
16:49:59	Mat2	Araq: We bring up Nimrod as language with help of my current vm to the duinomite (the original planning was to choose Pascal but Nimrod seem to be easier for beginners)
16:54:53	*	Associat0r joined #nimrod
16:54:53	*	Associat0r quit (Changing host)
16:54:53	*	Associat0r joined #nimrod
16:57:05	Araq	Mat2: awesome
16:57:06	*	zahary_ left #nimrod (#nimrod)
17:06:22	Araq	well Mat2 feel free to optimize the heck out of it once it's stable
17:06:47	Mat2	I will do :)
17:08:35	Araq	how is Nimrod easier than Pascal btw? :-)
17:10:01	Mat2	I think Pascal features some excentrical syntax shortcomings inherited from Algol
17:11:03	Araq	it surely confuses verbosity with readability
17:11:19	Mat2	I know beginners learning it are confused from using semicolons as sequence terminator for example
17:11:58	Mat2	there arr many details most of us don't think of
17:12:01	Mat2	^are
17:19:34	Mat2	on the other side, the language is somewhat restricted compared with Nimrod
17:20:23	Mat2	no prototype support etc.
17:20:34	Mat2	it is showing Age
17:55:55	*	Raynes quit (Excess Flood)
17:56:15	*	Raynes joined #nimrod
18:00:31	dom96	hello
18:00:44	Mat2	hi dom96
18:05:50	EXetoC	dom96: YOOOOOOOOOOOO
18:06:05	dom96	HELLOOOOOO Mat2 and EXetoC !!!!!!
18:07:02	Mat2	hi EXetoC *g*
18:07:58	dom96	So I am now free until the end of August!
18:18:59	EXetoC	Mat2: ohai
18:19:17	EXetoC	dom96: good stuff. watcha gonna do? write at least *some* Nimrod code I hope :>
18:21:01	dom96	oh dude, I will be writing nimrod ALL the time.
18:23:08	Mat2	a suggestion: write a demo and post it at pouet.net, this should attract more programmers to Nimrod
18:24:08	EXetoC	dom96: oic :p
18:26:07	EXetoC	wasn't I supposed to exercise?
18:26:16	EXetoC	I blame BitPuffin and his stupid code
18:41:40	*	EXetoC is now known as EXetoC_
18:41:59	EXetoC_	swizzling is implemented now. compile-time evaluation is pretty solid
18:42:43	Araq	"pretty solid" -- perhaps
18:42:53	Araq	but not good enough
18:43:56	dom96	Mat2: I actually am a bit interested in the demoscene. Maybe someday... I have too many things planned already.
18:44:38	Araq	dom96: we need bitbucket support :P
18:44:50	dom96	I know.
18:45:12	dom96	(You mean for babel right?)
18:46:53	Araq	right
18:48:14	EXetoC_	maybe not, but I haven't been blocked by anything so far. I'll re-evaluate it if I ever use it for rocket science :>
18:48:50	Araq	I ran a simple SDL demo with -d:useFFI
18:48:54	Araq	at compile time
18:48:58	Araq	:D
18:56:35	Sergio965	Woo!
18:58:00	Mat2	nice
18:58:49	Araq	key presses don't work though as the compiler still computes field offsets wrongly
18:59:21	dom96	wow, i'm getting a big deja vu
18:59:49	dom96	I guess this is with your new VM?
19:00:02	Araq	no ... *cough*
19:00:11	Araq	but I said "I ran" :P
19:00:14	dom96	lol
19:00:29	Araq	and EXetoC wasn't around back then
19:19:50	*	BitPuffin joined #nimrod
19:23:52	*	rubino123 joined #nimrod
19:24:12	Araq	hi rubino123, how is progress with websockets?
19:26:20	rubino123	I have not started the nimrod version yet
19:26:37	rubino123	Got another project that took over my freetime
19:26:52	rubino123	Its related to my need for the websocket though
19:26:52	Araq	aww
19:27:05	rubino123	how tight is the postgres wrapper?
19:27:38	Araq	I used it in production, kind of
19:27:50	rubino123	You are the lead dev?
19:27:51	Araq	switched to sqlite quite soon afterwards though
19:28:22	rubino123	Was it a postgres problem with untrusted langauges or something else?
19:28:32	Araq	yes, I'm the creator
19:28:44	rubino123	I have high hopes for nimrod
19:29:12	Araq	no problem with postgres really but too tedious to setup on a server
19:29:12	rubino123	Its impressive at delivering the performance at a reasonably productive syntax
19:29:22	Araq	thank you
19:29:32	rubino123	OK, I have a DBA so i just need to write the integrations
19:29:40	gradha	now that ~30 is usual on IRC, what's the next milestone?
19:30:49	Araq	gradha: ~50 of course and I'll be here more rarely so that my presence is a cool event to celebrate
19:30:57	EXetoC_	BitPuffin: The string stuff would probably be executed at run-time, unless optimized away, and I've dealt with that
19:31:37	*	EXetoC_ is now known as EXetoC
19:31:38	Araq	EXetoC_: use a macro to optimize it away
19:32:17	EXetoC	I used compileTime
19:34:18	rubino123	Araq: where can I find the milestones for releases if any?
19:35:47	Araq	rubino123: $nimrod/todo.txt
19:36:40	Araq	these are however my personal notes
19:37:01	Araq	so most of it is overly terse and not understandable :P
19:40:44	rubino123	good enough
19:40:45	rubino123	thanks
19:41:21	rubino123	Is numerical computing going to be emphasized in the ecosystem? I ask because I saw the libsvm bindings
19:42:43	Araq	why do you think we have this: http://nimrod-code.org/trmacros.html ? ;-)
19:44:46	rubino123	holy macros batman!
19:47:15	*	OrionPK joined #nimrod
19:54:52	*	EXetoC quit (Quit: WeeChat 0.4.1)
19:55:06	*	EXetoC joined #nimrod
19:56:42	gradha	Araq: see http://forum.nimrod-code.org/t/184/1#904 and please tell me I won't have to build manually the question mark list for the sql statement
19:59:20	dom96	That last example looks like a bug to me.
20:01:16	Araq	dom96: hu? & for sequences is a feature
20:01:29	Araq	it works exactly as its signature suggests
20:01:51	dom96	Shouldn't it output: 1374868327, one, two?
20:02:20	EXetoC	high and low seems to work at compile-time, but not len in some cases
20:08:26	Araq	dom96: good point
20:08:34	Araq	looks indeed like a bug
20:09:11	Araq	lol yeah
20:09:21	Araq	copy paste not enough modified
20:10:10	reactormonk	copypasta!
20:10:58	EXetoC	mmh, pasta
20:12:40	gradha	is it really a bug? I mean, for a & b I would expect b to be concatenated to a, so b would need to transform into a string first
20:13:15	gradha	unless the rule is "things get upgraded to the highest level object"
20:13:23	gradha	whatever that means
20:13:51	*	gradha sighs and starts writing his own quote concatenation query formatter
20:14:21	Araq	proc `&` *[T](x: T, y: seq[T]): seq[T] {.noSideEffect.} =
20:14:23	Araq	newSeq(result, y.len + 1)
20:14:24	Araq	result[0] = x
20:14:25	Araq	for i in 0..y.len-1:
20:14:27	Araq	result[i+1] = y[i]
20:14:35	Araq	that's how it looks like after the fix
20:14:58	Araq	no idea why you think the old buggy behaviour is valid
20:15:45	Araq	"Huh, this is something that sort of works automatically for other language/bindings I've tried, help this spoiled brat!"
20:15:49	gradha	I expect b to be transformed to the type of a, not the other way round
20:16:10	Araq	<-- no idea where it works this way
20:16:14	gradha	python
20:16:24	Araq	ok fair enough
20:16:32	Araq	it surely doesn't work this way in C#
20:16:45	Araq	well ... you're right it's a problem
20:17:10	Araq	we could introduce ?! as a way to insert stuff without quoting
20:17:23	Araq	or maybe just !
20:17:44	reactormonk	Araq, that gets hairy
20:18:02	dom96	Sounds dangerous.
20:18:20	dom96	I accidentally misplace my params and i'm fucked.
20:18:22	Araq	well I'm not happy with the ? syntax, I think it should use $# like format
20:18:45	Araq	dom96: you accidentally wrote ! instead of ?
20:19:05	dom96	You never know :P
20:19:12	gradha	this "feature" can also work in objc, most sqlite bindings require you anway to wrap simple values into objects so the wrapper can do these magical things
20:19:13	EXetoC	you don't like the printf syntax?
20:19:36	dom96	If I mix ! and ? then it gets messy doesn't it?
20:19:53	dom96	I have to be sure that I give it the params in the correct order is what I meant.
20:20:37	Araq	EXetoC: printf makes sense when you have no type safety, otherwise the 1 letter types after the % are stupid
20:20:57	gradha	it's surprising C# doesn't expand collections for the question mark
20:22:00	dom96	why is there no printf macro?
20:23:36	EXetoC	Araq: no need to use all of it, or at least not force that. $# is kinda ugly, but maybe I can get used to it
20:24:30	Araq	$# is excellent once you understand it IMHO
20:24:44	EXetoC	well, the annoying part is having to press two different modifier keys
20:25:01	Araq	get a german keyboard then
20:25:35	Araq	dom96: so ... what's your alternative? use !!! instead?
20:26:07	dom96	Araq: I don't really have an alternative. Just raising my concerns.
20:31:20	EXetoC	BitPuffin: linaGL? I see what you did there
20:32:33	Araq	gradha: what do you think of my '!' solution? it's 2 lines of code in db_sqlite.nim ;-)
20:32:53	Araq	ok, ok, 3 lines
20:33:15	gradha	wouldn't you also need to export dbQuote so I can prepare the params?
20:33:34	Araq	true
20:33:52	gradha	I like more the idea of creating n question marks
20:34:01	EXetoC	anyway, just a single character for the basic case would be nice. plenty of characters are not that commonly used
20:34:23	Araq	gradha: that requires no changes to db_sqlite then, right?
20:35:16	gradha	no changes required
20:35:29	dom96	oooh mercurial has a built-in web server
20:36:25	Araq	gradha: fine with me but dbQuote should be exported and in fact
20:36:32	Araq	now that we have 'export'
20:36:44	Araq	we can pull common code into a db_common
20:37:06	Araq	currently there is some duplication in all of these db_* wrappers
20:37:06	gradha	I think my memory is wrong, according to http://stackoverflow.com/questions/283645/python-list-in-sql-query-as-parameter this behaviour of list expansion doesn't work in python either
20:37:22	gradha	so either I'm confusing the language or I smoke good stuff
20:39:47	gradha	maybe python is like 8bit games: you remember it fondly, but your memories don't match reality
20:41:14	Araq	I surely don't remember matplotlib fondly :P
20:41:38	EXetoC	doesn't it? I thought you could expand in all sorts of places
20:42:03	gradha	I remember PIL and pyglet with special hate
20:49:38	Araq	I ended up drawling every single pixel on my own to generate the graphs exactly how I wanted them
20:50:31	gradha	so just to make sure: string & seq[string] produces seq[string] instead of string?
20:50:51	Araq	sure
20:50:57	gradha	aha
20:51:04	Araq	auto join is no concept in the stdlib
20:58:24	dom96	BitPuffin: Do you know if hg has a git-like --depth param for 'clone'?
21:00:13	Mat2	gradha: do you know the ABC80 ?
21:02:19	gradha	nope
21:04:10	Mat2	gradha: https://sv.wikipedia.org/wiki/ABC_800; It featured a half-compiling basic interpreter named ABC. Then take a look at the predecessor influence for pyhthon: The ABC language!
21:05:39	gradha	I remember typing hexadecimal programs into a spectrum without tape, FWIW
21:06:06	Mat2	nice hobby !
21:06:16	gradha	oh, the joy of magazine erratas
21:08:19	*	gradha goes thirsty into the kitchen to return with salty pistachoes
21:08:23	Mat2	some years ago, I learned most z80 instructions by coding a game for the zx81 in machine-code
21:25:56	*	Sergio965 quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…)
21:36:34	Mat2	ciao
21:36:42	*	Mat2 quit (Quit: Verlassend)
21:39:20	gradha	shame that the `?` operator requires backticks
21:39:37	gradha	have to write varname.`?` instead of varname.?
21:40:15	Araq	it's hard to support varname.? as .? would be a postfix operator
21:40:30	EXetoC	gradha: I've never needed them
21:40:59	gradha	is it possible to have ?varname or varname? without the dot?
21:41:51	EXetoC	the former should work
21:41:54	Araq	?varname is easy
21:42:22	Araq	template `?`(x: expr): expr = x
21:43:40	gradha	wonderful
21:44:12	Araq	in my example code I make ? a nop. I'm sure that's what you want.
21:49:02	Araq	so dom96 is right: http://memedad.com/meme/19953
21:49:09	dom96	ahaha
21:49:35	Araq	but I'm sure I soon get 'strutils.replace' to work in my VM
21:52:33	rubino123	Araq: I do not have lua on my system but the lua example runs?!?
21:52:46	*	Endy quit (Ping timeout: 276 seconds)
21:53:33	Araq	rubino123: it's magic ;-)
21:54:16	Araq	if you're on linux chances are high that you have it on your system though
21:54:29	Araq	nimrod doesn't require devel-packages
21:57:31	dom96	I am such a noob at mercurial...
21:59:04	dom96	I have to merge before I push huhhh
21:59:22	gradha	so like git?
21:59:28	dom96	or maybe only when I create a new tag.
21:59:53	dom96	gradha: huh? In git you just commit and push.
22:00:19	gradha	you can't push if the remote is out of sync
22:00:43	Araq	dom96: what about scriptability for the download?
22:00:44	gradha	maybe mercurial push is different from git push?
22:00:57	dom96	hrm, ok. Perhaps mercurial pull doesn't automatically merge origin with local
22:01:06	dom96	Araq: hrm?
22:01:18	dom96	Araq: Have you tested babel yet?
22:02:49	Araq	no ...
22:02:59	Araq	sorry
22:03:55	Araq	dom96: I mean instead of hardcoding git, mercurial and then svn, make it scriptable
22:04:48	dom96	uhh how would that work?
22:04:55	dom96	an embedded script in the packages.json?
22:05:08	Araq	yeah
22:05:14	Araq	perhaps
22:05:19	dom96	no. I'm not adding that.
22:05:22	Araq	or in the .babel file
22:05:48	dom96	What's wrong with it being hardcoded?
22:05:59	Araq	downloadCmd: "mercurial pullit merc://my.mercuiral"
22:06:23	dom96	What about tag switching?
22:06:27	dom96	what about tag lists?
22:06:40	Araq	what about an FTP download?
22:06:49	dom96	That will be added.
22:07:02	dom96	along with http
22:07:24	dom96	Why should every package specify the same downloadCmd?
22:07:41	gradha	I want a megaupload download too
22:07:56	dom96	but megaupload is dead :\
22:07:58	*	Associat0r quit (Quit: Associat0r)
22:08:08	gradha	ok, then napster is the way to go
22:08:29	Araq	what if I use codeBeamer?
22:08:37	Araq	are you going to add support for that too? :P
22:08:48	gradha	is a codeBeamer like a star treak tricoder for code?
22:09:17	Araq	codeBeamer is a web based Collaborative Application Lifecycle Management and Requirements management tool for distributed software development, written in Java.
22:09:28	dom96	maybe I should add support for mail, you know, in case someone wants to send you a CD with a babel package.
22:09:33	gradha	right, so it's jave, now I know you are joking
22:09:43	dom96	babel can then go to the post office and pick it up for you
22:09:54	Araq	dom96: now you got the spirit
22:10:07	gradha	dom96: floppies FTW https://www.youtube.com/watch?v=up863eQKGUI
22:12:05	Araq	but there is at least Bazaar, Darcs and Fossil to support too
22:12:27	dom96	screw that shit
22:12:49	Araq	downloadCmd is starting to look nice, hu? :P
22:12:59	dom96	no
22:21:52	dom96	What should be installed by default if the git repo has tags, git HEAD or the latest tag?
22:23:31	dom96	(And by git I mean any VCS)
22:24:48	dom96	i'm going to go for the latter
22:26:04	Araq	really?
22:26:08	Araq	not git HEAD?
22:26:15	Araq	what about living on the edge?
22:26:24	dom96	it's possible that HEAD will be unstable.
22:26:42	Araq	yeah but that's the point
22:26:59	Araq	otherwise the tag should have been included
22:27:07	dom96	I suppose it's easier to specify the specific version that 'latest'
22:27:17	dom96	*than
22:27:49	dom96	i'm not sure what you mean by what you just said
22:28:21	Araq	well I dunno, it seems we need 2 different "latest"
22:28:43	Araq	latest tag and HEAD
22:29:56	dom96	no, this makes sense.
22:30:02	dom96	you're confusing me
22:30:13	Araq	ok
22:30:14	dom96	You don't want your dependencies to be HEAD by default
22:30:22	Araq	I do
22:30:32	*	Araq doesn't care about tags
22:30:34	dom96	Safer to have them be the latest version
22:31:03	Araq	safer to include every last bugfix :P
22:31:21	dom96	many use the master branch as the dev branch...
22:31:29	dom96	like, well, you.
22:31:41	dom96	until recently perhaps
22:31:42	dom96	:P
22:33:00	dom96	it's a pretty bad idea to depend on HEAD.
22:33:56	dom96	haven't you read the Go story? :P
22:34:19	gradha	the Go story or the Go! story?
22:34:40	dom96	The Go!!! story!
22:34:55	gradha	don't know about it, sorry
22:35:32	Araq	dunno I think Go's packager manager is awesome
22:35:47	Araq	it's clearly broken but simple
22:36:03	Araq	and since every package manager is broken anyway, simplicity wins
22:36:11	Araq	of course except Babel
22:36:20	dom96	babel always wins
22:36:31	Araq	Babel will be the first non broken pakcage manager
22:36:53	dom96	Well thinking about this more. It's a bit odd.
22:37:01	dom96	If there are tags present in the repo
22:37:07	gradha	searching for "The Go!!! story!" I arrived to http://gointothestory.blcklst.com. Nice police clip.
22:37:13	dom96	babel will find the latest version and use it
22:37:17	dom96	if there are not it will use HEAD
22:37:53	dom96	This behaviour could be confusing.
22:38:53	dom96	But it makes sense I think.
22:41:32	EXetoC	your head is confusing
22:41:37	BitPuffin	EXetoC: yeah! too bad it won't be called linagl for much longer :)
22:41:57	BitPuffin	dom96: not entirely sure, ask in their irc perhaps?
22:42:13	dom96	it's not that important.
22:42:48	dom96	EXetoC: I'm way too tired to be thinking about deep package manager issues :\
22:43:02	BitPuffin	dom96: it's for submodules right?
22:43:06	EXetoC	I just said something random, really
22:43:44	EXetoC	and I meant the shape of your head. think Karl Pilkington
22:44:09	BitPuffin	Karl Pilkington is amazing
22:44:13	dom96	BitPuffin: It's to do with history. --depth 1 will only download the latest commit, it will not download the rest of the history.
22:44:26	EXetoC	yes, with a head like a fucking orange
22:44:27	BitPuffin	dom96: oooh, yeah that should be useful
22:45:21	EXetoC	BitPuffin: ok. I guess there's no need to mention OpenGL all that much
22:45:41	BitPuffin	dom96: maybe hg clone -r tip or head or whatever it is again
22:46:39	Araq	yay strutils.replace works
22:46:54	EXetoC	c(:)
22:47:08	BitPuffin	EXetoC: yeah, I hope to kind of somehow make it portable for other platforms too, the game api is ment to be able to run on pretty much everything, so say the native ps4 api ain't opengl and uses row major arrangement or whatever
22:47:28	BitPuffin	hey Araq, care to help a guy out?
22:47:44	Araq	BitPuffin: always
22:48:33	BitPuffin	Araq: I'm trying to implement swizzling right? But I'm getting a bunch of weird errors, it needs to be a template since I can't figure out the return type at compile time
22:48:53	EXetoC	and my pull request?
22:49:06	BitPuffin	EXetoC: oh, I didn't check
22:50:03	BitPuffin	EXetoC: why can't it take a TVector?
22:50:49	BitPuffin	EXetoC: I find it funny that you made a bunch of semantic changes after your preferences >_<
22:51:23	EXetoC	oops
22:51:42	EXetoC	I'll remove that
22:52:10	BitPuffin	EXetoC: and I'm fairly sure len(a) worked, why make it high - low + 1?
22:52:43	EXetoC	will only take a minute
22:52:53	EXetoC	BitPuffin: it didn't when I tried
22:53:07	BitPuffin	EXetoC: Do you have the latest nimrod?
22:53:08	EXetoC	it simply didn't work at compile-time at that point
22:53:30	Araq	EXetoC: I fixed the 'len' bug
22:53:31	EXetoC	no, but I'll install it
22:53:34	EXetoC	alright
22:53:50	EXetoC	so I'll commit some stuff after that
22:54:05	BitPuffin	Btw
22:54:15	BitPuffin	what's the type for these literals?
22:54:20	BitPuffin	{'x', 'X', 'r', 'R', 's', 'S'},
22:54:27	EXetoC	set[char]
22:54:31	BitPuffin	ah
22:54:34	BitPuffin	so it's sets
22:54:48	EXetoC	you can echo it. it'll give you an error, but with the type in it :p
22:55:02	EXetoC	usually easier than importing that module
22:57:56	BitPuffin	I'm fairly sure I tested it :s
22:59:01	EXetoC	it should tell you what type you tried to pass to it. typedesc[something]
23:00:56	NimBot	nimrod-code/babel master 07b7f46 Dominik Picheta [+1 ±5 -0]: Implemented download of packages residing in mercurial repos.... 5 more lines
23:01:08	dom96	BitPuffin: there you go.
23:01:24	dom96	oh yeah, the --depth stuff
23:03:32	*	gradha quit (Quit: bbl, need to watch https://www.youtube.com/watch?v=1ZZC82dgJr8 again)
23:03:47	NimBot	nimrod-code/babel master 4bfbc1a Dominik Picheta [+0 ±1 -0]: Improved efficiency of hg cloning.... 2 more lines
23:05:09	BitPuffin	dom96: awesome!
23:09:23	NimBot	nimrod-code/packages master d401396 Dominik Picheta [+0 ±1 -0]: Added info about new hg support.
23:11:01	BitPuffin	EXetoC: a.high instead of high(a) works fine though :)
23:11:02	dom96	BitPuffin: Let me know if you find any bugs :)
23:11:08	BitPuffin	dom96: sure!
23:11:40	BitPuffin	dom96: how do you suggest I package this api for babel, some people might only want to grab the math part of it for example
23:13:08	EXetoC	BitPuffin: I reverted all that, but ok
23:13:21	BitPuffin	EXetoC: ah, I'll change it later then
23:13:38	dom96	You could put it into separate repos, or use directory structure like 'mylib/math.nim'
23:13:41	dom96	*use a
23:14:13	BitPuffin	dom96: yeah I plan to do the second one, do you think people will mind downloading everything?
23:14:33	Araq	BitPuffin: I don't think so
23:14:47	Araq	I surely don't mind if it's 10K or 50K
23:15:07	EXetoC	BitPuffin: the issue was that it wouldn't recognize TVector as an array even though it's just an alias, and so overloading `==` this way seemed like the easiest solution
23:15:11	dom96	I guess it depends how big it is.
23:15:20	EXetoC	but it doesn't really matter since it's an alias
23:15:56	dom96	Some people may dislike depending on a library from which they only use one module.
23:16:56	BitPuffin	EXetoC: weird, but it worked when i tested it
23:17:08	EXetoC	I'll try again
23:18:05	EXetoC	that appears to have been fixed too
23:20:12	BitPuffin	EXetoC: the len thing?
23:21:31	Araq	Nimrod: 0.0790179
23:21:38	Araq	Python: 0.022048
23:21:54	Araq	for my shitty strutils.replace benchmark ;-)
23:23:29	BitPuffin	Araq: higher is worse right?
23:23:34	Araq	yeah
23:23:42	BitPuffin	Well, that can be optimized
23:23:47	dom96	not bad for a VM you wrote in a couple of weeks.
23:23:50	Araq	but python's string.replace is compiled code
23:24:00	Araq	nimrod's is interpreted
23:24:28	BitPuffin	huh?
23:24:29	BitPuffin	oh
23:24:30	BitPuffin	it's the VM
23:24:34	Araq	yeah
23:24:39	BitPuffin	I thought it was statically compiled nimrod
23:24:42	BitPuffin	how does that compare?
23:25:06	Araq	I don't care
23:25:20	BitPuffin	fair enough
23:25:27	BitPuffin	it's probably about the same one would assume
23:26:25	Araq	not really, python has some overhead
23:26:26	EXetoC	BitPuffin: that and the TVector stuff. fixing both
23:27:32	BitPuffin	Araq: so you think it's slower?
23:27:46	Araq	BitPuffin: yeah
23:28:52	NimBot	Araq/Nimrod master ce6df5f Araq [+0 ±1 -0]: bugfix: system.& for sequences
23:28:52	NimBot	Araq/Nimrod master 3f63f76 Araq [+0 ±3 -0]: new vm: lots of fixes
23:29:03	BitPuffin	Araq: hmm, yeah it might be, I was just thinking that they have more resources to optimize such things, but obviously python can't compare to a compiled language
23:29:52	Araq	BitPuffin: perhaps you're right
23:30:14	Araq	well my benchmark is silly anyway as it includes startup overhead in both
23:30:24	BitPuffin	hehe :)
23:30:26	BitPuffin	oh well
23:30:36	BitPuffin	It will get better I'm sure
23:30:49	BitPuffin	and the VM will help me work around apple shitty bullshit shit
23:31:01	Araq	well nimrod needs to process system.nim and strutils.nim at startup
23:31:06	BitPuffin	even though iOS gets a far slower version of the game
23:31:27	Araq	so it's suprising these numbers are comparable
23:31:48	Araq	and we're only a factor of 4 slower
23:35:37	BitPuffin	yeah truly
23:35:46	BitPuffin	and it's a very new VM
23:35:52	BitPuffin	that was written in a short time
23:36:08	BitPuffin	says a lot about nimrods productivity / performance ratio, hats off
23:36:17	Araq	indeed
23:36:33	Araq	I can't imagine debugging it somewhere else
23:37:49	Araq	now only ~20 missing features have to be implemented :P
23:38:57	Araq	good night
23:43:16	BitPuffin	night!
23:51:25	BitPuffin	EXetoC: why did you remove +? I know it's silly but I don't want compilation to fail just because someone wrote +a
23:52:03	EXetoC	huh
23:52:42	BitPuffin	oh wait
23:52:46	BitPuffin	github showed it as removed
23:52:49	BitPuffin	it was just moved
23:53:09	EXetoC	yeah
23:53:25	BitPuffin	EXetoC: also why `U`?
23:54:00	EXetoC	missed something. wait
23:54:21	BitPuffin	also ~= still takes an array
23:54:29	BitPuffin	and T is not a range, I is (interval)
23:58:18	EXetoC	I can't remember why I did that, but I can fix that in another branch
23:59:08	BitPuffin	well I haven't merged yet
23:59:15	BitPuffin	so go ahead and add :)