<< 17-03-2014 >>

---

00:00:07	OrionPK	Why doesnt this work; $(value.date.int) & "some other string"
00:00:15	renesac	but you could be sitting in a TChair or a PChair...
00:00:17	OrionPK	conversion from string to int is invalid
00:03:26	dom96	OrionPK: What type is value.date?
00:03:29	OrionPK	oh god, nevermind
00:03:35	OrionPK	the date is a string, thought it was a TTime
00:09:03	Skrylar	its more a matter that you're not going to name a module tfoo but you might name it after what it contains, and if the typename is the same as a module name nim gets confused
00:10:50	EXetoC	Skrylar: so rectangle/Rectangle then, but you need case sensitivity
00:11:31	Skrylar	i would rather have trectangle than case sensitivity
00:11:51	Skrylar	but we're not supposed to use T/P's (lol toiletpaper), so i'll probably rename the module to rects
00:12:02	runvnc	so I don't put T in front of the type name right
00:12:14	clovis	It's the other way around. Rectangle is the base entity, not the module.
00:12:31	EXetoC	Skrylar: well we're getting it
00:12:43	Skrylar	EXetoC: it sounded like araq nixed it last time i checked the thread
00:12:45	EXetoC	you can do that if you want but you need to enable case sensitivity
00:12:53	runvnc	yeah just realized that
00:12:55	EXetoC	uh ok
00:13:06	runvnc	because I have a TSession and a session var
00:13:09	Skrylar	he mentioned partial case sensitivity was more migraine than it was worth in the last post, but i haven't checked in a few days
00:13:30	runvnc	I think for now I just need to stick with T for practicality sake
00:13:30	EXetoC	so we're getting full case sensitivity then?
00:13:40	runvnc	personally I think it is good that way it is
00:13:46	Skrylar	^
00:13:59	clovis	Me too, except scrap the T/P
00:14:03	Skrylar	He said there won't be any major syntax changes soon when filwit was poking about moving some symbols around
00:14:18	Skrylar	I'd imagine case sensitivity is a 'major change' which is another hint that it was scrapped
00:15:02	Skrylar	i'd also imagine it would just be confusing since the existing blog posts already mention the style insensitivity
00:18:39	clovis	A variable called "rectangle" does a bad job at describing *what* rectangle. It only states the type of the thing.
00:19:06	EXetoC	clovis: T is necessary with case insensitivity
00:19:07	clovis	A module describing the type Rectangle is itself not a rectangle
00:19:50	Skrylar	EXetoC: i've been getting along fine without T/P so far (well, without T.)
00:19:56	Skrylar	I don't see the hate against P.
00:19:57	EXetoC	I think the problem was that he didn't enable partial case sensitivity
00:20:03	EXetoC	so Rectangle == rectangle
00:20:05	clovis	EXetoC, what I'm trying to say is that case sensitivity isn't the problem. The identifiers naturally have different names
00:20:40	clovis	One is Rectangle the other is something more descriptive, like windowFrame
00:21:33	Skrylar	EXetoC: nope, the problem is that 'rectangle.nim' and 'rectangle the type' have the same identifier
00:21:53	Skrylar	and nimrod isn't smart enough to figure out what i meant
00:22:15	EXetoC	oh you actually named it that way. that's not good
00:22:25	clovis	Skrylar, the type Rectangle has the correct name. The module which itself isn't a rectangle should have another name
00:22:30	Skrylar	it works in java, D and C# :\
00:22:58	dom96	indeed, nimrod seems to have more trouble dealing with module names which are the same as types/identifiers than it has dealing with types which are the same as identifiers.
00:23:19	dom96	I think this finally deserves an issue.
00:23:29	dom96	Because there have been many cases where people got confused.
00:23:34	EXetoC	clovis: it's an accurate name
00:23:44	Skrylar	dom96: well, the compiler crashed. thats a pretty confusing issue :)
00:24:02	Skrylar	I saw Zah's fix and his error message isn't much better; it just says "couldn't instantiate" which doesn't tell you that the symbol name conflicted
00:24:43	EXetoC	clovis: you could bundle it with similar things and call it something else, but sometimes it doesn't make sense
00:26:07	clovis	from Rectangle import Rectangle is confusing
00:26:19	Skrylar	you don't need to from rectangle import rectangle
00:26:37	Skrylar	the only thing in skylight.rectangle is the type and methods that operate directly on the type
00:27:56	clovis	I guess so
00:28:05	*	Matthias247 quit (Read error: Connection reset by peer)
00:28:36	EXetoC	clovis: not when the module name is lower case
00:29:01	clovis	EXetoC, I take it that you want case sensitivity
00:29:27	EXetoC	no, but you'd say "from rectangle import TRectangle" now, assuming that the conventions are adhered to
00:30:17	clovis	Or, why not "from RectangleModule import Rectangle"
00:30:18	EXetoC	s/no/don't really mind either way/
00:30:58	EXetoC	that'll get annoying fast
00:31:16	clovis	I'd refer to the type a lot more often than the module
00:31:31	Skrylar	I don't use the TFoo convention because its apparently taboo
00:31:33	clovis	I'd rather have the type be the natural name
00:31:37	Skrylar	even though I prefer it
00:31:39	OrionPK	dom96 is there a way to convert a TTimeInfo to a TTime
00:32:53	OrionPK	nm, found it. TimeInfoToTime
00:32:55	clovis	Why waste readability by having extra characters added to the type name when they instead could be added to the less used module name
00:32:59	EXetoC	Skrylar: everyone has been doing that for ages
00:33:26	clovis	Legibility is the most important aspect, imho
00:33:39	Skrylar	i don't lose readability from TRectangle
00:33:45	fowl	Skrylar, https://gist.github.com/fowlmouth/9584728
00:33:49	*	Skrylar is one of those pascal-y people
00:34:27	clovis	It's still a bit of misplaced noise
00:35:06	Skrylar	fowl: neat.
00:35:37	Skrylar	fowl: now the second problem is, "is that a good pattern for nimrod?" since the closures are going to end up making the GC keep the objects alive unless you explicitly disconnect
00:36:23	EXetoC	clovis: "variable names should be more descriptive than that" is too general I think
00:36:30	EXetoC	so what about collisions?
00:36:31	fowl	if thats a problem we can gcunref after connecting, but yes ill need tests written
00:37:21	Skrylar	gcunref means when you trigger you'll be calling dead pointers, that's also bad
00:37:41	clovis	EXetoC, I don't have the answer except I think that it might depend on the context. In the context of dealing with rectangles a particular rectangle should be better described with a more specific name, such as windowFrame
00:37:53	Skrylar	Qt deals with it with some magic in that a QObject's destructor goes through a list of backlinks and performs disconnects, which is invasive but it prevents sig/slots from ever segfaulting
00:38:02	Skrylar	instead you get log output that someone just signalled a dead window
00:38:40	fowl	yea
00:39:06	Skrylar	might be possible to do backlinks + unref
00:39:18	fowl	i need to store the object as well as a closure for the function
00:39:34	Skrylar	after i fix the rectangle packer i have a todo to look in to event patterns
00:39:59	Skrylar	signals and slots are the 'neatest' one but they also require a bit of participation from classes to do really well
00:40:11	Skrylar	event pumps are the safest but least pretty to use
00:40:26	fowl	i was hoping to avoid using a supertype but i dont think it will be possible
00:40:42	Skrylar	shouldn't need a supertype
00:41:14	Skrylar	should be able to use templates to make it a simple "plug this on to any S&S object" since you don't need *that* much
00:41:15	fowl	i need to be able to do signal.remove(obj)
00:41:38	EXetoC	clovis: in some cases it's obvious anyway, and a more descriptive name might collide with some other type. I don't know how common that'd be for large projects
00:41:42	clovis	fowl, have you looked at libsigc++ ? http://libsigc.sourceforge.net/
00:42:09	EXetoC	Anyway, I'm sure Araq will respond, so I'll see what he has to say about the status of case sensitivity
00:42:10	clovis	EXetoC, me neither
00:42:10	Skrylar	template function for "DisconnectSignals" which uses the seq of backrefs on self, sig.remove(blah) could use that
00:42:17	Skrylar	so there is no supertype, just methods plugged on to the receivers
00:42:18	EXetoC	and maybe I missed some post. let me check
00:42:34	fowl	clovis, no i was looking at qt, boost and http://sigslot.sourceforge.net
00:42:38	Skrylar	fowl: but yeah, i'll poke around with some of the research later for it
00:42:50	clovis	fowl, it's supposedly the fastest c++ implementation
00:43:00	fowl	Skrylar, "seq of backrefs on self" this is provided by the base type (qobject)
00:43:36	Skrylar	yes, but you could do it via templates so people weren't confined to using qobject
00:43:49	Skrylar	they could make an arbitrary object "qobject-like"
00:44:00	fowl	a template requires someone to put the right field: type in their object def
00:44:13	Skrylar	thats a matter of documentation :P
00:44:32	Skrylar	"btw if you don't want to be a child of X, do these two things"
00:45:03	Skrylar	its hardly a gobject level of boilerplate
00:45:57	EXetoC	Skrylar: where did Araq claim that he had changed his mind regarding case sensitivity?
00:46:40	EXetoC	also, we haven't even hit 1.0 yet, so major changes are to be expected
00:47:03	Skrylar	EXetoC: on the forum topic someone asked about it and he said he said partial case sensitivity was more troublesomse than expected
00:47:14	Skrylar	and in IRC he said there won't be any major changes to syntax for a while
00:48:22	EXetoC	not *the* thread apparently
00:48:24	clovis	EXetoC, what do you think about this: a local variable (e.g. windowFrame) would shadow a type with the same name in the local context/scope. And in doing so avoid collision?
00:48:41	EXetoC	well that's not the same thing as forever, but anyway just keep doing things the usual way
00:48:49	Skrylar	i poked firefox but it doesn't have the specific thread up
00:48:57	Skrylar	:\
00:49:10	Skrylar	anyhow, now that i can work on that module again i can go back to getting us a neat GUI
00:49:48	EXetoC	that win32-only GUI?
00:49:56	Skrylar	no, the gl one
00:49:57	fowl	oo lots of file changes today
00:50:18	EXetoC	ok
00:58:05	EXetoC	clovis: that should be fairly bulletproof, but the prefix doesn't really bother me
00:58:15	EXetoC	and in many cases the type of something can be omitted
01:02:07	clovis	A related thought: the compiler should, ideally, never be able to mistake a type for a variable and vice versa. Example: let foo:rectangle. Here rectangle can't possibly be a variable, can it?
01:05:14	clovis	I don't know the syntax enough to say for sure, but I think that the context in which an identifier is used is all that is needed to determine if it is a module, variable or type.
01:05:25	*	lanior joined #nimrod
01:06:27	clovis	If that's the case you could in principle do the following: from rectangle import rectangle ... let rectangle:rectangle = ... without confusing the compiler.
01:07:29	clovis	Of course, you wouldn't!
01:07:47	clovis	But you could safely remove all prefixes
01:09:16	clovis	Humans understand natural language by context. It seems that nimrod's syntax could, in principle, allow the compiler to do the same regarding identifiers
01:09:30	fowl	not really
01:09:38	Demos	also, we want tools to be easy to write
01:09:57	fowl	you cant tell whether rectangle() is a constructor or a variable with the `()` function defined
01:09:58	clovis	Demos, reuse the parser
01:10:14	clovis	fowl, oh.
01:10:15	EXetoC	you should really Capitalize either way, case sensitive or not
01:12:58	*	q66 quit (Quit: Leaving)
01:53:18	*	clovis quit (Quit: clovis)
02:25:07	*	brson quit (Quit: leaving)
03:35:09	*	psquid joined #nimrod
03:38:01	*	Guest23044 quit (Ping timeout: 240 seconds)
04:07:55	*	lanior quit (Quit: Page closed)
04:31:34	Skrylar	meep
04:33:49	fowl	Skrylar, gc works fine with this pattern
04:40:45	Skrylar	fowl: using unref?
04:41:01	Skrylar	and also with or without the backlinks?
04:43:11	fowl	no backlinks, im not sure how to do it
04:43:24	Skrylar	yeah, that will take some thinking
04:45:07	*	mal`` quit (Ping timeout: 264 seconds)
04:45:22	fowl	i decided to allow 1 argument to a signal
04:45:48	fowl	that way i dont need to write 9 overloads, if you need to pass more info pass a tuple/object
04:46:30	Skrylar	hrm, do we not have that weird trickery that lets you pass a tuple in to function parameters?
04:46:45	Skrylar	I remember that was added to D, where you could have a template with '...' as a parameter, and then fill it with a tuple
04:47:44	fowl	yea it works fine https://gist.github.com/fowlmouth/9584728#file-signals-nim-L51
04:49:43	Skrylar	neat
04:51:20	*	mal`` joined #nimrod
05:15:25	*	Skrylar growls at gnuplot
05:15:33	Skrylar	1px lines are really not helpful
05:42:16	fowl	hrm
05:42:50	fowl	i was going to add initialization to seq/string on use
05:43:02	fowl	the compiler code is scary :/
05:43:32	*	Demos quit (Read error: Connection reset by peer)
05:51:45	Skrylar	'tis a scary place
07:05:09	*	xenagi quit (Remote host closed the connection)
07:05:37	*	skyfex joined #nimrod
07:21:05	*	skyfex quit (Quit: Computer has gone to sleep.)
07:22:18	*	vendethiel quit (Remote host closed the connection)
07:52:23	Skrylar	fowl: got the rectangle packer working
08:28:53	*	nolan_d quit (Ping timeout: 240 seconds)
08:39:56	*	lanior joined #nimrod
08:45:07	Skrylar	o_o what an idiot... "upgrading" to SSD because his write cycles are too slow when recording video
08:45:33	Skrylar	just upgrade to a drive that has an exponentially shortened write lifetime instead of using a RAID, genius
08:51:24	*	BitPuffin joined #nimrod
08:58:36	BitPuffin	dom96: why play Civ 5 when we can play FreeCiv
09:01:35	*	noam_ quit (Disconnected by services)
09:02:00	*	noam__ joined #nimrod
09:06:07	*	zahary_office quit (Ping timeout: 264 seconds)
09:19:28	Skrylar	BitPuffin: i don't remember liking freeciv :(
09:19:35	Skrylar	and freeorion had the starlane sadness from MOO3
09:27:24	Araq	Skrylar: I'll let the community vote about the --cs:partial thing. We'll remove T/P either way.
09:29:25	Araq	with --cs:partial I have to go through the list of Babel packages and tell them to use foo not Foo. It's really annoying.
09:31:52	*	CarpNet joined #nimrod
09:33:20	Araq	and what's the point anyway. the people who enjoy fascism are all using Go already...
09:33:31	Skrylar	I've been coding without T prefixes; I think i considered using R instead of P
09:33:39	Skrylar	because P is technically supposed to be "pointer"
09:34:28	Skrylar	and.. yeah, I don't see the value in enforcing case sensitivity. I haven't had significant troubles with casing to see why its a big deal
09:34:35	*	Mat3 joined #nimrod
09:34:49	Mat3	Good Day
09:34:54	Skrylar	'morning.
09:35:17	lanior	Is there any way to enable null pointer dereference checks?
09:36:57	Araq	lanior: type Foo = ref int not nil
09:37:09	Araq	but the stdlib doesn't use it, making it hard to work with
09:37:40	Araq	we'll eventually update the stdlib but it will remain annoying for the guys of us who have complex invariants
09:39:05	Mat3	I want remark that the recent Git version of Nimrod depend on libdl, so compilation need an additional »-t='-ldl'« flag, otherwise ld abort though unresolved references to dlopen, dlclose and dlsym
09:39:42	lanior	Araq: So there is no magic compiler switch to force nil checks?
09:40:03	Mat3	(at least on Ubuntu)
09:44:31	Araq	Mat3: yes. why is that a problem?
09:45:01	Araq	lanior: no, but at least 'not nil' is checked at compile time, not runtime
09:45:08	Skrylar	Araq: Oh, because it doesn't add the -ldl flags properly there.
09:45:21	Skrylar	Back when I had linux on this computer I had to add it to get anything compiled.
09:46:06	Araq	hu?
09:46:36	Araq	I guess the ubu people decided -ldl needs to be at the end for "correctness"
09:47:16	Mat3	Araq: It would be nice at least actualising the documentation, otherwise new programmers which want to participate are forced into unneeded troubles
09:47:36	Araq	Mat3: no we will fix it instead
09:48:04	Araq	but that doesn't mean I don't enjoy ranting about this
09:48:19	Mat3	that's ok
09:49:24	BitPuffin	Araq: doesn't sound like something someone decided, more like, if you need libdl, you should link to it. You can't have the operating system guess what you want to link to, that would be stupid and error prone
09:50:02	Araq	BitPuffin: we do pass -ldl to gcc
09:50:06	Araq	since forever
09:50:30	Araq	we don't pass it "properly" though and gcc decided it should be stricter
09:50:37	Araq	something like that
09:50:55	BitPuffin	I'm not sure what you mean
09:50:59	lanior	Araq: any plans to add optional runtime check? Nimrod already has bundChecks so missing nilChecks looks a bit inconsistent
09:51:44	Araq	lanior: since you are the 3rd guy who wants it, I'll put it on my todo
09:52:02	lanior	Araq: this feature is particularly useful in web server. I dont want to reset all connections instead of serving page 500 to one user
09:52:11	Araq	I know
09:52:38	*	CarpNet quit (Remote host closed the connection)
09:55:00	*	io2 joined #nimrod
09:58:59	*	EXetoC quit (Ping timeout: 240 seconds)
10:02:10	*	CarpNet joined #nimrod
10:02:48	*	io2 quit ()
10:03:22	lanior	Coroutines are still not available?
10:06:55	lanior	if not, what is the recommended way to handle parallel requests (i'm talking about web server). Thread pool? Are there any library support to build DB connection pool or balance requests between threads?
10:08:37	BitPuffin	ping zahary
10:08:40	BitPuffin	only one zahary this time
10:15:43	Mat3	lanior: As I know, a decent coroutine implementation is planned
10:17:08	Mat3	what kind of coroutine concept do you prefer (or which is of most usage for you) ?
10:18:15	lanior	there are many of them?
10:19:14	Mat3	yes, parallel sheduling though process pools of cooperative tasks is one strategy for example
10:19:44	lanior	its just a different scheduling strategies
10:19:45	NimBot	dom96/jester master c88b792 Jason Livesay [+0 ±1 -0]: Add ttyl and finished
10:19:45	NimBot	dom96/jester master fa3a127 Jason Livesay [+0 ±1 -0]: Remove extra comments
10:19:45	NimBot	dom96/jester master 9f2a1eb Jason Livesay [+0 ±1 -0]: Fix comment
10:19:45	NimBot	dom96/jester master a60a4ff Jason Livesay [+0 ±1 -0]: finish as function
10:19:45	NimBot	2 more commits.
10:20:18	Mat3	that is what I asked for
10:20:22	lanior	web workload is primarily IO bound so i am fine with cooperative tasks
10:21:07	Mat3	good to know, thanks
10:21:19	lanior	... unless you are using Django)
10:22:52	Mat3	sorry, I am quite unfamiliar with most aspects of web related programming
10:23:32	lanior	this is likely to be a python problem
10:24:11	lanior	django relies on metaprogramming tricks to provide nice ways to define ORM models and forms
10:24:21	runvnc	Well, I don't know about parallel exactly, but I am using jester and putting long running requests in a redis queue
10:24:24	lanior	they are pretty slow
10:24:28	runvnc	that way if I get really successful
10:24:34	runvnc	I can even service them on multiple VPSs
10:24:47	runvnc	then I pull the data off the redis queue when it is done processing
10:24:50	runvnc	and send it with jester
10:24:54	runvnc	its not exactly parallel
10:25:02	runvnc	but jester is fast
10:25:12	runvnc	and can handle thousands of requests per second that way
10:25:18	lanior	i am not talking about offloading CPU intensive workload to background processes via queue
10:25:25	lanior	yes jester is fast
10:25:57	lanior	its definitely not the fastest but its fast enough
10:26:15	runvnc	my understanding is that message queues are kind of a standard best practice for scaling web servers
10:26:36	lanior	nope its about scaling web applications
10:26:57	runvnc	I had a project last year
10:27:00	lanior	modern web servers can handle more than 500k requests per second
10:27:04	runvnc	which was about scaling web applications
10:27:08	runvnc	lol
10:27:11	lanior	the problem is IO
10:27:13	runvnc	what are you talking about lanior
10:27:17	runvnc	"modern web server"
10:27:50	lanior	anything using appropriate async system API
10:27:58	runvnc	so anyway last year I had a Node.js project, about scaling.. we were taking advantage of the asynchronous IO in node
10:28:36	runvnc	doesnt matter what kind of tricks you pull with non-blocking io, if you go through the kernel you are not going to get 500,000 requests per second
10:28:48	runvnc	anyway we were taking advantage of async io in node, but we still were building it with a queue
10:28:51	runvnc	out of process
10:28:56	runvnc	and that is what twitter and others do
10:29:15	runvnc	I have seen a few articles about a guy who had 500,000 _connections_ on one node server
10:29:21	runvnc	thats different from requests per second
10:29:44	lanior	i guess that was whats up
10:30:18	runvnc	if you really want to push 100000 requests per second you have to go around the kernel
10:30:23	lanior	my toy libuv based webserver handles about 80k requests per second on 1 core
10:30:27	runvnc	most people dont go to the trouble
10:30:51	lanior	system calls are not THAT expensive
10:30:55	runvnc	ok thats news to me
10:31:03	runvnc	can you publish a benchmark of that on a blog somewhere
10:31:07	runvnc	and post to hacker news
10:31:17	lanior	wait a minute
10:31:18	runvnc	because I have not seen numbers like that from 1 core without avoiding the kernel
10:31:21	runvnc	its not a normal thing
10:31:49	Mat3	exist there a way for redefining an object (I mean no instanciation but object replacement) ?
10:31:52	lanior	you cant avoid kernel without using userspace network stack
10:32:08	lanior	which is not what normal people do
10:32:27	runvnc	yeah normal people dont say that 80k requests per second is a normal web server either
10:32:38	runvnc	very interested to see a benchmark comparing your program with "normal" web servers
10:33:24	lanior	http://www.techempower.com/benchmarks/#section=data-r8&hw=i7&test=plaintext
10:35:02	lanior	and my toy app web server: github.com/lanior/plum
10:35:18	lanior	it was built using libuv, joyent http parser and boost coroutine libraries
10:35:41	lanior	just to learn c++ tricks and web server perfomance limitations
10:36:10	lanior	but i am fine with 10-100 RPS
10:36:22	Mat3	lanior: thanks for the link
10:36:58	runvnc	what does the header column mean
10:37:04	runvnc	256 1024 4096 etc
10:38:38	lanior	concurrency requests
10:38:45	Mat3	however: The statistics show a larger difference between both hardware platforms
10:39:17	lanior	Dedicated server is a better choice
10:40:19	runvnc	I don't know where they are coming up with these numbers
10:40:28	runvnc	I can see maybe 10000 requests per second on node
10:40:37	lanior	javascript is slow
10:40:38	runvnc	dont know how they get 80000 never heard a number like that
10:40:59	runvnc	I have not seen numbers like this before
10:41:46	lanior	well V9 is faster than CPython but its targetting browsers so it cant apply heavy optimizations like Java VM
10:42:04	Skrylar	:3
10:42:17	Skrylar	I still find amusement in that the V-engine is using Smalltalk tech
10:42:31	runvnc	the nodel.js benchmark is clustered
10:42:43	lanior	BTW luajit shows better results using the same libuv network library
10:43:39	runvnc	s of Round 7, three Intel Sandy Bridge Core i7-2600K workstations with 8 GB memory each (early 2011 vintage) for the i7 tests; database server equipped with Samsung 840 Pro SSD
10:43:48	runvnc	if they are using 3 servers in clustered mode
10:43:54	runvnc	then I can buy 80000 requests per second
10:44:01	runvnc	so I think you are confused about what is being tested
10:44:06	runvnc	those numbers arent for one core
10:44:27	lanior	they are for one machine
10:44:40	lanior	benchmarks are run from a separate machines
10:45:24	runvnc	http://www.techempower.com/benchmarks/#section=environment&hw=i7&test=plaintext&p=ymz1dq-27
10:45:38	runvnc	says three sandy bridge
10:46:15	lanior	1 Web Server 1 DB server 1 Test Server AFAIK
10:46:34	dom96	'morning
10:46:55	runvnc	the node.js test is clustered
10:46:59	runvnc	and so are the rest of them
10:47:16	Skrylar	morning dom96
10:47:23	Skrylar	lanior: thats because luajit is awesome
10:47:24	runvnc	its not a single core test, doesnt say anyhing like that on the techempower site
10:47:25	lanior	its clustered on one machine
10:47:29	runvnc	yeah
10:47:34	runvnc	so that is using all of the processors
10:47:36	lanior	because nodejs cant use more than 1 core
10:47:37	runvnc	not 1 core
10:47:42	Mat3	morning dom96
10:47:58	runvnc	It can, it does, it uses all the cores, and that explains the 80k requests per second
10:48:06	lanior	oh man...
10:48:10	runvnc	suggesting that those numbers are for one processor is ludicrous
10:48:27	runvnc	lanior I have been coding in Node.js for three years
10:48:39	lanior	Node.js IS SLOW comparing to those web servers
10:48:40	runvnc	In clustered mode, node.js will use all of the processors you ask for
10:48:52	runvnc	did I say it was fast?
10:49:07	lanior	V8 much slower than c++ or Java
10:49:07	runvnc	I am just saying you don't get 80k requests per second from a "normal" web server on one core
10:49:24	lanior	what is "normal" web server?
10:49:38	lanior	Java web servers can five you that speed
10:49:59	runvnc	you seemed to be arguing that a single core, single thread "normal" server should get 80k requests per second
10:50:02	runvnc	which is ludicrous
10:50:06	runvnc	anyway its bedtime for me
10:50:08	lanior	but i am not talking about doing something useful. Just "Hello world" page
10:50:08	runvnc	goodnight people
10:50:15	runvnc	even with hello world
10:50:17	*	runvnc quit (Quit: Leaving)
10:51:45	Skrylar	Alright.. today... hrmm.
10:51:49	Skrylar	Event dispatch sounds important
10:54:08	Skrylar	Araq: any opinion on Qt signal/slots?
10:54:08	lanior	11k RPS using jester
10:54:12	Skrylar	fowl and I were talking about that earlier
10:54:14	lanior	hmm
10:54:35	lanior	does it use epoll on linux?
10:54:39	dom96	no
10:54:45	dom96	not yet
10:54:53	*	Skrylar also prods dom96 as he was interested in GUIs
10:55:02	dom96	it also isn't optimised at all
10:55:08	lanior	and its single threaded?
10:55:16	dom96	Skrylar: never used Qt so I don't know.
10:55:18	dom96	lanior: yeah
10:55:48	lanior	look like i have to compile my toy webserver to get a good baseline
10:55:56	Skrylar	dom96: maybe i should do a writeup of common event systems; but basically, signal/slots is putting callbacks in a queue
10:56:08	lanior	dom96: any way to bypass blocking on database calls?
10:56:09	Skrylar	so you have blah.onResized.connect(stuff) # i am now an event
10:56:38	dom96	lanior: I don't think the current database APIs support that.
10:56:41	lanior	dom96: looks like databases drivers from stdlib doesnt not support async interface
10:57:00	lanior	but we can use thread pool
10:57:21	lanior	this is basically what python/php webserver does
10:57:24	dom96	Skrylar: Isn't that what GTK uses too?
10:57:25	lanior	many processes or threads
10:57:45	Skrylar	dom96: i think gobject/gtk does that, a little clunkier
10:58:03	*	hgsdgbv joined #nimrod
10:58:06	Skrylar	Delphi "kind of" does sigslot, though it does a more java-ish version of it.
10:58:12	Skrylar	Java observers are terrible IMO
10:58:52	lanior	without that database connection will ruin perfomance
10:59:11	lanior	its a big problem if you have more than 5 QPS
11:00:31	hgsdgbv	Which gui lib shall i use? I'm not used to C.
11:00:40	lanior	i can still run multiple processes and do load balancing but that does not sounds good
11:00:50	Mat3	how I override an object member ?
11:06:33	*	easy_muffin joined #nimrod
11:06:50	lanior	dom96: did you see easy way to implement balancing requests in jester between worker threads?
11:07:16	dom96	no sorry
11:07:29	lanior	hard way?)
11:07:32	dom96	async API should be done soon and then things will be easy
11:07:34	BitPuffin	dom96: is the new async stuff you did in tfm?
11:07:44	dom96	BitPuffin: tfm?
11:07:50	BitPuffin	dom96: as in rtfm
11:07:51	dom96	BitPuffin: oh btw, FreeCiv sucks.
11:08:05	BitPuffin	dom96: oh btw, YOU suck
11:08:09	dom96	BitPuffin: It's in the stdlib if that's what you're asking.
11:08:23	BitPuffin	dom96: no
11:08:25	BitPuffin	tfm
11:08:31	lanior	dom96: you mean c# like async?
11:08:46	dom96	yes
11:08:57	dom96	BitPuffin: It's not in the manual and it won't be.
11:10:35	BitPuffin	dom96: oh so it's a lib thing
11:10:39	BitPuffin	I thought it was language stuff
11:10:42	BitPuffin	since it looked like it
11:10:45	BitPuffin	guess it's macros
11:10:48	lanior	looks like it converts function to promise-based variant
11:11:22	Araq	Mat3: use 'method' to override, but I'd use a proc var instead
11:11:42	Araq	I want to remove 'method' for nimrod v2 :P
11:12:50	Araq	hgsdgbv: we only have gtk and iup bindings afaik
11:13:45	lanior	dom96: using it would require wrapping async proc with async macro right?
11:13:59	dom96	yes
11:14:00	hgsdgbv	Yeah i was looking at gtk2 doc. Btw, why are you removing methods?
11:14:18	hgsdgbv	Or what will be the alternative?
11:14:26	Araq	because I want the macro system to generate the dispatchers
11:14:36	Araq	more flexible and faster on x86
11:14:58	lanior	but who uses x86?
11:15:04	hgsdgbv	Me
11:15:28	BitPuffin	dom96: I was just joking, you're a jolly good fella
11:17:07	hgsdgbv	Bear with me: i want to ¿instantiate? this: PWidget* = ptr TWidget
11:17:09	lanior	dom96: all the way down to the event loop callback?
11:17:19	hgsdgbv	Do i have to use new(PWidget) or something?
11:18:06	dom96	lanior: The macro handles the low-level details.
11:18:10	Araq	lanior: btw the database APIs are mostly designed for convenience. for performance you likely have to use the native APIs
11:19:44	lanior	Araq: thread blocking API is basically unusable without big thread pool
11:20:55	Mat3	Araq: thanks
11:24:05	Araq	lanior: so fix the wrapper to support asynio :P
11:24:43	Araq	it's not like we're opposed to making the harder stuff possible
11:25:45	lanior	Araq: having decent ORM seems to be a more important issue
11:26:16	lanior	Araq: and i'm trying to build one now.
11:26:37	Araq	actually I want to make a "decent ORM" as a commercial project ...
11:26:38	lanior	Araq: so far I am impressed with syntax I can get using macro system
11:27:04	lanior	Araq: commertial ORM wont make language popular
11:27:12	Araq	bah
11:27:24	Araq	commerical anything is bad for popularity
11:28:10	Mat3	Araq: I have an object type A which hold only state variables (a structure) and and object B with a member of these type C. Now I want to instanciate B and override C. Is this possible ?
11:28:21	Mat3	^an
11:28:39	*	EXetoC joined #nimrod
11:29:28	Araq	what is "this type C"?
11:29:33	hgsdgbv	With member you mean methid?
11:29:35	hgsdgbv	*method
11:29:48	Araq	you say A is struct-like and B inherits from C
11:29:56	Araq	typo?
11:29:56	Mat3	C is of type A, sorry
11:31:38	Araq	Mat3: just check the tutorial for 'method'
11:32:04	Araq	I can only repeat the examples here
11:32:50	Araq	hgsdgbv: in general 'ptr Foo' cames from C, so you can't 'new' it, you need to use the C function to create the object
11:33:14	Araq	bbl it's lunch time here
11:33:20	hgsdgbv	Thanks, i saw in a example i had to use "gtk_window_new"
11:33:28	Skrylar	I've not succeeded in getting IUP to build with GCC :(
11:33:36	Skrylar	Unless they changed it, since IUP uses some archaic windows headers
11:35:18	hgsdgbv	btw how can i get a list of args passed to nimrod?
11:40:19	*	EXetoC quit (Ping timeout: 264 seconds)
11:40:37	Skrylar	I've not actually tried.
11:41:37	hgsdgbv	That's a good thing to know
11:42:07	Mat3	I get it, thanks
11:42:35	lanior	hgsdgbv: http://forum.nimrod-lang.org/t/84
11:43:07	hgsdgbv	Ah! thanks!
11:43:29	lanior	forum really needs build-in search
11:43:40	lanior	or maybe just google search widget
11:43:44	hgsdgbv	That is from 2012, it is still neccessary to import os ?
11:43:52	lanior	idk
11:43:54	lanior	probably
11:44:35	lanior	why asyncio2.nim is located in lib/pure?
11:44:56	Skrylar	could always do the forum search as a call out to node.js with lunr
11:45:01	Skrylar	that would be heresy though :o
11:48:19	hgsdgbv	But a practical one.
11:48:32	lanior	we can use ElasticSearch
11:48:37	lanior	it has simple json api
11:48:52	lanior	and it is node.js free!
11:49:37	*	EXetoC joined #nimrod
11:51:14	*	clovis joined #nimrod
11:51:20	hgsdgbv	"could not load: libgtk-win32-2.0-0.dll" I suppose i have to download this
11:51:47	clovis	Sounds reasonable
11:51:53	Skrylar	Nimrod doesn't come with the gtk runtime
11:52:35	hgsdgbv	Oh joy
11:52:48	Skrylar	Wx probably would have been a better library for us to have than GTK, but oh well
11:53:02	EXetoC	yes. it doesn't come with any of the C components
11:53:48	hgsdgbv	Maybe ... in a future?
11:54:04	Mat3	well just copy the required DLL to your system directory :->
11:54:12	Skrylar	o_o
11:54:23	Skrylar	the gtk installers put it in your path you know
11:54:28	Skrylar	no need to shove random things in system
11:55:59	Mat3	I know, but its more fun to accumulate them in the global system directory
11:56:02	hgsdgbv	What if i want to distribute a binary that uses gtk2, there is an "easy" (looks with escepticism) way to distribute the dlls?
11:56:38	lanior	put them in the dir with executable file
11:57:16	hgsdgbv	Thanks, that sounds easy.
11:57:18	lanior	this should work in most cases
11:58:24	lanior	how tightly nimrod is bound to C standard library?
11:58:36	hgsdgbv	Humm the installer tutorial says nothing about installers that let you choose the path, besides manual action.
11:59:20	*	EXetoC quit (Ping timeout: 246 seconds)
12:00:36	Skrylar	no, its the GTK for windows installer
12:00:55	Skrylar	the usual way that GTK programs do it on windows is to tell you to run the GTK installer, or they include one, but its a large runtime
12:01:06	Skrylar	GTK is sort of a sub-optimal toolkit on Windows tbh :(
12:01:32	clovis	GTK is suboptimal anywhere
12:01:39	hgsdgbv	lol
12:01:43	hgsdgbv	Do i have a choice?
12:02:06	lanior	GTK looks so terrible on windows
12:02:23	lanior	QT feels slightly better
12:02:48	Skrylar	Qts not really usable from nimrod with any sanity
12:03:15	Skrylar	whatwith the whole MOC setup
12:03:55	lanior	but its usable from python
12:04:58	hgsdgbv	How many years have nimrod been developed?
12:04:58	clovis	How feasible is it to use Qt Quick / Qml with nimrod?
12:05:07	hgsdgbv	*in development
12:06:42	Skrylar	clovis: well the biggest problem with using Qt has to do with making sure the MOC macros are in the right places
12:07:09	Skrylar	Unless you find a way around it (@lanior)
12:07:35	BitPuffin	well it looks like zahary did something but I'm still getting an error
12:07:46	BitPuffin	proc `*`*[T; R, N, C: static[int]](a: TMatrix[T, R, N], b: TMatrix[T, N, C]): TMatrix[T, R, C] {.noSideEffect.} =
12:07:53	BitPuffin	matrix.nim(121, 49) Error: type expected
12:07:57	clovis	Skrylar, I see.
12:09:30	clovis	hgsdgbv, at least 3 years
12:09:54	clovis	indicated by first github commit
12:13:45	clovis	Oh. "Copyright (c) 2004-2014 Andreas Rumpf"
12:15:44	clovis	First github import was Jun 22, 2008
12:16:04	clovis	5 years, 8 months, 25 days
12:20:38	*	lanior quit (Ping timeout: 245 seconds)
12:23:29	*	lanior joined #nimrod
12:23:41	clovis	Someone who is interested in GUI development and Qt could look into how PyQt5 is implemented. If Python can use Qt (bypassing the MOC and the object model of C++) then should Nimrod.
12:24:35	lanior	or you can use it via python binding
12:24:54	clovis	Perhaps
12:25:17	hgsdgbv	And i used to hate swing
12:25:23	lanior	btw who knows state of python bindings?
12:25:46	clovis	"The latest iteration of PyQt is v5.2. It fully supports Qt 5.2.0"
12:25:47	lanior	i saw discussions about fluent property access syntax for pyobjects
12:26:10	Skrylar	I'm working on a GUI system, albeit its not a qt interface
12:26:17	Skrylar	we were talking earlier about adapting signals/slots to nimrod
12:26:38	lanior	Skrylar: custom drawn?
12:26:47	clovis	Skrylar, nice. Signals/slots is a good foundation
12:26:52	Skrylar	yeah, GL-backed
12:27:05	Skrylar	I was considering a WinAPI or Cairo variant a few times
12:27:32	lanior	whats about users without OpenGL support?
12:27:34	Skrylar	there's a lot of GUI that is agnostic to how it gets drawn, so its not hard to separate out the draw code to the rest of it all and have multiple versions
12:27:38	clovis	WinAPI - is it Windows only?
12:27:47	Skrylar	yes, winapi is windows native
12:27:56	clovis	I meant your library
12:28:12	Skrylar	no, any winapi code i put in is isolated
12:28:39	lanior	firefox even uses opengl to directx translator on windows because WinXP supports only OpenGL 1.1 out of box
12:29:00	Skrylar	i've been working on foundation code so far (glyph caching, texture atlases, rectangle management) so the platform stuff hasn't been reached yet
12:29:03	BitPuffin	the python qt binding probably exists because holy shit does python have a userbase
12:29:12	BitPuffin	it's not like it isn't possible in nimrod
12:29:15	BitPuffin	it's just a big task
12:29:35	lanior	and I guess you use modern OpenGL pipeline, so thats could become a problem
12:29:54	Skrylar	at the moment there is zero platform code, so i don't use any gl pipeline
12:30:01	Skrylar	i had planned on hiding the gl pipeline under a support lib
12:30:14	BitPuffin	lanior: what, it does support more than 1.1, as long as you have drivers
12:30:18	BitPuffin	you talkin cray cray
12:30:26	Skrylar	he said 'out of box'
12:30:29	lanior	BitPuffin: but many people doesnt have it
12:30:30	BitPuffin	well yeah
12:30:33	BitPuffin	but if you don't have drivers
12:30:40	BitPuffin	then you don't really have d3d either amirite?
12:30:51	Skrylar	I planned on doing a simple graphics layer like most game engines do; mostly involving shoving VBOs together
12:30:52	lanior	we are talking abound 2d UI
12:31:35	clovis	Build upon sdl2?
12:31:38	lanior	Skrylar: you can just use existent 2d drawing library
12:31:41	lanior	like skia
12:31:53	Skrylar	isn't skia 1) chrome's pet rasterizer and 2) C++
12:32:04	BitPuffin	Skrylar: to me it seems like d3d even requires MORE setup than opengl
12:32:16	Skrylar	AFAICT you can't set nimrod to only partially support C++, if any one C++ lib gets in then you *have* to use it for the whole project
12:32:23	BitPuffin	considering every time you run a fucking game it's like, hey, I need to set up directx ;D
12:32:58	BitPuffin	Skrylar: that is correct
12:33:04	Skrylar	I'm not hugely worried about people being unable to run an OpenGL program; mostly because the apps that I plan to use it with don't target computers that have been unpatched since 2001
12:33:08	BitPuffin	however there is no reason we couldn't work our way out of that one
12:33:16	BitPuffin	just that it's something that hasn't been done yet
12:33:27	Skrylar	Though the design I plan on using allows someone else to change the graphics code, so if they want to use Not GL, they can do so
12:33:42	Skrylar	I mean, 90% of a GUI is *not* the drawing
12:33:59	Skrylar	its dealing with text layouts, control layouts, usability triggers
12:33:59	BitPuffin	exactly: where the hood at
12:34:21	BitPuffin	bah, who needs em
12:34:30	clovis	The part that is ugly with gtk :|
12:34:35	Skrylar	I would be happy to have a Cairo+Pango backend, if it wasn't for Pango having the worst dependency in the world
12:34:50	lanior	but you wont get "native" user experience
12:35:16	Skrylar	so make a winapi backend? or ask windows to draw the controls?
12:35:20	Skrylar	Qt just.. asks windows.. to draw it
12:35:21	BitPuffin	Skrylar: what's the dep?
12:35:34	Skrylar	There's an API for asking the theme system in windows to rasterize components for you
12:35:38	lanior	yes it uses theme api
12:35:50	lanior	but its not about drawing but about user interaction
12:36:02	Skrylar	then those people can use the win32 backend instead of the GL backend?
12:36:06	lanior	like scroll behavior on mac
12:37:00	Skrylar	my design is not hugely different from a 2D scene graph; theres some helpers for doing layout control, and probably helpers so people don't do button.onClicked(..) crap but instead centralize action code
12:37:26	clovis	You had better go 4D from the start
12:37:41	Skrylar	what good is a time dimension going to do? :)
12:37:47	clovis	;)
12:38:20	Skrylar	(fwiw, i was going to name one of the programs using it 'Tesseract' because it was a multidimensional outliner, thought the name would be appropriate)
12:38:38	BitPuffin	isn't tesseract that fork of sauerbraten?
12:38:41	BitPuffin	2
12:39:33	Skrylar	i didn't know there was one
12:39:44	Skrylar	i stole it from a minecraft mod who stole it from a 4D mathematical object
12:40:19	clovis	Skrylar, perhaps the only public API could be the cetralised one?
12:41:51	Skrylar	clovis: maybe, though i don't want people to complain that it gives them no control
12:42:09	Skrylar	i still have to iron out the design docs for it all
12:42:44	Skrylar	but I ultimately want people loading their GUIs from data files and only interacting with the GUI through a series of actions, because it ensures buttons/menu items have consistent labels as well as makes undo/redo/recording user actions much simpler
12:42:47	clovis	Declare Button. Declare Action. Connect Button to Action. Connect Action to Program
12:42:55	Skrylar	yep
12:43:04	Skrylar	Qt half-supports this, but I want it to support it more
12:44:28	Skrylar	i need to benchmark resizing anchors against unified coordinates
12:44:58	Skrylar	CEGUI does this thing where each coordinate point has a relative to parent and absolute point, which makes it super easy to put a box "at the top left, always 25% of the parent width"
12:46:03	*	hgsdgbv quit (Ping timeout: 245 seconds)
12:46:08	clovis	Closures that are run automatically on resize
12:46:56	Skrylar	Traditional GUIs have anchors/fill rules; I'm thinking about benchmarking the complexity and time for both and seeing which one is best. The good thing with unified is that they basically replace both anchors *and* fill rules entirely within the coordinates, so there is no longer a concept of "i fill this dimension" but rather its coordinates naturally specify that they do so
12:47:16	Skrylar	but this requires a multiply+add for every coordinate
12:47:48	Skrylar	parent.x*relative+absolute
12:47:56	clovis	Skrylar, you also have to deal with constraints like maximum and minimum size
12:49:13	Skrylar	clovis: i know, one step at a time
12:49:37	Skrylar	I want to do this right, because it feels like most GUI libs suck pretty badly
12:49:45	clovis	Perhaps this is all one big constrain-satisfaction problem
12:50:45	clovis	All the declarations are constraints like obj A must be 25% of obj B. Obj B must be at most 30% of obj C...
12:51:27	clovis	There are mathematical models and algorithms that deal specifically with generalised, linear constraint satisfaction problems
12:51:37	Skrylar	well with cegui's unified coordinates, if i tell a widget that its 30% of the width, i do so by telling it to be (0,0) (0,0) (0.33,0) (1,0)
12:51:47	Skrylar	It will always be 33% the width of the parent, regardless of size
12:52:22	clovis	But what if it has a constraint to be at least 200px
12:52:28	*	EXetoC joined #nimrod
12:52:30	clovis	`Then what do you do?
12:52:48	*	akghf joined #nimrod
12:52:55	Skrylar	I would probably suggest that the GUI be designed more like CSS4+ flexboxes, so that if the width is violated it changes to stacked tabs
12:53:20	Skrylar	Users get annoyed when they can't make their windows the size they want, because it decided it won't obey
12:53:44	Skrylar	I say its probably better that the application cope in a meaningful way, because its the job of a computer to obey
12:54:08	Skrylar	such as reorienting the buttons in the window, changing from a grail-style layout to stacked tabs, using icons instead of text if its really small
12:54:46	clovis	That requires conditionals, if-statements?
12:55:08	Skrylar	yes, you'd have to check the constraints on component resize
12:55:08	clovis	E.g: if x<200px then use layout-B
12:55:20	Skrylar	however you can hide most of this with precooked layout widgets, i think
12:56:44	Skrylar	I suppose as a last resort, you could make it so that generic widget contains enforce minimum sizes by making scrollbars appear
12:56:56	Skrylar	so if you have a data entry forum that just really can't degrade further, it scrolls instead
12:57:07	clovis	Yeah, usually
12:57:28	Skrylar	what i hate is when i see programs that decide to have a minimum width and will not let me shrink them, for no apparently good reason even
12:57:49	clovis	Another neat thing - which I have not actually seen - would be to rescale a surface (gl rescale the pixels)
12:57:56	akghf	How do i get a pointer from whatever?
12:57:59	clovis	to shrink a surface
12:58:22	EXetoC	akghf: addr
12:58:31	akghf	Nice, thanks
12:58:32	EXetoC	in whichever context it makes sense
12:59:04	Skrylar	clovis: thats a thing that a GL interface would probably be better at than a CPU one
12:59:19	Skrylar	since you can fiddle around with scale cheaper
12:59:39	Skrylar	a big reason that i want GL is 1) for game GUIs and 2) because CPU-based diagram software tends to laaaaag
12:59:42	clovis	Skrylar, suppose the widget has a min with of 200px and you resize it to 100px. Perhaps the GPU could do it for you in real time
13:00:06	Skrylar	clovis: i'll put in a note for that
13:00:08	EXetoC	yeah just update the coords
13:00:28	EXetoC	in some situations at least
13:00:49	Skrylar	one thing that i take inspiration from is "what does a pro gui actually *need*?"
13:00:52	clovis	EXetoC, yeah you might want to get the global mouse coordinates translated
13:01:10	Skrylar	because a lot of GUI toolkits feel like they stop after they give you 30% of what a massive program does
13:01:34	Skrylar	like how many DON'T have a text editing component?
13:01:39	Skrylar	despite the amount of programs that need them
13:01:47	clovis	Right
13:02:00	EXetoC	not that I have used many widget toolkits, but MyGUI is pretty nice
13:02:15	Skrylar	I forget what problem I had with that toolkit
13:02:22	Skrylar	I think I had weirdly low performance in some of the demos
13:02:30	clovis	The thing I hate the most about text editing is when the blinking cursor becomes invisible while you move it with the cursor keys
13:03:01	Skrylar	but yeah. i need to figure out how to deal with backlinks tomorrow, since fowl wrote a simple signal/slot module
13:03:22	akghf	I wrote the following sentence: let args = [0..(paramCount - 1), string] and i receive "type mismatch, got (proc(): int, int literal(1))"
13:03:27	Skrylar	comparing the merits of an anchor/constraint system to just using unified coords is also a todo item; CEGUI did an interesting thing there
13:03:51	akghf	Is because i did not use () on paramCount ?
13:03:59	EXetoC	yes
13:04:04	clovis	My advice for you would be to view GUI lay out from a general constraint satisfaction perspective.
13:04:30	akghf	Humm, it always happens or just inside array definition?
13:04:37	EXetoC	always
13:04:49	EXetoC	akghf: the property syntax can only be used when the proc takes at least one arg
13:04:59	EXetoC	x.val
13:05:34	Skrylar	clovis: i've been looking at it as a simple draw tree system that just obeys; IMO time spent dealing with GUI code is wasted time
13:05:45	akghf	aaaah, that explains a lot. Si, if there does not exist a paramCount(arg1), you cant use that syntax?
13:05:47	EXetoC	where val might be defined as "proc val(x: Type)"
13:05:49	Skrylar	it should just do what you want it to (e.g. show the interface) instead of require 500 special knobs
13:06:10	akghf	*so
13:06:18	EXetoC	akghf: exactly. it's a reasonable limitation
13:06:21	Araq	hi akghf welcome
13:06:25	akghf	Thanks again EXetoC.
13:06:35	akghf	Hello there Araq.
13:06:44	Araq	Skrylar: interesting. I thought about an opengl ui toolkit too.
13:06:58	clovis	Skrylar, perhaps I got the wrong impression. You're perhaps writing a library at a lower level - closer to the hardware
13:07:08	Araq	then I tried to make a sweet lazarus interface and failed :P
13:08:02	Araq	should have listened to the guy years ago who told me to build nim on top of FPC ...
13:08:15	Skrylar	FPC is pretty nice
13:08:26	Skrylar	clovis: its an indeterminate level for the time being; as i mentioned earlier, still in design docs
13:08:30	clovis	What䳠FPX?
13:08:36	clovis	FPC?
13:08:45	Skrylar	I'm working on low-level stuff while assembling concepts and weighing decisions
13:08:49	Araq	it's still reasonably easy to build the UI with lazarus and link nimrod to it
13:09:05	Araq	the repo even had an example but somehow we lost it
13:09:20	EXetoC	Araq: did you say before that you wanted to change the convention from TType to Type, while keeping case insensitivity?
13:09:37	Araq	that's what I said, yes
13:10:02	Skrylar	measuring simplicity is going to be a nightmare
13:10:15	Skrylar	i'm not sure how to compare two modules to see which one that does the same job is "simpler"
13:10:31	Araq	Skrylar: than ask me :P
13:10:35	Araq	*then
13:11:10	akghf	How do i resolv the error: type mismatch: got (typedesc[string]) but expected 'TSlice[int]' in "let args = [0..(paramCount() - 1), string]"
13:11:30	Araq	clovis: FPC = Freepascal compiler, Lazarus = delphi like environment
13:11:45	Araq	produces a native UI on macosx, linux, windows
13:11:48	akghf	wow, works even on gba
13:11:56	clovis	Interesting
13:12:02	Araq	with a decent UI builder
13:12:05	EXetoC	Araq: won't this increase the occurrence of symbol shadowing?
13:12:38	Araq	EXetoC: looks like we more disambiguation rules anyway for module vs. other symbol
13:13:00	Skrylar	yep. I was an FPC+Lazarus derp for a while
13:13:12	Skrylar	I don't like Lazarus though.. the IDE is nice, but the LCL is a pain
13:13:26	Skrylar	Its just a clone of Borland, so all the annoying crap is still annoying and writing custom widgets is unfun
13:13:55	clovis	It sounds likely that the name of a module would be the same as that of its defining type, in many cases.
13:14:13	Araq	actually the convention is to add an 's' to the module name
13:14:20	Araq	tables, sets, etc.
13:14:28	clovis	That's quite good
13:14:39	clovis	Had not thought about that
13:14:42	Araq	rectangles vs rectangle
13:14:48	Araq	solves the problem
13:15:44	EXetoC	Araq: how will foo() for example be disambiguated?
13:16:02	Araq	how can foo() be a module name?
13:16:08	EXetoC	what about types?
13:16:29	EXetoC	type constructors for example, vs call operators
13:16:58	Araq	these simply shouldn't have the same name createFoo, newFoo, allocFoo vs Foo
13:17:12	Araq	like it is today
13:17:17	clovis	I think there are languages where the ctors do not share the name of the type
13:17:36	*	noam__ is now known as noam
13:19:50	clovis	As long as a global identifier's name is shadowed by a local declaration I'm happy. E.g. Global type WindowFrame is shadowed by: local var windowFrame:Rectangle = ...
13:19:53	EXetoC	akghf: the error is on another line, right? 'args' ends up being a typedesc
13:20:11	EXetoC	akghf: you might want : rather than =
13:20:15	clovis	This way you can drop the T from TRectangle and TWindowFrame
13:20:46	EXetoC	but that's an array with a runtime size. does that work?
13:20:59	EXetoC	I guess we'll find out soon
13:21:02	akghf	I want to iterate over the params and insert them into the array.
13:21:10	Araq	clovis: I think what you mean already is implemented but it's inherently complex to this properly thanks to macros
13:21:27	clovis	Oh
13:21:27	EXetoC	akghf: just replace the = with a :
13:21:35	akghf	i'll try.
13:22:03	Araq	var s = newSeq[string](paramCount())
13:22:19	Araq	for i in 0 .. < paramCount(): s[i] = paramStr(i)
13:22:31	Araq	I don't think you can use a shortcut for that
13:23:07	EXetoC	maybe when we get that lambda macro
13:23:23	Araq	you may want to use the parseopt2 module which everybody dislikes instead
13:23:35	Araq	which makes you write string cases and so it sucks
13:24:10	Araq	much better would be a table[string, closure] instead.
13:24:52	Araq	this way you can have the same syntax but slightly slower code which adheres to the nonsensical "switch statements are bad" rule
13:26:56	*	Araq wonders if he is hard to follow
13:27:55	akghf	Nice. I did it with seqs, thanks. Now, i have to pass a pointer to gtk's init method
13:28:17	akghf	addr did not do the work
13:28:32	EXetoC	Araq: nonsensical indeed..
13:28:42	EXetoC	table[string, string] would be more reasonable
13:29:55	EXetoC	https://github.com/docopt/docopt
13:30:51	EXetoC	or maybe values that are variants
13:43:07	akghf	how can i get a "pointer" from an object? using addr gives me "ptr seq", not "pointer"
13:43:29	*	Mat3 quit (Ping timeout: 255 seconds)
13:43:42	OrionPK	is thre a way to quit an iterator early?
13:43:47	EXetoC	akghf: it won't matter in many cases currently, but you can just cast it to a pointer
13:43:49	OrionPK	return / yield break
13:44:03	EXetoC	return?
13:44:21	OrionPK	"return not allowed here"
13:44:28	EXetoC	oh
13:44:32	akghf	it is because the compiler is complaining. How can i cast it? with Pointer(addr whatever) ?
13:44:53	EXetoC	akghf: actually, try x.addr.pointer
13:45:10	EXetoC	and if that fails, try cast[pointer](typedPtr)
13:45:16	Araq	that's a type *conversion* though
13:45:20	EXetoC	OrionPK: just break then I guess
13:45:29	akghf	Thanks EXetoC again, you are most helpful
13:45:31	Araq	casting looks like cast[pointer](addr x)
13:45:39	clovis	yield exits from the iterator, break from the loop.
13:45:41	OrionPK	invalid control flow: break
13:45:47	Araq	but surely a type conversion works here
13:46:02	clovis	OrionPK, break from the loop not the iterator
13:46:29	Araq	OrionPK: there is no alternative to a nested 'if', I think
13:46:32	OrionPK	yeah.. basically that's my only option I guess. a "yield break" would be nice
13:47:38	clovis	I don't think it's the iterators responsibility
13:48:09	OrionPK	iterators can have logic too
13:48:27	clovis	They should be able to be reused elsewhere
13:48:28	EXetoC	it's always the responsibility of the iterator to determine the *max amount* of values to yield
13:48:49	BitPuffin	Araq: will we ever get rid of the inconsistency between named params and constructors?
13:48:51	EXetoC	and then the actual user can break earlier if necessary
13:48:59	clovis	They shouldn't need to know how the calling loop is implemented. Perhaps there is more code in the loop that must be executed
13:49:07	akghf	How can i get an address from an integer?
13:49:09	OrionPK	EXetoC yes but based on logic during iteration, there might be fewer of something
13:49:10	akghf	Or create one
13:49:14	OrionPK	so breaking early is necessary
13:49:17	BitPuffin	akghf: addr theint?
13:49:40	akghf	that gives me "expression has no address"
13:49:52	BitPuffin	wait
13:50:02	BitPuffin	are you trying to get the addr of 1 + 3 for example
13:50:04	BitPuffin	or a variable
13:50:16	akghf	From a var
13:50:17	EXetoC	akghf: do you really need that? if so, then make sure that the integer is a var. 'let' won't allow you to take the address of it
13:50:25	akghf	aa, ok
13:50:26	akghf	i used let
13:50:53	BitPuffin	EXetoC: wasn't there a time where it did let you get address of a let?
13:51:02	EXetoC	OrionPK: I think I implied that. I'm just saying that both the iterator and the caller can break whenever
13:51:11	BitPuffin	or wait
13:51:16	Araq	akghf: the gtk module has 'nimrod_init' that you should call to setup the command line for gtk
13:51:21	BitPuffin	it's that you are allowed to change what's the value of a pointer
13:51:23	OrionPK	right.. but the iterator cant unless it's got an inner loop
13:51:24	Araq	no need to do it yourself
13:51:30	BitPuffin	so if you have let foo = PBar()
13:51:39	BitPuffin	you could do foo[] = ...
13:52:06	Araq	and you need to use toCStringArray anyway otherwise
13:52:09	clovis	In my opinion. The iterator should not be able to control when and how the loop is exited. It's not its responsibility. The calling loop might have to do other operations during the current iteration before exiting.
13:52:12	akghf	My god, at last i see a fucking window.
13:52:50	BitPuffin	Araq: are we ever gonna have a recursive let? which would force immutability even if you dereference or access an array or something whatever? :P
13:53:02	*	BitPuffin coughs at dom
13:53:07	EXetoC	clovis: you're breaking implicitly anyway, whenever the iterator is done
13:54:00	clovis	EXetoC, suppose you're using an iterator defined in a library. How could the library author possibly know all the use cases?
13:54:32	clovis	All the iterator can say is: hey, i'm done here. Then it's up to the loop to decide what to do
13:54:33	Araq	BitPuffin: we'll get something better instead. A 'discard' pragma that checks the implementation is only allowed to consist of a single 'discard' statement
13:54:54	EXetoC	clovis: I didn't imply that it should know anything about the caller
13:55:13	clovis	EXetoC, then I must have misunderstood you.
13:55:24	BitPuffin	Araq: is that even related?
13:55:37	BitPuffin	I think you are joking
13:55:42	BitPuffin	it's always hard to tell with you
13:55:49	BitPuffin	dom96: I was coughing at you
13:55:51	BitPuffin	pay attention damn it
13:55:59	Araq	I'm joking, yeah
13:56:00	BitPuffin	dom96: write an article about the async stuff, naoooo
13:56:18	BitPuffin	Araq: :)
13:56:23	Araq	BitPuffin: we'll get write-effect tracking which amounts to the same
13:56:53	BitPuffin	Araq: will this help out with the way that C interferes?
13:57:05	Araq	what?
13:58:04	BitPuffin	Araq: ie like when you mark a proc with noSideEffect, and you call a C function which you know in your heart causes no side effects to your program so you still want the proc to be considered noSideEffects by other procs that are marked with noSideEffect?
13:58:48	clovis	I've never know anything this profound in my heart
13:58:55	BitPuffin	like a {.noMrCompilerYouDoNotNeedToCheckForSideEffectsBecauseYouWillThinkThatThereAreButThereReallyAreNotAnyIPromisePleaseBabyDoNotHurtMe.} pragma
13:59:33	BitPuffin	guess it's not entirely related
13:59:54	BitPuffin	I guess what you mean is that with the write tracking let will actually enforce immutability because it tracks if something writes to it?
14:01:23	clovis	That's a nice pragma, though
14:01:48	BitPuffin	well I guess while I wait for zahary to rise from the dead and my colleague to write some code so that I can deploy it I'll rtfm
14:01:55	BitPuffin	tfgfm
14:02:02	BitPuffin	tfgm*
14:02:11	EXetoC	clovis: the iterator always controls when the loop is exited, and so 'break' inside the iterator wouldn't make a difference. besides, break is nothing more than a shorter/faster way of exiting from a loop
14:02:40	EXetoC	A loop is probably the most common way of implementing an iterator, but it could be anything really. 5 yields in a row for example
14:03:14	EXetoC	did I misunderstood *you*?
14:04:23	clovis	EXetoC, I see the iterator as a function to be called by the loop. Therefore the loop has the final say. Perhaps it has to do something else before exiting?
14:05:56	EXetoC	all break inside the iterator would do is possibly result in fewer values yielded. that doesn't change anything in this regard.
14:06:13	clovis	Right
14:06:26	EXetoC	and I don't know why you'd want to do something just before an iterator exits, but there's currently no way to know if you are at the end anyway
14:07:45	clovis	I think you're right. I didn't think it through clearly enough
14:12:53	dom96	BitPuffin: Weren't you supposed to have a blog done like months ago? :P
14:12:54	EXetoC	I'll be experimenting with a library-based alternative to 'iterator' soon, similar to D's ranges, which means you'll be able to query the state
14:13:21	EXetoC	dom96: did you try the lambda macro recently? :>
14:14:11	EXetoC	hopefully we can have it ready for 0.9.4
14:14:41	clovis	EXetoC, lazy evaluation also?
14:14:44	*	Varriount joined #nimrod
14:15:02	dom96	EXetoC: no but zahary fixed that bug so I will try it ASAP
14:15:35	Araq	Varriount is back! :-)
14:16:01	Araq	Varriount: you're on the criticial path now. are you happy?
14:16:46	Skrylar	Hrm. I saw the HN discussion about Nimrod getting AFD'd on wikipedia
14:16:49	EXetoC	dom96: yay
14:16:53	Skrylar	Just looked at the Buzz page again
14:17:07	Skrylar	they have a list of "buzz clones" that are all unfinished except for one
14:17:24	BitPuffin	dom96: shh
14:17:25	Skrylar	they sure do a great job curating those non-notable projects don't they -_-
14:17:26	Varriount	Araq: Sorry, I've been taking a break, playing some games, etc.
14:17:36	BitPuffin	dom96: well I can't fucking finish the blog because of this mac os x bug with redirect
14:17:39	BitPuffin	so that's my excuse
14:17:48	*	darkf quit (Quit: Leaving)
14:18:02	BitPuffin	and how am I supposed to be able to write an article about YOUR async stuff ;D
14:18:18	dom96	You're not. You're suppose to write an article about Nimrod.
14:18:27	dom96	Skrylar: buzz page?
14:18:41	BitPuffin	well yeah when zahary fixes the bugs that are currently holding linagl back from it's new release I will
14:18:45	BitPuffin	and it's not far off
14:18:47	BitPuffin	the code is there
14:18:50	Araq	Varriount: please add that nimbuild feature. I depend on it
14:18:51	BitPuffin	just not the compiler :P
14:19:47	EXetoC	clovis: they are lazy by definition I think
14:21:48	EXetoC	clovis: you can pass it around however you like, and the range will be exhausted only after popping a certain amount of times, and that can be hidden using some foreach macro
14:21:56	Varriount	Araq: Testing bootstrapping you mean?
14:22:10	*	Demos joined #nimrod
14:24:42	Araq	no the "still failing tests" features
14:25:10	EXetoC	I can't remember what I was going to call it though. I don't think range is a good name, and it's occupied already
14:25:34	Skrylar	dom96: the wikipedia entry for jeskola buzz
14:25:39	Skrylar	its an old tracker
14:26:07	Skrylar	but they have out-links to a lot of abandoned code :|
14:27:27	Araq	EXetoC: closure iterators already provide what you seek (you can pass them around)
14:27:31	*	psquid quit (Ping timeout: 264 seconds)
14:27:57	*	psquid joined #nimrod
14:28:01	Araq	but a "cursor" might be a good name, perhaps?
14:28:19	Araq	that's what a c++ iterator is anyway
14:28:32	Araq	a D-styled range is a cursor_pair
14:29:16	*	Skrylar might write a tracker in nimrod someday :/ i hate the 'pattern command' thing where you have to type in hexcodes as though this was 1960 and you were punching raw binary code on to paper <_<
14:32:12	EXetoC	yeah ok
14:33:18	BitPuffin	hey right now int / int gives back an int right?
14:34:05	EXetoC	Araq: I like the idea of having a library implementation though. I think I said this before, but it will at the very least be a nice way to demonstrate the UDTC feature
14:34:15	EXetoC	with sensible error messages and all
14:34:19	*	vendethiel joined #nimrod
14:34:24	EXetoC	BitPuffin: it works?
14:34:44	BitPuffin	EXetoC: what are you talking about?
14:34:52	EXetoC	BitPuffin: yeah it does, but it gives you a float
14:34:56	EXetoC	again, see system.nim :)
14:35:06	BitPuffin	well yeah I didn't feel like looking it up
14:35:16	BitPuffin	I didn't know that int / int made float
14:35:38	EXetoC	proc `div` *(x, y: int): int {.magic: "DivI", noSideEffect.}
14:35:40	BitPuffin	so we don't have a way to divide and get back int? or maybe that's the div keyword or something like that fuck I don't remember
14:35:57	Demos	does int / int at least do the correct conversion to float?
14:36:09	Araq	Skrylar: what do you mean with "pattern command"?
14:36:15	BitPuffin	Demos: no Araq decided to opt for the incorrect conversion
14:36:21	EXetoC	I got used to it, but then I found it annoying when attempting to write a generic interface
14:36:40	BitPuffin	EXetoC: I was thinking we could have / and /.
14:36:42	Araq	no? 0 / 2 produces 0.5
14:36:53	Araq	how is that the incorrect conversion?
14:37:04	BitPuffin	Araq: I was being sarcastic :)
14:37:08	Araq	I mean 1 / 2 of course
14:37:16	BitPuffin	yeah I was like what
14:37:20	BitPuffin	wow such correct
14:37:35	Demos	hehe well I figured maybe the bitpattern was just interpreted as a float. hey it could happen :D
14:37:36	BitPuffin	0 / 2 = 0.5 - Araq 17 March, 2014
14:37:54	Varriount	Araq: Are you sure that you don't' just want to add some javascript table sorting to the test results page, so that you can see all the failing tests that way?
14:37:58	EXetoC	Demos: brilliant
14:38:09	EXetoC	you know Araq. he ain't stupid :p
14:38:31	Araq	Varriount: yes. I don't want to look at nimbuild's website
14:38:44	Araq	because that is yet another tab in my browser
14:39:12	Araq	#nimbuild is a tab in my irc client. much better ;-)
14:39:46	BitPuffin	he ain't? :o :O but I wanted to use a stupid person's programming language
14:39:53	BitPuffin	Araq: any recommendations?
14:39:53	Varriount	Araq: Wait, so how exactly do you want these "still failing tests" results displayed?
14:40:25	BitPuffin	Varriount: on silver plate with condoms next to it
14:40:31	Araq	dump it into the gist and then get rid of the gist and make nimbot tell us in #nimbuild
14:41:07	BitPuffin	so in other words, add the data and delete it? much accomplished by the sounds of it
14:41:41	Araq	dom96 is going to hate me for this ...
14:41:53	Araq	but I really dislike the gist generation part
14:42:05	Araq	I click on the link just to see that nothing changed
14:42:31	BitPuffin	why are we generating gists?
14:42:35	Araq	that's bad, nimbot should dump the list of changes into #nimbuild
14:42:37	BitPuffin	make jizz, not gist
14:42:52	dom96	No. NimBot should give a summary.
14:42:55	BitPuffin	no but what are we making gists for
14:43:02	dom96	Not spam the damn channel/
14:43:12	Araq	dom96: it's not spamming at all
14:43:16	BitPuffin	can't it just generate a summary based on a timeline you ask for
14:43:23	Varriount	Or we should just have a proper test results page, with filters, searches, etc.
14:43:29	Araq	it's 0-3 lines of changes, on average
14:43:41	BitPuffin	so if you go to build.nimrod-lang.org/summary/from/0/to/100/ it will list the changes?
14:44:02	BitPuffin	pretty sure git can already give you what you need there
14:44:10	Araq	BitPuffin: join #nimbuild and then you might be qualified to comment
14:48:17	Araq	Varriount: a proper test results page is fine, but not essential
14:56:56	Araq	bbl
15:05:05	akghf	http://www.pvv.ntnu.no/~steinl/vitser/evolution.html
15:05:07	Skrylar	do you ever have one of those moments where something seems really crippled and you're like "well that can't be nearly as good as something else, its obviously just gimpy" and then realize its actually meant to be used with something else that makes it actually considerablly powerful <_<
15:06:25	BitPuffin	gosh darn it the new white chapel song is pretty catchy
15:06:29	BitPuffin	whitechapel*
15:09:47	Demos	Skrylar: you have not used UNIX much then have you...
15:10:03	*	Demos remembers replaceing most of a matlab heavy workflow with gnuplot and awk
15:10:39	Skrylar	i used gnuplot earlier today
15:11:09	Demos	it is pretty awesome, and not actually that much harder to use than excel, esp when you need slightly complex graphs
15:11:37	Skrylar	i still don't know how to do proper I/O in nimrod, so i augmented the hashtable code with debugEcho and a tag name, dumped with > blah.log, used vims v/LOAD>/d to delete all lines that don't have "LOAD>" in them and spent 40 minutes getting gnuplot to obey
15:12:27	Skrylar	60% load before the hash table decides it is full is still very plebian though :(
15:15:03	Demos	60% sounds reasonable for some situations
15:25:41	*	Endy joined #nimrod
15:28:58	*	lanior quit (Ping timeout: 245 seconds)
15:31:21	Skrylar	hrm. apparently i have seemingly unreasonable requirements
15:31:49	Skrylar	i was looking in to varying-length tracks, and there's only one tracker that actually does it
15:40:11	Skrylar	Demos: maybe, though i saw people reporting 80-90% with a 4-key cuckoo
15:40:31	Skrylar	mine might do better when i figure out how to do the arranging portions
15:40:44	Demos	Skrylar: yeah, it is a tradeoff. Some hashtables let you tune the load factor
15:40:59	EXetoC	BitPuffin: PHP?
15:59:49	*	clovis quit (Quit: HydraIRC -> http://www.hydrairc.com <- \o/)
16:04:11	*	Demos quit (Ping timeout: 252 seconds)
16:07:15	*	Ransel joined #nimrod
16:07:36	Ransel	How do I echo a string without a newline at the end?
16:08:54	dom96	stdout.write
16:10:52	*	Demos joined #nimrod
16:16:30	Ransel	Thank you, dom96!
16:17:09	dom96	np
16:20:13	*	akghf quit (Ping timeout: 245 seconds)
16:22:59	*	gfdbgd joined #nimrod
16:29:36	*	easy_muffin quit ()
16:30:51	dom96	hi gfdbgd
16:31:17	gfdbgd	Hi there dom96
16:36:49	*	Varriount quit (Ping timeout: 240 seconds)
16:40:35	*	Varriount joined #nimrod
16:40:48	gfdbgd	How can i tell nimrod to load dlls from a specific path?
16:41:11	BitPuffin	EXetoC: what?
16:41:16	Varriount	Sorry about that. I probably shouldn't have asked microsoft paint to resize a 4000x4000 pixel image by 500%
16:47:07	Demos	gfdbgd: I think nimrod loads dlls according to the normal shared library search path rules on your platform. You can use all the normal linker options after using --dynlibOoverride:libname, you may be able to use rpath even with nimrod's dlopen based loading
16:47:21	Demos	windows tends to be a bit more primitive wrt dll load paths
16:50:11	gfdbgd	Thanks, does nimrod support sdl 1 or 2 ?
16:51:08	Demos	not sure. if you want SDL2 it should not be too hard to wrap
16:51:17	Demos	or maybe it is already wrapped, I dun know
16:51:26	dom96	it is already wrapped
16:55:43	*	Demos quit (Ping timeout: 264 seconds)
16:59:52	*	gfdbgd quit (Quit: Page closed)
17:02:57	*	DAddYE joined #nimrod
17:12:28	Varriount	dom96: Which parts of nimbuild post stuff to github?
17:15:04	dom96	you mean post to gist?
17:15:07	dom96	website.nim
17:39:32	*	BitPuffin quit (Ping timeout: 246 seconds)
17:45:55	*	q66 joined #nimrod
17:54:11	*	q66 quit (Ping timeout: 252 seconds)
17:56:04	*	q66 joined #nimrod
18:11:45	*	jbe joined #nimrod
18:12:14	*	BitPuffin joined #nimrod
18:25:22	*	Matthias247 joined #nimrod
18:36:40	*	brson joined #nimrod
18:36:54	*	brson quit (Client Quit)
18:37:02	*	brson joined #nimrod
18:50:05	*	zielmicha joined #nimrod
18:50:46	fowl	lol http://gizmodo.com/5980842/there-is-blatant-racist-and-sexist-language-in-github-code
18:51:29	fowl	nimrod repository is clear i just checked it
18:52:28	EXetoC	qd
18:52:33	EXetoC	xd
18:54:10	Varriount	dom96: Ping
18:54:26	dom96	yea?
18:55:10	Varriount	dom96: Is there any feasable way for me to check my additions to the nimbuild code?
18:55:16	Varriount	(website.nim)
18:57:09	Araq	hey DAddYE is back :-)
18:58:18	dom96	Varriount: Not really. You would have to setup everything manually.
18:58:27	dom96	Well.
18:58:30	dom96	Just redis actually.
18:58:35	dom96	Since you got builders setup anyway
18:59:18	Araq	Varriount: just PR it and let dom96 do the work. His server, his work. :P
18:59:48	dom96	I thought that the reason Varriount was doing this was because I didn't have time myself to do it?
19:01:13	Varriount	If nothing else, it gives me an opportunity to look at the rest of the code.
19:01:46	dom96	Please don't refactor everything though.
19:03:44	Araq	dom96: yeah
19:03:59	Araq	that's still the reason
19:04:03	*	Mat3 joined #nimrod
19:04:10	Araq	but obviously he can't deploy on his own
19:04:30	Araq	and thanks to static typing he doesn't need to test anything anyway *cough*
19:06:14	fowl	Araq, i wanted to add initialization for seq/string to add/len, is adding if(seqvar == 0) initialize(); to the generated c acceptable
19:07:03	Araq	fowl: in principle yes
19:07:15	Araq	make a PR and I'll review it, obviously
19:14:16	Araq	BitPuffin: join the VNUG
19:15:24	Mat3	Araq: I tested some integer vector routines against SSE equivalents. The SSE code suffers from the inpossibility of direct scalar accesses (need at least an expensive shuffling instruction) whereby for integer registers this is no problem (RAX <-> EAX: lower half). Because of these and despite the fact that most SSE instructions have longer latencies and/or can only be combined with integer instructions (dependent on the architecture) I thin
19:15:24	Mat3	k that approach is advantageous for integer and fixpoint SIMD processing
19:18:48	*	Mat3 thinks Intel's SIMD extension is basical ill designed compared to alternative approaches like Altivec
19:19:26	Araq	Mat3: I think the point is also to free the integer units so they are available for other things
19:19:48	Araq	so using SIMD can still be advantageous
19:20:05	Araq	even if not faster in comparison
19:20:38	Mat3	hmm, good point
19:20:41	*	io2 joined #nimrod
19:21:44	Mat3	hi io2
19:21:59	io2	hi Mat3 :)
19:30:16	Mat3	if I think about it: x86 CPU's are out-of-order and hold some ~128 shadow registers, we would only need only one direct register assignement so why not reuse the same one for all operations?
19:30:40	*	Mat3 is now known as Mat3-testing
19:31:57	*	Mat3-testing is now known as Mat3
19:32:47	Mat3	two, fixed register allocations are sufficient
19:37:54	Mat3	I do not know if this would be problematic for C related code optimization however
19:38:06	Mat3	^optimizations
19:38:27	Araq	what's the problem again? personal dislike of compiler specific intrisitics?
19:39:06	Mat3	hold the library compiler independent
19:40:15	BitPuffin	Araq: woa okay
19:40:35	*	silven quit (Remote host closed the connection)
19:40:56	BitPuffin	Araq: ah, it was a trap
19:41:18	Araq	Mat3: there are only 4 compilers left, 1 (intel) being a niche, it's not hard to have 4 different 'importc' pragmas, but a bit tedious
19:41:37	Araq	BitPuffin: nope, sorry my TS-client keeps crashing
19:42:35	BitPuffin	Araq: how lame :/
19:42:41	BitPuffin	that's why we should use mumble instead :P
19:44:10	Mat3	Araq: That is a x86 centric agumentation. The GCC SIMD library is not usable for other CPU architectures
19:44:29	Mat3	(as I know)
19:45:19	BitPuffin	Araq: well tell me if you get it working
19:47:29	Araq	Mat3: ok but if you're concerned with SIMD on ARM, only 2 compilers are left and the intristics are compatible, I think (gcc and clang)
19:49:21	Mat3	Araq: Supported platforms are: x86 (MMX,SSE,SSE2), Freescale (Altivec) and only *one* ARM architecture (Cortex-A, NEON)
19:49:38	Araq	well your "lets use integer registers instead they are just as fast" argument is x86 centric too
19:49:52	Araq	so you lost me
19:50:21	EXetoC	dom96: have you completed the macro and blogged about it yet?
19:52:24	Mat3	Araq: my point was, that this is a processor independent approach which work without adaption for not so uncommon CPU architectures like SH, MIPS (loongson!) and last but not least all recent ARM architectures above MCU's with somewhat 4 kB ram
19:52:37	renesac	Araq, for ARM there is also the proprietary ARM compiler, that is pretty much in the same position as ICC on x86
19:53:45	Matthias247	SH is near death
19:53:57	Matthias247	since renesas is also using ARM now
19:54:32	flyx	so I just tried to bootstrap nimrod from github master. koch boot fails with „compiler/semtypes.nim(207, 32) Error: invalid indentation“
19:55:11	Araq	flyx: interesting
19:55:14	flyx	I opened the file and looked for the error, but I don't see any error there
19:55:55	Araq	which OS?
19:56:10	flyx	OSX 10.9
19:56:37	Araq	bootstrap with --gc:markAndSweep please
19:57:25	flyx	same error
19:57:43	Araq	well which iteration fails?
19:58:13	flyx	wait, at which step should I add this parameter?
19:58:26	fowl	let intType = getSysType tyInt
19:58:48	fowl	flyx, try replacing that line with let intType = getSysType(tyInt)
19:58:57	Araq	ah fowl thank you
19:59:14	fowl	i wonder why it doesnt error for me though
19:59:17	Araq	the old compiler doesn't support the command syntax here
19:59:18	Mat3	Matthias247: I dont think so (military)
19:59:32	Araq	I wonder why our testers do not complain about this
19:59:46	Matthias247	Mat3: yeah, we also have some flying around in Automotive
19:59:52	flyx	now I get „compiler/semexprs.nim(924, 36) Error: ')' expected“
20:00:12	Matthias247	Mat3: but these are areas were anyway nobody would use Nimrod in the next 10 years
20:00:12	Araq	fowl: you're not up to date, command syntax can now be used in more contexts
20:00:49	fowl	flyx, else: (internalAssert(typ.kind == tyCompositeTypeClass); typ.sons[1])
20:01:33	flyx	ah, I get the pattern
20:01:58	fowl	Araq, usually the compiler doesnt use any new features like this though
20:02:37	Araq	fowl: zahary enjoys the new stuff too much :P
20:02:45	flyx	okay, now it succeeded. however it gives „[Warning] executables are still not equal“
20:03:15	fowl	Araq, that means the csources need to be updated?
20:03:52	Araq	no, he should fix the compiler instead to not use these bleeding edge features
20:04:32	fowl	does this mean i can do `discard f x` now?
20:04:40	Araq	try it
20:04:46	BitPuffin	Araq: do you have the latest version?
20:04:50	Araq	you should be able to do that
20:05:04	fowl	sweet
20:05:43	EXetoC	BitPuffin: I think he's on 0.9.2 still
20:06:02	BitPuffin	EXetoC: was referring to TS but ok
20:06:19	Mat3	Matthias247: Well, if we can buy some cheap ARM boards like Raspberri Pi, have a (even cheaper) PIC32 (MIPS3) board [duinomite], some 5-10 AVR based platforms, the firebee (Freescale) and even 2 Propeller based ones, I can envision some demand for an SH board too
20:07:16	Araq	renesac: ok, but does the syntax of the commercial arm compiler's intrisitics differ?
20:07:36	Araq	btw I'll write that as intr from now on
20:07:38	renesac	I have no idea
20:07:44	Araq	abbrevs ftw
20:07:53	Matthias247	Mat3: AVR and the PI are mostly popular due to wide availability, easy tools, existing code and good tutorials for beginners. All that doesn't apply to SH afaik
20:07:59	renesac	still, SIMD instructions exists for a reason
20:08:13	Matthias247	Mat3: even ARM Cortex-M has a hard stand
20:10:56	*	skyfex joined #nimrod
20:11:56	Mat3	yes that is true, however, my point is Nimrod as system language should be attractive for hackers and the maker society - these are people who will even misuse DSP's as CPU
20:13:22	Araq	Mat3: enough of this. make an emulatedSimd.nim module and it will be added
20:13:42	Mat3	ok
20:17:17	BitPuffin	dom96: where the hell did you poke me even
20:17:53	BitPuffin	dom96: but you are not even in ts
20:17:53	Ransel	Uhm, I get an error when trying to do "src\babel1.exe install". Something with that there's no floppy disk inserted, and that I should insert one in \Device\Harddisk\DR4.
20:20:31	BitPuffin	Araq: get yo ass in the VNUG, dom96 is there too now
20:20:47	EXetoC	Ransel: :E
20:21:17	Araq	hi Ransel welcome
20:21:20	EXetoC	yeah a floppy disk is now required. sorry
20:21:40	EXetoC	joking of course. odd error
20:23:36	Ransel	Yeah. :/
20:24:20	Araq	Ransel: babel1.exe? not babel.exe?
20:25:00	Ransel	Araq: Yeah, it's on the installation instructions for Windows. "This is required because Windows will lock the process which is being run."
20:26:42	BitPuffin	EXetoC: do you have CS:GO?
20:27:38	Araq	Ransel: try to run it in an admin shell
20:28:43	Ransel	Araq: Same error :/
20:31:23	Araq	well I don't think you even need to *install* babel
20:31:41	Araq	try to install some other package with babel and see if that works
20:38:04	Ransel	I still get the same error.
20:38:21	Ransel	Same if I run it without any arguments.
20:40:20	Araq	deactivate your anti virus software and check again
20:40:31	BitPuffin	haha
20:40:33	BitPuffin	such a windows problem
20:40:42	Araq	also check that you're not trying to run a 64bit application on a 32 bit OS
20:42:45	*	Varriount is going to the VNUG
20:42:53	Ransel	I have no AV installed and I am running a 64-bit OS.
20:48:28	*	clovis joined #nimrod
20:49:09	Araq	Ransel: how did you compile babel?
20:50:05	Ransel	Araq: "nimrod c src\babel" as the installation instructions say. https://github.com/nimrod-code/babel#windows
20:51:08	Araq	try nimrod c -d:release src\babel
20:52:40	Varriount	(nimrod c -d:release .\src\babel.nim") && (copy .\src\babel.exe .\)
20:53:37	*	Endy quit (Ping timeout: 240 seconds)
20:54:07	Ransel	Still get the error.
20:54:22	Varriount	Ransel: What exactly is the error?
20:54:34	Ransel	But I forgot the console error (I am extremely sorry), but that is "Auto configuration failed
20:54:34	Ransel	6436:error:02001015:system library:fopen:Is a directory:.\crypto\bio\bss_file.c:
20:54:35	Ransel	126:fopen('g:/phpbuild/apps_install/openssl.cnf','rb')
20:54:35	Ransel	6436:error:2006D002:BIO routines:BIO_new_file:system lib:.\crypto\bio\bss_file.c
20:54:35	Ransel	:131:
20:54:35	Ransel	6436:error:0E078002:configuration file routines:DEF_LOAD:system lib:.\crypto\con
20:54:36	Ransel	f\conf_def.c:199:"
20:55:06	Varriount	Er. Is this a windows machine?
20:55:09	Ransel	Apparently it's looking for a file in g:\
20:55:14	Ransel	Varriount: Yes
20:55:23	fowl	flyx, does it work now?
20:55:28	Trixar_za	Shouldn't it be openssl.conf?
20:55:51	EXetoC	BitPuffin: Like I said, I only use Linux so I can't play it yet
20:55:57	Varriount	Ransel: So this a gcc/c compiler error?
20:56:20	Ransel	I don't know?
21:00:13	Varriount	Ransel: According to stackoverflow, you need to set your OPENSSL_CONF environment variable to a proper config
21:02:05	Varriount	Ransel: try "(nimrod c -d:release -u:ssl -u:openssl .\src\babel.nim")"
21:07:56	Varriount	Ransel: http://stackoverflow.com/questions/4106035/what-does-this-openssl-error-mean
21:08:48	Ransel	It works! :D
21:09:12	Ransel	Thank you very much!
21:11:08	flyx	fowl: yes, after I altered the two lines it compiled
21:11:25	Varriount	Ransel: There is a downside - the parameters I gave disable SSL functionality in the modules used by babel.
21:11:50	Varriount	(-u undefines a symbol)
21:11:51	fowl	flyx, good stuff
21:37:10	Varriount	dom96, Araq: Sent the PR for the 'still failing tests' feature.
21:39:10	Araq	yaaaay
21:39:45	EXetoC	weeee
21:41:19	BitPuffin	EXetoC: ah right
21:47:00	Varriount	Also, I found out my git frontend has a "send pull request" feature.
21:47:18	*	skyfex quit (Quit: Computer has gone to sleep.)
21:47:54	*	skyfex joined #nimrod
21:48:31	EXetoC	BitPuffin: I'll buy it if you enable remote access to some windows machine. the added latency might even things up a little
21:50:11	*	nolan_d joined #nimrod
21:56:14	EXetoC	BitPuffin: dude, you gotta write more code
21:56:48	EXetoC	audio-centric code in particular, so that I'll bother to write that high level interface
21:57:00	EXetoC	and maybe some mongo code as well
22:01:44	dom96	Varriount: Honestly, that will spam the channel.
22:01:52	dom96	You won't be able to tell what's what.
22:02:06	Varriount	dom96: I cooked to Araq's order.
22:02:35	Araq	and I made a statistic about it
22:03:14	dom96	I'll give you a statistic:
22:03:34	dom96	There are currently 96 tests failing.
22:03:51	dom96	You want to send 96 lines of text to #nimbuild?
22:03:53	dom96	For each platform
22:03:57	dom96	for each commit built?
22:04:29	Araq	hmm now thats a good point
22:05:03	Araq	I only considered the annoying "nothing changed" gists
22:05:24	Araq	but indeed these are not empty anymore when we always list what still fails
22:05:35	Araq	Varriount: I'm sorry but dom96 is right
22:07:59	Varriount	Araq: Well, to be fair, I anticipated this change, which is why I moved the "still failing" reports into the gist as well.
22:08:16	Mat3	ciao
22:08:19	*	Mat3 quit (Quit: Verlassend)
22:08:22	Varriount	I only have to delete some lines. 3 lines, to be precise. :D
22:08:25	dom96	The only new data that NimBot should announce is the number of "Now failing" tests.
22:10:11	dom96	Also i'm not sure I like your code formatting change.
22:11:10	Varriount	dom96, Araq: PR updated.
22:11:33	Varriount	Now, I go eat.
22:25:07	*	Ransel quit (Ping timeout: 264 seconds)
22:25:25	*	Demos joined #nimrod
22:28:31	BitPuffin	EXetoC: I want to, blame zahary
22:29:19	BitPuffin	no but
22:29:21	BitPuffin	bugz
22:29:40	BitPuffin	EXetoC: you mean for PA?
22:30:03	BitPuffin	EXetoC: I will be writing my own abstraction on top of PA for this project, since it will possibly be ported to console etc
22:32:05	zahary	ha! why am I for blame?
22:33:36	zahary	BitPuffin: have you tried my weekend fixes already?
22:44:54	BitPuffin	zahary: yes! They dun diddely work
22:44:57	BitPuffin	well maybe some of them worked
22:44:59	BitPuffin	but something doesn't
22:45:44	BitPuffin	matrix.nim(121, 49) Error: type expected
22:45:45	EXetoC	BitPuffin: well if you want to deal with untyped pointers directly
22:46:04	BitPuffin	proc `*`*[T; R, N, C: static[int]](a: TMatrix[T, R, N], b: TMatrix[T, N, C]): TMatrix[T, R, C] {.noSideEffect.} =
22:48:29	zahary	well, post bugs please, don't just complain on IRC
22:48:46	BitPuffin	zahary: just really hard to know what's already known
22:48:48	BitPuffin	but yeah
22:48:53	BitPuffin	okay I'll post it as an issue
22:52:25	Varriount	Hi demos
22:52:29	BitPuffin	https://github.com/Araq/Nimrod/issues/1013
22:52:36	BitPuffin	zahary: are you proud dad? ;-;
22:53:03	Demos	hi Varriount
22:58:49	*	clovis quit (Quit: clovis)
22:59:38	BitPuffin	no but zahary it's thanks to the work you do on generic that I can almost code ;D
23:01:59	zahary	is this the last issue so far? if you comment (or don't use) that proc are there other problems?
23:02:01	BitPuffin	EXetoC: I guess technically I can work on audio stuff while I wait for the last bugs
23:02:07	BitPuffin	good question sir
23:02:35	BitPuffin	man I wish we had block comments
23:03:31	Araq	discard """
23:03:33	EXetoC	'when False'? 'discard """ """'?
23:03:37	Araq	verbose block comment"""
23:04:29	BitPuffin	odd
23:04:38	BitPuffin	not sure which line causes this tbh
23:04:40	EXetoC	visual editing in vim?
23:04:40	BitPuffin	ector.nim(7, 42) Error: ordinal type expected
23:04:54	BitPuffin	zahary: that's when running matrix.nim, feels like I've seen it before
23:05:51	zahary	I'm currently implementing a swizzle example and I think I stumbled on this one
23:05:51	BitPuffin	my hunch is that It's because like a generic variable is passed to a TVector type and it goes like hey, yo, you didn't say that it's a static int, so I'm gonna say it wasn't an ordinal type for shitzels
23:08:02	EXetoC	BitPuffin: program-specific audio code?
23:08:43	EXetoC	or the lib
23:09:25	BitPuffin	EXetoC: for the engine
23:10:00	BitPuffin	zahary: very annoying that it doesn't show the line it happens on :/
23:10:57	zahary	this is an internal error
23:11:28	zahary	gotta go, see you tomorrow
23:11:42	Demos	zahary, I got an internal error when I tried to use `.` overloading to write a swizzle implementation
23:11:46	Demos	I gotta go as well though
23:14:09	*	darkf joined #nimrod
23:14:22	BitPuffin	zahary: Yeah I know but you still wanna know which line causes the internal error lol
23:14:29	BitPuffin	see you :) have a good one
23:28:11	*	xenagi joined #nimrod
23:30:57	*	jbe quit (Quit: Leaving)
23:33:23	*	zielmicha quit (Quit: Connection closed for inactivity)
23:35:58	*	Demos_ joined #nimrod
23:38:17	Demos_	hmm I am having issues with type Vector[N: static[int]; T; O: Options] = TMatrix[N, 1, T, O], it says "can not instanciate TMatrix"
23:38:26	Demos_	Similar issue BitPuffin?
23:44:26	*	io2 quit ()
23:44:27	BitPuffin	Demos_: I am not allowed to disclose open source secrets to competitors sorry
23:44:37	Demos_	hehe
23:44:43	BitPuffin	no but I don't think so :o
23:46:06	*	skyfex quit (Quit: Computer has gone to sleep.)
23:46:42	*	skyfex joined #nimrod
23:46:59	Demos_	well I will use your library if it meets my needs (Column major data storage, no extra data in fixed size matrices/vectors)
23:48:26	Demos_	hm it may be, my test case error'd with the same message as #1013
23:49:39	Demos_	also who was it who was really into note-takeing apps?
23:52:19	*	seubert quit (Ping timeout: 264 seconds)
23:53:10	Demos_	right now I have a pretty sketch set of linear algebra functions that just take a specific sized vector, and I use the good ol copy-n-paste generics. GOOD ENOUGH FOR C GOOD ENOUGH FOR ME :D