00:02:10 -!- carado has quit (Ping timeout: 246 seconds). 00:05:22 ^show list 00:05:23 (Taneb atriq Ngevd Fiora nortti Sgeo alot)S 00:05:38 ^def list ul (Taneb atriq Ngevd Fiora nortti Sgeo ThatOtherPerson alot)S 00:05:39 Defined. 00:05:59 Who is ThatOtherPerson? 00:06:05 fizzie: needs saving hth 00:06:15 ^save 00:06:27 no? 00:06:31 ok 00:06:35 i do not think that command responds to anyone but fizzie 00:09:58 hi fungot 00:09:59 kmc: he is so silly! 00:10:14 fungot: me or you? 00:10:15 kmc: for a while it was frustrating. most of john's dad, who broke a huge wizard is. you don't like is my mother's obsession with clocks. the sooner all these idiots stop being alive the better of you. 00:10:26 ^style 00:10:26 Available: agora alice c64 ct darwin discworld europarl ff7 fisher fungot homestuck* ic irc iwcs jargon lovecraft nethack pa qwantz sms speeches ss wp youtube 00:10:40 ^style fungot 00:10:40 Selected style: fungot (What I've said myself) 00:10:48 fungot: hi 00:10:48 kmc: is that something you know and and the cases that required to actually mutate the original ( sorted, perhaps, it may be said that particularly here, parliament will give a single instance, 00:15:32 Mr. President, I think fungot is mixing metaphors again 00:15:33 oerjan: ( that is, levinson went to kish, disappeared, had on your behalf of the commission is not allowed to use the crane, enter any two of these letters, a b y. every night, brings a new world of good! i was your problem? maybe you just can't handle is designed" if continuations are not unmodular in the same 00:26:58 -!- Taneb has quit (Quit: Leaving). 00:42:35 Oh hey someone wrong my mamb for Scala years ago 00:42:42 https://github.com/urso/embeddedmonads 00:43:05 Was that supposed to be "wrote"? 00:43:25 yes, yes it was 00:45:15 Have you considered writing a library called Mambo? 01:18:53 -!- copumpkin has quit (Ping timeout: 245 seconds). 01:19:24 -!- copumpkin has joined. 01:22:19 ^save 01:22:19 OK. 01:28:14 yay! 01:28:36 * oerjan didn't really think fizzie was around 01:28:58 I have been playing a game, is why. 01:29:17 you'd also been idle for over a day 01:30:15 Oh, but I've been checking the channel every few hours, all sneaky-stealth-like. 01:30:37 (IRC should have some kind of attention notification protocol, perhaps?) 01:31:21 IM things have that "user is typing" indicator, but even they don't generally (AFAIK) have a "user is looking" indicator. 01:32:37 we've been thinking about that a lot for the business communication tool we're building 01:35:40 it works as realtime chat, but (unlike say IRC) it strongly encourages you to read every message 01:35:56 Incorporating eye-tracking technology sound like the obvious way to go about it, of course. 01:35:58 so that it can also take the place of asynchronous things like email 01:36:35 so we kind of want to deliberately avoid a 'user is around right now' notification, because it would discourage people from using it in that second way 01:36:55 on the other hand it's something people really really want 01:37:22 xmpp's combo of away statuses and typing notifications seems fine to me 01:38:01 Beating the users until they stop wanting things you don't want them to want seems fine to me. 01:40:13 but in the limit you end up with something nobody wants 01:40:28 we've been thinking about that a lot for the business communication tool we're building <-- as long as it is better than Samtime in the god awful Lotus Notes... 01:40:40 in the limit you work somewhere else, i thought 01:41:26 what's Samtime like 01:41:30 Bike: well ok 01:41:41 kmc, well I guess it actually isn't the worst part of notes 01:42:11 kmc, of course it suffers the general problems with confusing UI of notes that means you have no idea where to find the setting you want. 01:42:49 kmc, oh and I have no idea how to change spell check language in it. It seems stuck to English, while the rest of Notes is stuck to Swedish 01:43:08 kmc, anyway it is an IM system integrated into Notes 01:43:12 use it at work 01:43:23 does show the away status 01:44:57 -!- Nisstyre-laptop has quit (Quit: Leaving). 01:46:37 -!- Nisstyre-laptop has joined. 01:49:15 -!- DHeadshot has quit (Read error: Connection reset by peer). 01:49:21 -!- DH____ has joined. 02:01:13 http://www.reddit.com/r/Python/comments/17rfh7/warning_dont_use_pip_in_an_untrusted_network_a/ 02:01:22 yay people are finally talking about these problems 02:02:12 "It's a bit worse than that." security sure is a fun field 02:04:43 yeah, the default python HTTP libraries make it basically impossible to check SSL certs 02:04:47 who would ever want to do that right 02:05:02 Python: explosive, acid-leaking batteries included 02:05:11 -!- Nisstyre-laptop has quit (Quit: Leaving). 02:17:38 Would x + y = y + x be a valid last pattern match, so that all cases where x is specific also apply to where y is specific instead? 02:17:50 what 02:17:59 oh i see what you mean 02:19:11 yes, you can do things that way, but I remember concluding that it's not worthwhile 02:19:19 Say I have (x,True) + (y,False) = something and then don't want to write the case for (x,False) + (y,True) = something 02:19:39 one problem is that if you forget a case, there's no compiler warning and you get an infinite loop at runtime 02:20:42 Yeah, that's what I was thinking 02:21:25 I'm making silly Num instances =P 02:21:39 instance Num ([()],Boolean) 02:21:48 that is silly 02:22:03 Boolean is the sign, the list is the magnitude 02:22:04 you should introduce a new constructor rather than using (,) 02:22:09 As in the length 02:22:13 otherwise it will overlap with other (,) instances 02:22:35 I was thinking of making a new constructor, but was lazy =P 02:22:42 Sure, I will 02:23:39 also if you want to be sillier, use Fix Maybe 02:23:47 data SillyNum = Positive [()] | Negative [()] 02:23:49 instead of [()] 02:24:01 :t Fix Maybe 02:24:04 Not in scope: data constructor `Fix' 02:24:04 Not in scope: data constructor `Maybe' 02:24:12 :k Fix 02:24:13 Not in scope: type constructor or class `Fix' 02:24:15 :k Mu 02:24:16 (* -> *) -> * 02:24:18 ok lambdabot calls it Mu 02:24:23 why 02:24:25 @src Mu 02:24:25 newtype Mu f = In { out :: f (Mu f) } 02:24:36 because μ is used for least fixed points in maths sometimes 02:24:46 > In (Just (In Nothing)) 02:24:48 In (Just (In Nothing)) 02:24:59 Is Fix Maybe a chain of justs that possibly ends in Nothing 02:25:08 yes 02:25:25 Mu Maybe ≈ Maybe (Mu Maybe) 02:25:36 where by ≈ i mean "isomorphic to" 02:25:42 data Fix a = a (Fix a) ??? 02:25:55 newtype Mu f = In { out :: f (Mu f) } 02:26:05 Ah 02:26:09 you need a value constructor 02:26:20 Yeah, I see 02:26:27 also you need to use 'newtype' if you want it to be an actual isomorphism 02:26:34 otherwise ⊥ ≠ In ⊥ 02:26:44 so each level of recursion has one more "value" 02:27:03 i wrote some stuff about this stuff here http://mainisusuallyafunction.blogspot.com/2010/12/type-level-fix-and-generic-folds.html 02:28:15 oh, so it's the same recursive-functions-are-actually-fixedpoints thing as with, uh, functions. 02:28:22 yeah 02:28:29 I think I'll use [()] for now 02:29:06 [()] means you can use list functions for arithmetic... 02:29:16 I like how the definition of fix there relies on recursion 02:29:49 well in a statically typed language, you can't write fix unless you are already given some kind of recursion construct 02:29:58 > const id <$> [(),(),()] <*> [(),()] 02:30:00 [(),(),(),(),(),()] 02:30:29 oh wait 02:30:36 > [(),(),()] *> [(),()] 02:30:38 [(),(),(),(),(),()] 02:30:50 Need to consider negative values too =P 02:31:09 > replicate 3 [()] >> replicate 5 [()] 02:31:11 [[()],[()],[()],[()],[()],[()],[()],[()],[()],[()],[()],[()],[()],[()],[()]] 02:31:15 though in Haskell there are various accidental "recursion constructs" you can use 02:31:18 http://okmij.org/ftp/Haskell/types.html#fix 02:31:23 > replicate 3 [()] >> replicate 5 [()] >>= id 02:31:25 [(),(),(),(),(),(),(),(),(),(),(),(),(),(),()] 02:31:42 Oh, duh 02:31:46 > replicate 3 () >> replicate 5 () 02:31:47 [(),(),(),(),(),(),(),(),(),(),(),(),(),(),()] 02:32:11 anyway if you have a simple typed lambda calculus, you can't write fix 02:32:29 note that its type, (a -> a) -> a, is unsound as a logical proposition 02:32:59 > sequence_ $ [(),(),()] <$ [(),()] 02:33:01 [(),(),(),(),(),(),(),(),()] 02:33:10 once you allow arbitrary recursion, your programs are no longer really proofs, because they can loop forever instead of yielding a value of their declared type 02:33:42 Somehow I can't see {-# RULES forall m n. length (replicate m () *> replicate n ()) = m * n #-} becoming a thing 02:33:59 If you don't allow arbitrary recursion, your programs aren't turing-complete 02:34:01 > (mapM_ . const) [(),(),()] [(),()] 02:34:04 [(),(),(),(),(),(),(),(),()] 02:34:44 @let voided n = replicate n () 02:34:46 Defined. 02:34:50 > void [1..] > voided 5 02:34:53 True 02:34:56 Jafet: Wouldn't that give a different result for negative values of m and n 02:35:47 > (compare `on` void) (1:2:3:4:5:undefined) [10,20,30] 02:35:49 GT 02:36:05 :t on 02:36:07 (b -> b -> c) -> (a -> b) -> a -> a -> c 02:36:33 > null . 02:36:35 :1:7: parse error (possibly incorrect indentation) 02:36:41 :t null .: void 02:36:43 Couldn't match expected type `[a0]' with actual type `()' 02:36:43 Expected type: g0 a1 -> g0 [a0] 02:36:43 Actual type: g0 a1 -> g0 () 02:37:09 I love expressive type systems 02:37:26 :t null 02:37:28 [a] -> Bool 02:37:42 > (null .) 02:37:43 No instance for (GHC.Show.Show (f0 [a0] -> f0 GHC.Types.Bool)) 02:37:44 arising f... 02:38:03 :t (null .) 02:38:04 Functor f => f [a] -> f Bool 02:38:19 kmc: i know all that much at least, you don't need to tell me again :P 02:39:11 but do you know the type newtype T = T (T -> T)!! 02:39:23 Ah, lambdabot does the (.) = fmap thing 02:40:03 shachaf: T id 02:40:04 FreeFull: http://en.wikipedia.org/wiki/%CE%9C-recursive_function you only need search!!! 02:41:44 Bike: ok 02:42:00 it's hard to remember what people know and don't know 02:42:14 i suppose 02:42:18 nothing wrong with talking it through again 02:42:47 Back to my num instance 02:45:46 hm, i had this idea that cata(morphism) was actually reasonably widespread in haskell world 02:46:40 more of an agda thing maybe? 02:46:56 -!- Phantom_Hoover has joined. 02:48:13 Maybe I shouldn't have used separate constructors for positive and negative 02:48:20 It's bloating my num instnace 02:50:39 data Sign = Negative | Zero | Positive deriving Show data SillyNum = SillyNum Sign [()] deriving Show this might be better, probably should derive more instances but whatever 02:50:54 so what's this all for 02:51:11 Sillyness 02:51:48 Actually, screw Sign, I'll just use Integer or something 02:52:48 i'll make my own Num instance, with blackjack and hookers 02:55:33 -!- DH____ has quit (Read error: Connection reset by peer). 03:04:00 data SillyNum = SillyNum Integer [()] deriving Show 03:04:11 instance Num SillyNum where { fromInteger x = SillyNum (signum x) (genericReplicate (abs x) ()); negate (SillyNum x y) = SillyNum (-x) y; abs (SillyNum x y) = SillyNum (abs x) y; signum (SillyNum x _) = SillyNum x [()]; (SillyNum x a) + (SillyNum y b) = fromInteger (x * genericLength a + y * genericLength b); (SillyNum x a) * (SillyNum y b) = SillyNum (x*y) (a >> b) 03:04:46 } 03:05:33 :t (>>) 03:05:35 Monad m => m a -> m b -> m b 03:05:36 Appears to work 03:06:21 Bike: Same as *> but for monads 03:11:45 What happens when you call length on a list larger than maxBound :: Int 03:12:15 Assuming you wait for it to actually finish 03:12:35 > length ([1.. (fromIntegral (maxBound :: Int) + 1)]) 03:12:39 You realize that you need to buy a 64-bit CPU 03:12:39 mueval-core: Time limit exceeded 03:12:58 FreeFull: I think the semantics are not defined by the standard 03:13:03 > length ([1.. (fromIntegral (maxBound :: Int))]) 03:13:07 mueval-core: Time limit exceeded 03:13:12 Let me look up the standard implementation of length 03:13:15 in practice it will depend on how length is implemented and how Int overflow works on your Haskell implementation 03:13:21 length [] = 0 03:13:21 length (_:l) = 1 + length l 03:13:24 from the report 03:13:37 so that will pretty directly depend on what (+) :: Int -> Int -> Int does 03:13:46 > 1 + (maxBound::Int) 03:13:47 Are there laws for (+) 03:13:48 -9223372036854775808 03:13:53 nice! 03:14:07 Jafet: if there are any laws for numbers, they are almost certainly violated by Double ;P 03:14:12 i mean equality is not even reflexive 03:14:39 Since there are no laws, length can return anything on a non-null list 03:14:43 hm, i do wonder about that sometimes. like should you have * work on both integers and matrices even though they have different properties. how do you fit this into a type system that makes any sense at all. 03:14:54 CHECKMATE 03:15:00 Jafet: i think the semantics of Int specifically are specified within the range [-2^29..2^29-1] 03:15:01 So you'll eventually end up with a negative value for length :D 03:15:18 And it's possible to get a list large enough that length will say it's length is 0 03:15:19 > let n :: Double; n = read "NaN" in n == n 03:15:21 False 03:15:29 good representation 03:15:33 int overflows are a grave security concern in C 03:15:49 lotta code will do like buf = malloc(n * sizeof(foo)) 03:16:02 haha oops 03:16:04 if that multiplication overflows, you allocate less memory than you expect 03:16:27 kmc: Should I be surprised Haskell has adopted the NaN /= NaN thing? 03:16:28 sscanf(buf, "%d", &n) 03:16:50 I shouldn't 03:16:52 isn't it just using ieee like everyone else in the universe? 03:16:56 Makes sense for ieee 03:17:02 except ps2 gpus maybe 03:17:18 In other words calloc is more secure than malloc??? 03:17:34 But then 03:17:35 Jafet: how's that 03:17:38 > Nothing == Nothing 03:17:40 True 03:17:52 calloc(n, sz) 03:18:01 ah yes 03:18:11 does the haskell report specify that floating point is IEEE754? 03:18:14 FreeFull: what's the relevance 03:18:22 > isIEEE (undefined :: Float) 03:18:23 True 03:18:30 okay so it's not spec'd but you get a runtime check 03:18:56 Double should be an algebraic type 03:19:11 > isIEEE (3 :: Real) 03:19:11 and NaN /= NaN does make sense, it's just unfortunate 03:19:13 Expecting one more argument to `GHC.Real.Real' 03:20:01 kmc: unfortunate? 03:20:14 :k Real 03:20:16 NaN is almost like ⊥ -- it means you have no information about the "actual" result 03:20:16 * -> Constraint 03:20:22 :k Constraint 03:20:23 BOX 03:20:55 ow my brain 03:20:55 > (Nothing == Nothing) :: Maybe Float 03:20:57 Couldn't match expected type `Data.Maybe.Maybe GHC.Types.Float' 03:20:58 ... 03:21:12 > (Nothing == (Nothing :: Maybe Float)) 03:21:12 oh right, nan means "not representable" in addition to "doesn't make sense". gosh i'm bad with floats 03:21:14 True 03:22:08 however Infinity == Infinity which seems a little bit wrong 03:22:21 because two numbers which are too big to represent might still be unequal 03:22:33 > (1/0) == (2/0) 03:22:33 Isn't no information about the actual result undefined 03:22:35 True 03:22:42 positive infinity, infinity on the pi/4 vector, what's the diff 03:22:53 Multiply both sides by 0, 1 == 2 03:23:07 Jafet: yeah, most systems allow configuring the floating point unit to treat NaN as an exception 03:23:14 -!- constant has changed nick to function. 03:23:16 though iirc it's pretty nasty with GHC haskell 03:23:27 because this FPU state is not saved by the green-thread scheduler 03:23:33 > 2.0^10000 == 2.0^20000 03:23:35 True 03:24:14 Why do most programming languages have mediocre type systems ): 03:24:23 any attempt to represent real numbers in a computer is doomed to incoherence 03:24:38 FreeFull: most things are bad 03:24:38 Also it's weird to blame type systems for reals being weird. 03:24:40 There are always the unreal computers 03:24:59 many explanations are offered but I prefer to think that success is simply random, and most things people try are bad 03:25:03 so most things that succeed are also bad 03:25:31 Bike: Nah, I'm not 03:25:45 There are only so many reals you can represent in a limited space 03:26:09 there are only so many reals you can represent in countably infinite space 03:26:14 which is to say almost none of them 03:26:25 lebesgue measure is so passé 03:26:30 Well, any real is countably infinite 03:26:32 almost every real number contains an infinite amount of information 03:26:50 yeah it's true Jafet 03:26:55 i was speaking imprecisely 03:27:01 "Damn you cantor" 03:27:08 however the problem of representing reals in a computer is more fundamental than the problem of representing integers 03:27:24 the latter is just about space whereas the former is about the fact that reals are crazy and fucked up 03:27:25 ya think? representing integers is a pretty neat problem too. 03:27:58 of course you don't really have to think about it unless you're dealing with trillion-digit numbers, i guess. how boring 03:28:33 (and not that reals aren't crazy and fucked up) 03:28:47 You're lucky if your real is just transcendental, you can still approximate that with a rational 03:29:26 Most reals can't be and all you can do is assign some symbol to it 03:29:59 “eog or the Eye of GNOME is a simple graphics viewer for the GNOME desktop which uses the gdk-pixbuf library. It can deal with large images, and zoom and scroll with constant memory usage.” (eog:21228): GLib-ERROR **: /build/buildd/glib2.0-2.34.1/./glib/gmem.c:165: failed to allocate 18446744071773880320 bytes 03:30:17 uh aren't most reals transcendental 03:30:37 eog failed to open my 32768x32768 image 03:30:41 gnome sucks 03:30:51 FreeFull, except for the ones that no string of symbols can be assigned to 03:30:57 Which is most of them 03:31:10 Well, hmm 03:31:16 Not a good way to phrase that 03:31:27 You can assign them. Omega = chaitin's constant on 2,3 machines. 03:31:36 but that's not what you mean. 03:31:42 Different meaning of symbol 03:31:44 No finite amount of symbols and description suffices to describe. 03:31:50 There you go. 03:32:09 Also isn't the golden ratio algebraic or am I being dumb 03:32:30 Chaitin's constant is described as the real number containing the probability of halting for 2,3 machines 03:32:30 Sgeo: to describe as distinctive from other reals you mean 03:32:43 FreeFull, good point 03:32:47 Because I could just say x is a real 03:32:51 Jafet: most numbers are undefinable too, though. 03:32:54 -!- Phantom_Hoover has quit (Remote host closed the connection). 03:33:01 Those don't exist 03:33:11 Monster. 03:33:21 Everything exists. 03:33:34 If all numbers are definable, how many numbers are between 0 and 1? 03:33:50 Existence is easy 03:33:54 Actually, nevermind 03:33:56 (Therefore: everything is easy) 03:33:57 FreeFull: cardinality of the naturals duh 03:34:48 FreeFull: also, the golden ratio is algebraic but all rational approximations of it suck (for reasonable definitions of suckage) 03:35:17 How would you make a function that takes a floating value of Inf and makes it 1 using just arithmetic operations? 03:35:39 * function returns 1 to FreeFull 03:35:42 Bike: Good enough 03:35:43 f _ = 1 03:36:00 Bike: I didn't specify enough 03:36:09 Heh, heh. 03:36:17 It has to return 1 only for Inf and 1, for everything else it's id 03:36:55 No ifs? 03:36:57 Hm 03:37:04 -!- Nisstyre-laptop has joined. 03:37:09 f n | n /= n = 1 | 1>0 = n 03:37:12 > let f n | isInfinite n = 1; f n = n in (f 3, f (1/0)) 03:37:13 (3.0,1.0) 03:37:15 Not necessarily id 03:37:36 -!- WeThePeople has joined. 03:37:39 Surely guards count as ifs 03:37:43 Not necessarily 1 for 1, but has to return different values for different inputs 03:37:44 > let f n | (isInfinite n && (n > 0)) = 1; f n = n in (f 3, f (1/0), f (-1/0)) 03:37:45 (3.0,1.0,-Infinity) 03:37:46 You could probably say it has to be a composition of possibly partially applied arithmetic functions. 03:38:04 Guards count as non-arithmetic operations here, and isInfinite does too 03:38:24 NaN is non-arithmetic 03:38:30 Or you could just get to your point instead of adding more and more constraints as kmc works around yours. 03:38:38 Non-arithmetic number 03:38:43 i'm not really paying attention 03:38:46 Ignore NaN 03:38:52 just fuckin' around 03:38:55 on the internet 03:38:57 same thing 03:38:57 But then what's Inf-Inf? 03:39:04 > Inf - Inf 03:39:06 That's an arithmetic operation 03:39:06 Not in scope: data constructor `Inf' 03:39:06 Perhaps you meant one of these: `In' ... 03:39:09 hth 03:39:30 > let Inf = 1/0 in Inf - Inf 03:39:32 Not in scope: data constructor `Inf' 03:39:32 Perhaps you meant one of these: `In' ... 03:39:42 > let inf = 1/0 in inf - inf 03:39:44 NaN 03:39:57 There, it's bread 03:40:01 Unless you want to take the utter crazy view that Inf-Inf = 0, in which case, (1+) . (`subtract` inf) 03:40:23 well, let's see, on the riemann sphere such a function wouldn't be continuous so fuck it 03:40:25 I think J does that or something 03:41:03 smokin' a J 03:41:16 Oh it doesn't 03:41:18 ) _-_ 03:41:19 Sgeo: |NaN error 03:41:19 Sgeo: | _ -_ 03:41:32 j's syntax is a thing of beauty 03:41:56 I was thinking there are functions that squish -Inf..Inf to -1..1 or some other range 03:42:20 But wouldn't IEEE floats foil that without explicitly considering infs 03:42:55 Wouldn't any non-infinite number get squished to, say, 0? 03:42:57 well it wouldn't be injective 03:43:42 Sgeo: Not necessarily 03:43:56 Sgeo: The closer the number is to 0, the less it'd get squished 03:44:11 hrm 03:44:29 Shouldn't say "arithmetic" operations. Makes me want to exclude logarithmic stuff 03:44:57 Ok, include logarithmic and trigonometric stuff too then 03:45:05 Say, atan :) 03:45:05 Since you're dealing with reals you could just say analytic functions. 03:45:09 ) $%97^3$@>.$(Po.uo.godeto.go4i$*%#@( 03:45:09 kmc: |spelling error 03:45:09 kmc: | $%97^3$@>.$(Po.uo.godeto.go4i$*%#@( 03:45:09 kmc: | ^ 03:45:19 god kmc, check your spelling 03:45:23 ) $%97^3$@>.$( 03:45:23 kmc: |syntax error 03:45:23 kmc: | $%97^3$@>.$( 03:45:43 But I don't know if there's a reasonable extension of the complex plane with two points at infinity. 03:46:22 > atan (1/0) 03:46:23 1.5707963267948966 03:46:33 Close enough 03:46:45 ) _3 o. 1%0 03:46:45 Sgeo: 1.5708 03:46:50 > tan (atan (1/0)) 03:46:51 can't find file: L.hs 03:46:59 > tan (atan (1/0)) 03:47:00 1.633123935319537e16 03:47:09 Not close enough 03:47:17 ) 3o._3o._ 03:47:18 Sgeo: |ill-formed number 03:47:25 ) 3o._ 3o._ 03:47:25