Available formats: content-negotiated html turtle (see SIOC for the vocabulary)
Back to channel and daily index: content-negotiated html turtle
These logs are provided as an experiment in indexing discussions using IRCHub.py, Irc2RDF.hs, and SIOC.
| 00:00:23 | <ray> | besides function application, which goes up to 11 |
| 00:00:33 | <Baughn> | 10 |
| 00:00:35 | <mauke> | I thought application was 10 |
| 00:00:39 | <alexsuraci> | I'm curious how haskell does the infix parsing, it seems to me that you'd want to first scan for infixr/infixl/infix to determine how to parse the operators |
| 00:00:47 | <ray> | i skipped 10 so i could say "goes to 11" |
| 00:00:59 | <Baughn> | ray: No, that's token parsing |
| 00:01:12 | <Baughn> | It's why mapconcat isn't map concat |
| 00:02:23 | <mauke> | ray: Haskell does go to 11 but only if you leave function application at 10 |
| 00:25:34 | <A1kmm> | Hmm, I have a program which segfaults early on when run with ghc head on threaded mode... any tips on how to go about debugging this? |
| 00:26:15 | <A1kmm> | (the problems only happen if I use +RTS -N2 or higher to enable parallel Haskell...) |
| 00:29:20 | <Tobsan> | what is the main use of monoids in Haskell? |
| 00:30:54 | <Apocalisp> | Tobsan: Summing over functors, deriving other monoids. They're rather useful, really. |
| 00:31:33 | <Tobsan> | Alright. I was writing a tree structure just now and i happened to implement Monoid |
| 00:31:41 | <Tobsan> | but then i realized i didnt know why |
| 00:31:52 | <Apocalisp> | @type Data.Foldable.foldMap |
| 00:31:53 | <lambdabot> | forall a m (t :: * -> *). (Monoid m, Data.Foldable.Foldable t) => (a -> m) -> t a -> m |
| 00:31:56 | <Apocalisp> | that's why! |
| 00:32:00 | <Tobsan> | oh snap! |
| 00:33:22 | <Apocalisp> | You'll want to instance Foldable for your tree structure, I think. |
| 00:34:46 | <Baughn> | > foldMap show [1..4] |
| 00:34:47 | <lambdabot> | Not in scope: `foldMap' |
| 00:34:53 | <Baughn> | > Data.Foldable.foldMap show [1..4] |
| 00:34:55 | <lambdabot> | "1234" |
| 00:41:20 | <Peaker> | does the lack of axiom of choice in intuition(SP?) logic (which total value types are isomorphic to) not a problem, in dependent types language, that use this isomorphism? |
| 00:42:42 | <opqdonut> | well AC is non-constructive |
| 00:43:03 | <opqdonut> | constructiveness is generally a nice property in logics relating to programming languages |
| 00:43:17 | <Peaker> | opqdonut: I guess I have a lot more reading to do? |
| 00:43:23 | <Peaker> | opqdonut: or is that easy to explain? |
| 00:43:34 | <opqdonut> | probably yes (more reading) |
| 00:43:49 | <opqdonut> | it might be easy to explain, however i am in no condition to do so |
| 00:44:09 | <Peaker> | ok, thanks |
| 00:44:53 | <opqdonut> | AC is only relevant when the logic we are studying has a rich enough model under it (sets, for example) |
| 00:45:39 | <opqdonut> | i |
| 00:45:53 | <opqdonut> | 'm not really sure what the c-h correspondence looks like wrt. dependent types |
| 00:47:42 | <conal> | Baughn, Peaker: i just released new versions of checkers, reactive, and reactive-glut. if you find version problems etc, please let me know. |
| 00:48:17 | <abbe> | hi everyone |
| 00:48:35 | <abbe> | is there any way to perform mapM_ like operation on Data.Map ? |
| 00:48:54 | <opqdonut> | yes |
| 00:48:59 | <opqdonut> | ?src Traversable |
| 00:49:00 | <lambdabot> | class (Functor t, Foldable t) => Traversable t where |
| 00:49:00 | <lambdabot> | traverse :: Applicative f => (a -> f b) -> t a -> f (t b) |
| 00:49:00 | <lambdabot> | sequenceA :: Applicative f => t (f a) -> f (t a) |
| 00:49:00 | <lambdabot> | mapM :: Monad m => (a -> m b) -> t a -> m (t b) |
| 00:49:00 | <lambdabot> | sequence :: Monad m => t (m a) -> m (t a) |
| 00:49:06 | <abbe> | opqdonut: how ? |
| 00:49:22 | <opqdonut> | there's a Traversable instance for Data.Map |
| 00:50:00 | <opqdonut> | http://hackage.haskell.org/packages/archive/base/4.1.0.0/doc/html/Data-Traversable.html#t%3ATraversable <- the relevant documentation |
| 00:50:48 | <opqdonut> | however, mapM_ f $ elems m might also be sufficient |
| 00:51:17 | <abbe> | opqdonut: oops, didn't thought of that. thanks :) |
| 00:51:38 | <opqdonut> | np |
| 00:51:40 | <Baughn> | conal: That fixes my most serious problem (event->behavior->snapshot working very badly) nicely, but the monad instance for Event now locks up at 100% cpu use |
| 00:52:03 | <Baughn> | conal: (Earlier it deadlocked at no use, iirc. Or maybe I just didn't notice.) |
| 00:52:17 | <Baughn> | Oh.. yeah, it deadlocked and GHC killed my program. Worked fine without -threaded. |
| 00:52:21 | <Baughn> | Now it doesn't work, period. |
| 00:52:48 | <Baughn> | I'll look closely at your patches, I guess. :) |
| 00:53:29 | <conal> | Baughn: sigh :( |
| 00:53:40 | <conal> | Baughn: thx. |
| 00:54:06 | <Baughn> | conal: Oh, I almost forgot. Inside Timer.hs, specifically in the adaptE implementation (where it sleeps), there's a > comparison that I'm pretty sure should be >= |
| 00:54:15 | <Baughn> | conal: Doesn't cause anything worse than a spurious sleep, though |
| 00:54:23 | <Baughn> | (A spurious 0-length sleep) |
| 00:55:34 | <abbe> | > mapM (putStrLn . show) (Data.Map.insert 5 10 (Data.Map.empty)) |
| 00:55:35 | <lambdabot> | Couldn't match expected type `[a]' |
| 00:55:46 | <Baughn> | conal: Make that Timing.hs and sleepPast |
| 00:56:04 | <abbe> | any ideas what is error in this ? |
| 00:56:20 | <conal> | Baughn: i see it now. hm. the comment says "-- Snooze until strictly after the target." |
| 00:56:21 | <mauke> | yeah, this mapM takes a list |
| 00:57:10 | <abbe> | mauke: oh, so i've to import from Traversable. and explicitly qualify it, right ? |
| 00:57:22 | <Baughn> | conal: Which is a questionable idea. How do you even know you're "strictly after" after calling sleep with *0*? |
| 00:57:33 | <mauke> | yeah, or hide the prelude version |
| 00:57:41 | <mauke> | also, putStrLn . show = print |
| 00:57:44 | <Baughn> | conal: But if that's important, better replace the loop |
| 00:58:22 | <conal> | Baughn: i guess i don't. that whole input timing thing is still shaky. not just in the details. |
| 00:58:23 | <abbe> | mauke: any ideas how to assign import a module in GHCi under a separate name. equivalent of import Data.Map as M ? |
| 00:58:31 | <abbe> | s/assign i/i/ |
| 00:58:41 | <Baughn> | conal: That said, for what adaptE's using it for, I don't think I'd want "strictly past" |
| 00:58:43 | <MonadBaastad> | BTW, isn't Data.Map an instance of Monad? :) |
| 00:58:52 | <mauke> | abbe: I don't know a way to do that |
| 00:59:01 | <mauke> | MonadBaastad: no |
| 00:59:15 | <abbe> | okay, thanks mauke :) |
| 00:59:18 | <Baughn> | conal: If I ask for an action to be executed at 2 seconds, I don't really wish for it to wait for 2.00001 seconds. ;) |
| 00:59:57 | <conal> | MonadBaastad: maybe no one thought to implement the Map Monad instance. |
| 01:00:21 | <Baughn> | It's not entirely obvious what it'd do |
| 01:00:28 | <mauke> | return x = ? |
| 01:00:33 | <Baughn> | singleton |
| 01:00:33 | <conal> | Baughn: maybe there are other uses. i forget. |
| 01:00:41 | <mauke> | Baughn: what |
| 01:01:04 | <Baughn> | mauke: Or maybe not. I haven't give it much thought. :) |
| 01:01:12 | <Baughn> | mauke: I got stuck trying to implement Applicative for Map |
| 01:01:27 | <conal> | my "denotational design" paper answers this question. |
| 01:01:31 | <conal> | about Map monad. |
| 01:02:19 | <conal> | there's a simple, general principle for answering questions like this one. |
| 01:02:41 | <MonadBaastad> | There *is* an Applicative inst. for Map, though. :) |
| 01:02:43 | <Baughn> | "First implement it, then notice you implemented a monad"? |
| 01:02:51 | <conal> | :) |
| 01:03:15 | <mauke> | what is pure for Maps? |
| 01:03:22 | <MonadBaastad> | Baughn: The Force is _strong_ with you. :) |
| 01:04:02 | <copumpki> | Map as in Data.Map? |
| 01:04:21 | <Baughn> | mauke: ..am I allowed to have it read a global variable updated with the last key to be written into a map? |
| 01:04:30 | <mauke> | Baughn: hah |
| 01:04:45 | <mauke> | that doesn't even work |
| 01:05:09 | <duke83> | hi there.. how can I read a file, which contains some haskell code, and execute what it has read? is it possible. using import a module is not an option for me... the file mainly contains some variable declarations |
| 01:05:42 | <Baughn> | @src Ord |
| 01:05:43 | <lambdabot> | class (Eq a) => Ord a where |
| 01:05:43 | <lambdabot> | compare :: a -> a -> Ordering |
| 01:05:43 | <lambdabot> | (<), (<=), (>), (>=) :: a -> a -> Bool |
| 01:05:43 | <lambdabot> | max, min :: a -> a -> a |
| 01:06:12 | <mauke> | duke83: if it's just some variables, why do you have to execute code? |
| 01:06:35 | <duke83> | you are right... execute is the wrong word |
| 01:06:45 | <Baughn> | duke83: Easier (and safer) to use something like JSON, perhaps |
| 01:06:53 | <mauke> | sounds like you want a config parser |
| 01:07:04 | <mauke> | for a quick and dirty solution, see Read |
| 01:08:15 | <duke83> | thx, I thought it will read everything as a string, but I am probably wrong |
| 01:08:20 | <Baughn> | mauke: I've got it. pure = undefined. |
| 01:09:06 | <solidsnack> | It's dawned on me that a container which is also a monoid is filterable. |
| 01:09:21 | <solidsnack> | However, a container that is a MonadPlus is also filterable. |
| 01:09:29 | <solidsnack> | So say we want a generic filter. |
| 01:11:05 | <solidsnack> | So say I create a Filterable class and then set instances for `(MonadPlus m)` and also `(Monoid (m t))` -- then they will overlap in some cases. |
| 01:11:11 | <solidsnack> | What will happen? |
| 01:11:40 | <shapr> | ACTION grumbles |
| 01:11:52 | <shapr> | Ubuntu server only has ghc 6.8, I guess I'm switching to debian. |
| 01:12:17 | <solidsnack> | Maybe is is best to insist that an MonadPlus be given Monoid instance. |
| 01:12:28 | <conal> | solidsnack: theoretically (ignoring how instance selection really works in haskell), there's no problem unless the instances disagree. |
| 01:12:28 | <Baughn> | solidsnack: To begin with, you can't do that. IIRC, "instance Foo a => Bar a" works the same way as just "instance Bar a", presumably on the idea that you can always define a Foo /somewhere/, so if it didn't you might get overlaps |
| 01:12:36 | <solidsnack> | s/an MonadPlus/any MonadPlus/ |
| 01:12:54 | <conal> | solidsnack: secondly, what Baughn just said. |
| 01:12:58 | <Baughn> | solidsnack: ISTR there was some incredibly complicated solution involving overlapping and incoherent types for that, but... |
| 01:13:34 | <Baughn> | solidsnack: You're better off just picking one and putting the restriction in the class definition |
| 01:13:41 | <solidsnack> | Ah. |
| 01:13:43 | <solidsnack> | Okay. |
| 01:13:48 | <conal> | not having the instance constraints considered during instance selection is a real pain sometimes. |
| 01:14:04 | <conal> | e.g., two functors compose to a functor, but to contravariant functors compose to a functor also. |
| 01:14:33 | <conal> | also a contra-functor composed with a functor is a contra-functor; and a functor composed with a contra-functor is a contra-functor |
| 01:14:52 | <FunctorSalad> | conal: add a dummy parameter to the typeclass? |
| 01:14:53 | <solidsnack> | conal: So this would be in an instance of which class? |
| 01:15:10 | <FunctorSalad> | (assuming you're using your own functor class) |
| 01:15:14 | <conal> | FunctorSalad: to the Functor typeclass? |
| 01:15:29 | <conal> | solidsnack: Functor in the first example and Co(ntra)Functor in the second |
| 01:16:06 | <FunctorSalad> | not sure whether this would work |
| 01:16:07 | <Baughn> | What's the type signature of contrafmap? |
| 01:16:33 | <FunctorSalad> | class Functor tag f where fmap :: (a->b)->f a->f b |
| 01:16:42 | <FunctorSalad> | or does tag have to occur in the body? |
| 01:16:51 | <conal> | Baughn: cofmap :: (a -> b) -> (f b -> f a) |
| 01:16:57 | <mauke> | FunctorSalad: only if you want to actually use fmap |
| 01:17:16 | <Baughn> | conal: ..that's even /possible/? |
| 01:17:17 | <conal> | Baughn: it's in the TypeCompose package. very handy! |
| 01:17:30 | <conal> | Baughn: i run into it a lot. |
| 01:17:48 | <conal> | Baughn: for any *reciever* of values. |
| 01:17:54 | <FunctorSalad> | mauke: wasn't there some trick to make it choose the instance for any tag? |
| 01:18:11 | <mauke> | unknown |
| 01:18:28 | <solidsnack> | So the problem in this case is that we want an instance like `instance (CoFunctor a, CoFunctor b) => Functor (something) where fmap = cofmap . cofmap` ? |
| 01:19:00 | <pumpkin> | isn't it contramap? |
| 01:19:03 | <conal> | Baughn: if i know how to receive b and i know how to turn an a to b, then i know how to recieve a. nice, huh? |
| 01:19:24 | <conal> | solidsnack: yeah |
| 01:19:32 | <solidsnack> | pumpkin: Maybe so. I'm asking because the context for the example is a little unfamiliar. |
| 01:19:39 | <pumpkin> | http://comonad.com/haskell/category-extras/dist/doc/html/category-extras/Control-Functor-Contra.html#v%3Acontramap |
| 01:20:16 | <Baughn> | conal: Right you are. Huh.. |
| 01:20:25 | <solidsnack> | I guess something would be, uhm, well, I have no idea. Would it be `TypeCompose a b` ? |
| 01:20:27 | <Baughn> | conal: ..that's my dose of brain-mangling for today, I think. |
| 01:20:47 | <conal> | pumpkin: oh, thx. didn't know about that class. geez, there's a lot in category-extras. |
| 01:20:50 | <solidsnack> | (Or something else to accomplish type composition.) |
| 01:20:53 | <conal> | Baughn: :) |
| 01:21:03 | <pumpkin> | I remember making a contrafunctor instance for the composable folds thing |
| 01:21:09 | <pumpkin> | but it was a bit of a stretch |
| 01:21:23 | <conal> | solidsnack: yes. see the TypeCompose package. |
| 01:21:35 | <conal> | solidsnack: @hackage TypeCompose |
| 01:21:40 | <conal> | @hackage TypeCompose |
| 01:21:41 | <lambdabot> | http://hackage.haskell.org/cgi-bin/hackage-scripts/package/TypeCompose |
| 01:22:08 | <solidsnack> | Aha, thank you -- that would be the missing context. |
| 01:22:11 | <conal> | i got type (constructor) composition from the applicative functor paper. |
| 01:22:23 | <conal> | functors compose to functors, and applicatives compose to applicatives. |
| 01:22:33 | <conal> | monads sometimes compose to monads, but sometimes not. |
| 01:22:42 | <conal> | and even when monads compose, they don't compose uniformly. |
| 01:22:51 | <conal> | unlike functors & applicatives |
| 01:23:07 | <conal> | that's why we don't need "applicative transformers". |
| 01:23:17 | <conal> | applicatives + compose are their own transformers |
| 01:23:53 | <conal> | in practice the composition newtype wrapper ("O") is awkward. |
| 01:26:32 | <solidsnack> | Interesting. |
| 01:28:54 | <conal> | using type composition, i've been able to replace code with type compositions. |
| 01:30:06 | <solidsnack> | Through default instances in type classes? |
| 01:31:33 | <conal> | not defaults. if a type is defined to be a composition, then it doesn't need (and can't have) an functor or applicative instance. |
| 01:32:16 | <solidsnack> | So through the, uhm, "obvious" composition of their respective operations? |
| 01:32:45 | <conal> | solidsnack: i don't know if it'd be obvious to you or not. it's in the applicative functor paper. |
| 01:33:09 | <solidsnack> | Ah, okay. |
| 01:33:11 | <idnar> | hooray for applicative functors, I guess |
| 01:33:34 | <pumpkin> | yay |
| 01:33:44 | <solidsnack> | So it looks like I'm wrong about this monoid/filterable thing. |
| 01:33:58 | <conal> | by getting familiar with the functor & applicative instances for maybe, pairing, functions, etc, and composition, i've been able to make lots of code just vanish. |
| 01:34:25 | <solidsnack> | I need `toList` or something `:: monoid (monoid t) -> monoid t` (`join` would suffice in many cases) |
| 01:34:48 | <mauke> | :t mconcat |
| 01:34:49 | <lambdabot> | forall a. (Monoid a) => [a] -> a |
| 01:35:02 | <solidsnack> | I don't get to start with a list, though. |
| 01:35:07 | <pumpkin> | a joinable monoid :o |
| 01:35:15 | <idnar> | solidsnack: `:: monoid (monoid t) -> monoid t` doesn't make any sense |
| 01:35:20 | <conal> | solidsnack: hm. Monoid classifies types, not type constructors. |
| 01:35:23 | <pumpkin> | Sum (Sum Int) |
| 01:35:31 | <solidsnack> | filter :: (a -> Bool) -> monoid t -> monoid t |
| 01:35:32 | <pumpkin> | :P |
| 01:35:40 | <solidsnack> | monoid is a type variable |
| 01:35:45 | <solidsnack> | Monoid is a type class |
| 01:35:48 | <mauke> | oh, so it's not a Monoid? |
| 01:35:56 | <idnar> | solidsnack: right, so monoid t is impossible |
| 01:35:58 | <mauke> | \o_O/ |
| 01:36:06 | <solidsnack> | right, |
| 01:36:23 | <solidsnack> | i mean, the thing such that `monoid t` is in `Monoid` |
| 01:36:38 | <idnar> | I wonder if type constructors that construct monoids are a useful class by themselves |
| 01:36:57 | <conal> | TypeCompose has the class Monoid_f |
| 01:37:00 | <conal> | for that purpose |
| 01:37:23 | <idnar> | ah, so (Monoid_f m) => m (m t) -> m t |
| 01:37:42 | <pumpkin> | joinoid! |
| 01:37:49 | <conal> | maybe. though i'd guess Monad & join |
| 01:37:49 | <pumpkin> | moin! |
| 01:37:59 | <idnar> | but you'd need some kind of extension to include a join-like operation |
| 01:38:21 | <conal> | solidsnack: why not use Monad & join? |
| 01:38:35 | <solidsnack> | Well, I need it to be MonadPlus, I think. |
| 01:38:43 | <solidsnack> | That's the route I'm going to take, actually. |
| 01:39:04 | <idnar> | @type msum |
| 01:39:05 | <lambdabot> | forall (m :: * -> *) a. (MonadPlus m) => [m a] -> m a |
| 01:39:11 | <idnar> | @type mplus |
| 01:39:12 | <lambdabot> | forall (m :: * -> *) a. (MonadPlus m) => m a -> m a -> m a |
| 01:39:17 | <idnar> | mplus isn't always the same thing as mappend, though |
| 01:39:53 | <idnar> | uhm, do you have an example Monoid instance in mind? I'm trying to think of a concrete scenario here, and I'm not coming up with anything |
| 01:39:57 | <conal> | and join isn't necessarily related to mplus, is it? |
| 01:40:11 | <idnar> | conal: yeah... |
| 01:40:19 | <solidsnack> | I'll be done in a second and paste it. |
| 01:40:28 | <idnar> | uhm, and msum is only for lists |
| 01:40:36 | <solidsnack> | Not using it :) |
| 01:40:41 | <Peaker> | conal: src/FRP/Reactive/GLUT/Adapter.hs:48:39: Not in scope: type constructor or class `:+->' |
| 01:40:48 | <idnar> | ACTION twists his brain some more |
| 01:40:48 | <Peaker> | conal: but the quickcheck version issues seem to be fixed |
| 01:40:49 | <conal> | oops! |
| 01:41:04 | <amckinley> | hey, would anyone like to help me with a small parsec parser? :) |
| 01:41:05 | <idnar> | > join [[5, 6], [7, 8]] |
| 01:41:06 | <lambdabot> | [5,6,7,8] |
| 01:41:07 | <idnar> | > msum [[5, 6], [7, 8]] |
| 01:41:11 | <lambdabot> | [5,6,7,8] |
| 01:41:13 | <amckinley> | http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5108#a5108 |
| 01:41:58 | <idnar> | oh man, parsec; I think that signals sleeptime for me |
| 01:42:03 | <conal> | Peaker: oh, i see. i missed a version dependency. |
| 01:42:24 | <idnar> | amckinley: uhm, you seem to have a spelling or search/replace problem there |
| 01:42:32 | <amckinley> | idnar: come on; its fun :) |
| 01:42:36 | <amckinley> | orly? |
| 01:42:45 | <idnar> | oh, wait, this is worse than I thought |
| 01:42:56 | <idnar> | my poor head |
| 01:43:01 | <amckinley> | :D |
| 01:43:19 | <amckinley> | i decided to learn haskell by writing a parser for bind config files |
| 01:43:46 | <amckinley> | the problem im having is with input that looks like this: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5109#a5109 |
| 01:43:58 | <idnar> | uhm, so, I'm confused; is the mangling in that paste a joke, or what? |
| 01:44:18 | <idnar> | or is my browser doing something silly? |
| 01:44:27 | <idnar> | oh, my browser is doing something silly |
| 01:44:40 | <amckinley> | idnar: not sure; it looks reasonable to me, but that might be wishful thinking :) |
| 01:44:44 | <conal> | Peaker: released reactive-glut 0.1.9, with reactive version dependency fixed. |
| 01:44:58 | <Peaker> | conal: reinstalling.. |
| 01:45:10 | <idnar> | this is what firefox is doing: http://slipgate.za.net/~mithrandi/dear-firefox-wtf.png |
| 01:45:25 | <Peaker> | conal: It really seems like this cabal business needs more automation of the dependencies... |
| 01:45:28 | <idnar> | but, for example, if I copy the text into gnome-terminal, it comes out okay |
| 01:45:35 | <amckinley> | madness |
| 01:45:42 | <conal> | Peaker: indeed! :( |
| 01:45:56 | <conal> | Peaker: we need more static checking. |
| 01:46:40 | <idnar> | what about static electricity! |
| 01:46:45 | <conal> | :) |
| 01:46:49 | <Peaker> | conal: yeah.. a type signature of the exported modules to check against -- verifying the version dependency holds |
| 01:46:53 | <conal> | maybe if we had a real module system, like ML |
| 01:46:58 | <Peaker> | conal: seems to compile now |
| 01:47:01 | <conal> | :) |
| 01:47:05 | <Peaker> | conal: :-) |
| 01:47:05 | <solidsnack> | Okay, this works for lists. |
| 01:47:06 | <amckinley> | idnar: very weird; my firefox is fine. i could switch over to regular pastebin... |
| 01:47:06 | <solidsnack> | http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5110#a5110 |
| 01:47:16 | <idnar> | amckinley: I think it's because I haven't restarted it since I upgraded the binary |
| 01:47:30 | <Peaker> | conal: hey cool, you fixed fieldtrip? |
| 01:47:34 | <conal> | this reactive version may have broken integral. one example in reactive-fieldtrip just sits there. |
| 01:47:38 | <stepnem> | please, what is the most painless way to get only the elements occuring more than once in a list? |
| 01:47:42 | <idnar> | amckinley: another fun thing is that some dialogs just have XML parse errors instead of their expected contents |
| 01:47:46 | <conal> | Peaker: great :) |
| 01:47:59 | <amckinley> | idnar: yeah, ive seen that before :) |
| 01:48:02 | <Peaker> | conal: reactive-fieldtrip fails, though |
| 01:48:07 | <idnar> | amckinley: another fun thing is that some dialogs just have XML parse errors instead of their expected contents |
| 01:48:37 | <Peaker> | src/FRP/Reactive/FieldTrip.hs:30:9: Could not deduce (Bounded tr) from the context (Transform (BehaviorG tr tf xf) (BehaviorG tr tf a), Ord tr, Transform xf a) |
| 01:48:54 | <solidsnack> | So it looks like things in `MonadPlus` have enough structure to be generically filtered; monoids don't because you can't walk them and aggregate the walk. |
| 01:48:56 | <conal> | Peaker: hm. i'll release a new reactive-fieldtrip then. |
| 01:49:07 | <Peaker> | conal: awesome :-) |
| 01:49:14 | <idnar> | solidsnack: yeah, monoids are values not containers |
| 01:49:41 | <Peaker> | conal: I am really anxious to start toying with Reactive.. Feel more comfortable with my understanding of things now to actually debug stuff even, hopefully |
| 01:49:48 | <idnar> | solidsnack: it might be useful for the values /in/ the container to be monoids, though |
| 01:50:03 | <Peaker> | conal: ghc could use more debugging tools.. especially for threads |
| 01:50:11 | <solidsnack> | idnar: How? |
| 01:50:30 | <Peaker> | Something visualizing threads/mvar state could be pretty cool |
| 01:50:31 | <idnar> | solidsnack: I just mean (Monad m, Monoid a) => m a |
| 01:50:45 | <conal> | Peaker: and how. this stuff drives me nuts. |
| 01:50:49 | <solidsnack> | idnar: I don't see that as helping with filter, though. |
| 01:50:52 | <codebliss> | So what do you guys say when you read (::)? I got a new idea |
| 01:50:54 | <Baughn> | conal: checkers-0.1.4 appears broken; you've forgotten to depend on quickcheck at all |
| 01:50:55 | <idnar> | solidsnack: no, not really |
| 01:51:06 | <conal> | really?! |
| 01:51:06 | <codebliss> | "Has the type of" is pretty good, but when you think functions, it's kinda meh |
| 01:51:07 | <idnar> | solidsnack: what you really want is some kind of sequence / enumerable / traversable / whatever typeclass |
| 01:51:08 | <solidsnack> | codebliss: I say "has type". |
| 01:51:21 | <codebliss> | solidsnack: I just thought of "represents." It seems to fit nice. |
| 01:51:22 | <idnar> | solidsnack: MonadPlus isn't really enough to "walk" it either |
| 01:51:38 | <solidsnack> | idnar: Well, it seems I just need `MonadPlus` |
| 01:51:47 | <idnar> | codebliss: (::) reads silently in my head |
| 01:51:56 | <Baughn> | conal: No, not really. Hmm |
| 01:51:59 | <codebliss> | idnar: Of course, what if you were explaining to newbies? =P |
| 01:52:01 | <solidsnack> | idnar: `fmap` let's me walk it and `join` let's me collapse it again. |
| 01:52:17 | <Baughn> | conal: You've forgotten to specify the version, and installing reactive pulled in a newer version at one point |
| 01:52:24 | <Baughn> | conal: So the one checkers actually depends on is hidden |
| 01:52:38 | <idnar> | codebliss: I'd just say that it's syntax for a type declaration |
| 01:52:39 | <codebliss> | idnar: map represents an A to a B to an A list to a B list |
| 01:52:44 | <solidsnack> | idnar: The presence of a zero element is what allows the collapse to accomplish the filtration. |
| 01:52:49 | <idnar> | codebliss: I'm not big on translating programming language syntax to english |
| 01:52:52 | <Peaker> | Baughn: sounds like an installed-package-depends-on-specific-version-it-happens-to-be-compiled-with issue |
| 01:52:54 | <conal> | Baughn: i think someone (you?) suggested i leave off the version. i'm in a muddle about what version to use for what. |
| 01:52:57 | <codebliss> | idnar: Yeah, it doesn't work well. |
| 01:53:09 | <codebliss> | idnar: I was teaching one of my professors and kind of hit a wall once |
| 01:53:15 | <Peaker> | conal: I suggested that, and I think if you are version neutral, it is a good idea |
| 01:53:25 | <conal> | okay new reactive-fieldtrip |
| 01:53:28 | <Peaker> | conal: but maybe I'm missing something |
| 01:53:30 | <solidsnack> | (It was the zero element that led me to monoid in the first place.) |
| 01:53:35 | <idnar> | codebliss: I'm not sure why "has type" is any worse for functions, though |
| 01:53:37 | <Baughn> | conal: It's a good idea only if all the possible versions of quickcheck would actually work with checkers |
| 01:53:43 | <Peaker> | Baughn: they don't? |
| 01:53:53 | <Baughn> | conal: quickcheck-2.1.0.1 doesn't; they've changed the interface |
| 01:53:57 | <idnar> | f :: a -> b would read something like "f has type function from a to b" |
| 01:54:16 | <Baughn> | conal: Say QuickCheck<2 or something, at least |
| 01:54:24 | <Peaker> | conal: reactive-fieldtrip installed successfully, yay! |
| 01:54:40 | <conal> | Baughn: i think QuickCheck<2 is what it said before i updated. |
| 01:54:46 | <codebliss> | idnar: I think more of a currying sense. Functions represent conversions of types to me, so when I think foo has type (A list to B list) to A list to B list, it kinda clashes |
| 01:55:03 | <Peaker> | conal, Baughn: If that's truly the case then reactive truly conflicts with Stream >= 0.3.1 ? |
| 01:55:09 | <conal> | clearly, i could use some help sorting out this stuff. |
| 01:55:26 | <idnar> | codebliss: any non-trivial function type seems like a mess to me if you try read it out loud ;) |
| 01:55:28 | <Peaker> | conal, Baughn: Maybe its better to update checkers/reactive to use the new quickcheck interface so that bugfixes/etc in Stream can be used? |
| 01:55:37 | <codebliss> | idnar: Very true XD |
| 01:55:41 | <conal> | Peaker: yes! |
| 01:56:01 | <Peaker> | Baughn: do you know exactly what interface changed in quickcheck 2.1.0.1? |
| 01:56:24 | <amckinley> | Cale: around? im back for more parsec abuse :) |
| 01:56:38 | <idnar> | codebliss: I used to work with a friend, and every now and then I'd shout some code out to him, but I'd sorta leave out half of it because we both knew what I meant; everyone else in earshot used to get horribly confused, though |
| 01:56:43 | <Baughn> | Peaker: No, but if you try compiling it it'll become obvious in short order. :) |
| 01:56:50 | <codebliss> | idnar: Hahahah XD |
| 01:57:09 | <codebliss> | idnar: You got haskell in a real job? |
| 01:57:26 | <Peaker> | Baughn, conal: Ok, trying... |
| 01:57:26 | <idnar> | codebliss: unfortunately not, this was mostly Python code |
| 01:57:36 | <Peaker> | aha, no more: Test.QuickCheck.Utils |
| 01:57:39 | <codebliss> | Python is much better than C's/Java |
| 01:58:02 | <Baughn> | conal: Meanwhile, I'm just specifying Stream-0.3 manually. THat's the quick fix to get it working on hackage; you can port to quickcheck-2 later |
| 01:58:02 | <idnar> | I'm still hoping that Haskell is going to be my primary language in the long run, but I'm wondering if it'll ever be ready for that |
| 01:58:16 | <conal> | Baughn: which package? |
| 01:58:24 | <Baughn> | conal: Basically, wherever you depend on Stream, say Stream == 0.3 |
| 01:58:25 | <conal> | hm. how does one get QuickCheck 2? |
| 01:58:31 | <Baughn> | cabal-install. :P |
| 01:58:41 | <conal> | just tried. |
| 01:58:49 | <conal> | maybe i need a cabal update. |
| 01:58:54 | <Baughn> | Oh, right.. |
| 01:59:00 | <Baughn> | No, it defaults to quickcheck-1 |
| 01:59:05 | <conal> | oh |
| 01:59:10 | <Baughn> | You need to say cabal install quickcheck-2.1.0.1 |
| 01:59:17 | <conal> | ah! thx. |
| 01:59:29 | <idnar> | is that version number in trinary? >:) |
| 01:59:34 | <Baughn> | Yes |
| 01:59:41 | <Peaker> | ACTION tries to figure out exactly what old quickcheck stuff checkers needs and where they are in the new quickcheck |
| 01:59:52 | <idnar> | that's, what, 0x64? |
| 02:00:03 | <idnar> | er, 0x40 |
| 02:00:27 | <idnar> | I'd better find a bed before I pass out on my keyboard and spam the channel |
| 02:02:01 | <Zao> | conal: Kudos for the post on C, it had me laughing a bit. |
| 02:02:05 | <Peaker> | hmm.. testAssocativeBy/etc from Test.QuickCheck.Utils just disappeared in the new quickcheck? |
| 02:02:13 | <Peaker> | Zao: where? |
| 02:02:17 | <conal> | Zao: thanks! :D |
| 02:02:25 | <Zao> | Peaker: http://conal.net/blog/posts/the-c-language-is-purely-functional/ |
| 02:02:34 | <Zao> | Trying to catch up on my RSS, two days old :) |
| 02:03:05 | <conal> | Peaker: i see the same thing (no Test.QuickCheck.Utils) |
| 02:03:28 | <conal> | coarbitrary moved out of Arbitrary |
| 02:03:42 | <Peaker> | conal: I can't find any alternative either, and nothing on google about those functions |
| 02:04:00 | <Peaker> | conal: which quickcheck had some release notes or such |
| 02:04:09 | <Peaker> | conal: can just duplicate the old library code into checkers :-( |
| 02:05:30 | <p_l> | conal: That post is ... I lack the words to describe its awesomness |
| 02:06:08 | <conal> | p_l: wow! :D |
| 02:06:39 | <Baughn> | conal: And now I can go back to my original proof-of-concept code. server :: Event Command -> Reader (Event Reply) |
| 02:07:03 | <gnuvince_> | If I have the following code: do { t1 <- getPOSIXTime; let x = f y; t2 <- getPOSIXTime; return (x, t2 - t1) }, is the time difference the time it took to run the f function or the time it took to generate the thunk? |
| 02:07:06 | <p_l> | conal: the possibilities for Mind Screw among other langauge lawyers are too big for lower grade :D |
| 02:07:08 | <Baughn> | p_l: Also, link to post of Kamina-like awesomness? |
| 02:07:08 | <Peaker> | conal: nice post :-) The Haskell functional part is significantly stronger than cpp, though :-) |
| 02:07:15 | <Baughn> | Wait.. |
| 02:07:20 | <Baughn> | The C=FP one? o_O |
| 02:07:23 | <p_l> | Yeah |
| 02:07:27 | <conal> | Peaker: it sure is. |
| 02:07:32 | <Baughn> | Yeah, that was amusing |
| 02:07:38 | <p_l> | Imagine the possibilities for screwing with people's minds :D |
| 02:07:50 | <Peaker> | conal: any idea why the quickcheck guys would eliminate seemingly useful functions like testAssociativeBy? |
| 02:07:56 | <Baughn> | I always thought C was more functional than a lot of other languages, though. |
| 02:08:01 | <Baughn> | At least it /has/ function pointers. ;) |
| 02:08:05 | <conal> | Peaker: the functional parts differ in *degree* only. so if you call one pure you have to call the other pure. |
| 02:08:23 | <Baughn> | conal: ..and my intelligence differs in degree only from a rock. |
| 02:08:25 | <conal> | Peaker: that's the real point of my post. i'd been wondering for a while how to make that point in a way that people would hear. |
| 02:08:38 | <conal> | Baughn: then you're equally "pure" |
| 02:09:40 | <p_l> | Baughn: And some stuff that can be done with those when using C as second stage can be quite amusing :D |
| 02:10:18 | <Peaker> | conal: I do think that there's an interesting point to be made about Haskell's power/referential transparency, even when dealing with IO. As you say, its probably mostly a quantitive rather than a qualitive thing, but also there's the issue of cpp handling possibly invalid syntactic elements and pure Haskell handling type-safe imperative actions |
| 02:11:01 | <conal> | the post is my way of challenging the common description of Haskell-with-IO as "purely functional". the defense that people often make, when pressed, applies also to C. |
| 02:11:12 | <conal> | that's what i was getting at. |
| 02:11:39 | <conal> | perhaps with humor, i won't be flame-broiled when i point it out. |
| 02:11:46 | <Peaker> | conal: what if you ignore the existence of CPP? |
| 02:11:48 | <sohum> | so I'm trying to use cabal for the first time, and I get 'Could not find module Data.Data' |
| 02:12:08 | <gwern> | Peaker: personally, I'm wondering what you get by combining CPP+Haskell |
| 02:12:10 | <conal> | Peaker: then you're left with just IO (C) |
| 02:12:22 | <Baughn> | sohum: What GHC version are you on? o_O |
| 02:12:23 | <conal> | gwern: without IO? |
| 02:12:31 | <gwern> | sohum: need more details. possibly you are missing a dep in the .cabal |
| 02:12:41 | <Baughn> | No, Data.Data is in base |
| 02:12:54 | <sohum> | Baughn: "The Glorious Glasgow Haskell Compilation System, version 6.8.2" |
| 02:12:57 | <gwern> | Baughn: maybe he's missing base as a dep! |
| 02:12:59 | <conal> | Baughn: hey -- when someone acts strangely here, we can ask "What (GHC version) are you on?" |
| 02:13:34 | <gwern> | conal: heck, I dunno. you're the one with all this C IO stuff. (I didn't follow it even reading the blog post, so how would I know?) |
| 02:13:53 | <sohum> | it was trying to install containers-0.2.0.1, if that helps |
| 02:14:31 | <sohum> | ghc and co are installed via debian, and now I'm just starting to try to use cabal |
| 02:14:42 | <Peaker> | is the quickcheck maintainer in #haskell ? |
| 02:14:54 | <gwern> | Peaker: don't think so |
| 02:15:16 | <gwern> | sohum: containers is one of the core packages; trying to install it generally says you're doing something wrong |
| 02:15:24 | <Peaker> | gwern: the point is that even the C AST is a referentially transparent value that when compiled/executed, will yield a value, has similar semantics to IO () values in Haskell, even if the combinators on it are pretty crappy |
| 02:15:45 | <conal> | Peaker: spot on! |
| 02:15:55 | <gwern> | Peaker: see, I don't know what it means to say an ast is a referentially transparent value |
| 02:15:57 | <sohum> | gwern: ok, so odds are that cabal doesn't know how to find the preinstalled packages? |
| 02:16:00 | <Peaker> | So C is purely functional like Haskell, except the purely functional part is really really weak and cannot do anything with these action values |
| 02:16:39 | <gwern> | sohum: I really have no idea. I ran into so many issues installing anything with debian I quickly gave up |
| 02:16:50 | <conal> | Peaker: you got it. :) so if we want to talk about how Haskell is different from C, we cannot (honestly) use the word "purely". |
| 02:16:57 | <gwern> | as far as I'm concerned, if you're intermingling debian and cabal-install, don't waste #haskell's time asking about it |
| 02:17:03 | <Peaker> | gwern: the AST is stateless, calling a CPP macro or using the expanded AST instead is equivalent. Or without CPP, you simply cannot call functions, so its referentially transparent in an .. empty way |
| 02:17:05 | <conal> | more honestly, we can say that haskell is a "very functional programming language" |
| 02:17:33 | <Peaker> | conal: I think the qualitive difference is the fact Haskell has powerful purely functional combinators on these IO values |
| 02:17:41 | <conal> | and if we really want to say "purely", we can stop using IO and come up with something genuinely functional to replace it. |
| 02:18:03 | <conal> | Peaker: yes. *more* powerful than cpp's |
| 02:18:09 | <conal> | it's still relative. |
| 02:18:11 | <solidsnack> | conal: I think this is the point where we all say "and what is that?" |
| 02:18:15 | <conal> | not absolute. |
| 02:18:53 | <conal> | solidsnack: for now, it's getting back to exploring that very question, as people used to do before monadic imperative functional programming was invented. |
| 02:18:53 | <Peaker> | conal: Yeah, I agree -- so the words "purely functional" are not that great to describe "powerful combinators on actions as first-class values" |
| 02:19:04 | <Peaker> | conal: which is a true, genuine advantage of Haskell as an *imperative language* |
| 02:19:10 | <conal> | Peaker: yeah. because "purely" is an absolute, not a relative. |
| 02:19:22 | <conal> | Peaker: yep. haskell is a better C |
| 02:19:29 | <conal> | so we can call it that. |
| 02:19:33 | <mmorrow_> | and we have a better CPP |
| 02:19:39 | <mmorrow_> | , ppDoc `fmap` fmaps 8 |
| 02:19:42 | <lunabot> | fmap . (fmap . (fmap . (fmap . (fmap . (fmap . (fmap . fmap)))))) |
| 02:19:43 | <solidsnack> | As far as I can tell, Haskell without unsafePerformIO is pure, per Amr Sabry's paper. |
| 02:19:46 | <conal> | on haskell.org: "Haskell: a better C" |
| 02:19:55 | <Peaker> | conal: but this advantage is so huge that it makes others' languages Haskell libraries. It deserves a name, more than "better C" :) |
| 02:20:03 | <conal> | solidsnack: yes. just like cpp+c. |
| 02:20:20 | <solidsnack> | Oh, so CPP is a purely functional langauge? |
| 02:20:30 | <conal> | solidsnack: did you read my post? |
| 02:20:33 | <solidsnack> | That's a little different. |
| 02:20:36 | <conal> | Peaker: i agree. "better C" is honest, but lacks something. |
| 02:20:40 | <Peaker> | solidsnack: whose output is an action, that when executed will perform side effects and yield a result, like Haskell's main :) |
| 02:20:53 | <solidsnack> | Peaker: Sure. Haskell is a better CPP. |
| 02:21:17 | <solidsnack> | Or a better `++` |
| 02:21:30 | <mmorrow_> | (+1) |
| 02:21:38 | <solidsnack> | I'd much rather get my abstraction from Haskell :) |
| 02:21:50 | <conal> | the reason i suggest "Haskell: a better C" is that we typically use "haskell" to mean "haskell+IO", and use "C" to mean "cpp+C" |
| 02:21:54 | <Peaker> | conal: In Haskell, I can have an imperative program be typed via some newtype or type synonym, which is actually some monad transformer, and by changing this monad transformer in one place, I'm changing the statement bind operator behavior, something unheard of in other imperative languages.. Perhaps Haskell is an "overridable semicolon language" :) |
| 02:21:58 | <jeffz`> | sohum, you might find it easier to ditch the debian specific stuff and use the haskell platform instead, that way things "just work" |
| 02:22:02 | <solidsnack> | conal: Sure, I understand that. |
| 02:22:09 | <solidsnack> | However, grauenwolf doesn't. |
| 02:22:19 | <conal> | Peaker: yes. that's part of what makes haskell a better C. |
| 02:22:23 | <solidsnack> | Altough you don't mean to, you are feeding the trolls. |
| 02:22:27 | <dibblego> | solidsnack, grauenwolf doesn't understand very much -- it's best to accept that fact |
| 02:22:38 | <Peaker> | conal: But its a huge, qualitive, and not just quantitive advantage. Even if the quality happens not to have much to do with functional-ness or purity |
| 02:22:57 | <solidsnack> | dibblego: He's unique in his aggressiveness but not in his confusion. |
| 02:23:16 | <conal> | Peaker: yep. i don't seriously want us to change hakskell's motto to "a better C". i want it to be honest. |
| 02:23:24 | <Peaker> | conal: I think that maybe because Haskell tends to be mind bending for a while, until you grasp the concepts well (and then its pretty simple, in retrospect :-) there's a lot of bad terminology and communication around |
| 02:23:35 | <conal> | Peaker: yeah. |
| 02:24:10 | <Peaker> | so we ought to rename "purely functional" to "overridable language semantics" or some such :) |
| 02:24:32 | <conal> | something. let's keep working on it until it's both honest and compelling. |
| 02:24:52 | <conal> | now that i've spilled what my post is really about, maybe some of you will want to add comments on the post. |
| 02:24:59 | <Baughn> | "A Kamina-like language"? |
| 02:26:31 | <conal> | when i've talked about these issues more directly on #haskell, i get some defensive reactions. |
| 02:27:11 | <dibblego> | conal, s/wisdon/wisdom |
| 02:27:44 | <conal> | dibblego: thx! |
| 02:28:22 | <solidsnack> | conal: I think this topic is under-developed. |
| 02:28:51 | <conal> | solidsnack: yeah? please say more. |
| 02:29:07 | <conal> | dibblego: fixed |
| 02:30:01 | <solidsnack> | It seems that this problem basically boils down to how deterministic we can make things. |
| 02:30:19 | <solidsnack> | Or rather, how much of their determinism can we make observable. |
| 02:31:19 | <solidsnack> | It seems like their should be some dovetailing with distributed systems -- in particular, the proof techniques being developed there. |
| 02:32:13 | <solidsnack> | This is something which I still need to learn more about, naturally. |
| 02:32:19 | <conal> | solidsnack: i'm not tracking you at all |
| 02:32:47 | <mmorrow_> | 444444444444444444444555555555555555555555555555555555555555555555555555555555555555555555555555555555553333333 |
| 02:32:54 | <mmorrow_> | heh |
| 02:32:57 | <mmorrow_> | good kittay |
| 02:33:14 | <copumpkin> | :o |
| 02:33:19 | <copumpkin> | u haz a cat |
| 02:33:25 | <solidsnack> | conal: Well, my ideas are not very well developed, either. |
| 02:34:02 | <conal> | solidsnack: ok. np. |
| 02:34:04 | <p_l> | Baughn: whut? |
| 02:34:31 | <solidsnack> | Having said that, the root of my objection is that, as far as I can tell, as long as you expect components to communicate across a network with unsynchronized time and lossy links, IO is the best that there is. |
| 02:35:12 | <conal> | solidsnack: best that there is or the best that we know? |
| 02:35:30 | <solidsnack> | Given lossy links and unsynched time, best that there is. |
| 02:35:43 | <solidsnack> | This could be a very interesting discussion, actually. |
| 02:36:16 | <conal> | offhand, i can't imagine how you could prove "best there is". |
| 02:36:34 | <solidsnack> | Fair enough. |
| 02:36:36 | <conal> | pretty strong claim. |
| 02:37:08 | <conal> | solidsnack: nice points to keep in mind while looking for something better than IO. |
| 02:37:16 | <conal> | where better == semantically tractable |
| 02:37:52 | <conal> | Peaker: eep. lots breaks in checkers with quickcheck 2. |
| 02:38:10 | <Peaker> | conal: except Utils dependency? |
| 02:38:12 | <copumpkin> | thoughtpolice: around? |
| 02:38:21 | <conal> | Peaker: yeah. i've fixed the Arbitrary/CoArbitrary instances. |
| 02:38:35 | <conal> | Peaker: and now i get to the Utils dependency. |
| 02:38:46 | <Peaker> | conal: I'm not very familiar with quickcheck. I couldn't find the functions from Utils in the new quickcheck |
| 02:39:04 | <solidsnack> | conal: That's fine to cast doubt on my certainties; but you can see how I look at these remarks about an alternative and remain unconvinced. |
| 02:39:06 | <Peaker> | conal: I guess the quick&dirty fix would be to just port that module to new quickcheck and put it in checkers? |
| 02:39:26 | <conal> | Peaker: i could, though i'm not a big fan of q&d. |
| 02:40:12 | <conal> | Peaker: interested in giving it a try? |
| 02:40:51 | <solidsnack> | conal: I guess a fair direction to take this is, what is intractable about IO and can I prove that it falls out of the traits that I described? |
| 02:40:51 | <conal> | Peaker: or we could ask koen or andy or dons |
| 02:41:12 | <solidsnack> | So maybe you can give me a property or something and I can look at that. |
| 02:41:14 | <conal> | solidsnack: i like that direction! |
| 02:42:13 | <conal> | solidsnack: i measure tractability by looking at the denotational domain (a mathematical type). if it's simple, then it's tractable. that's my measure. |
| 02:42:22 | <solidsnack> | Okay. |
| 02:43:06 | <Peaker> | conal: I'm not sure I can do it today (way beyond my sleep time), but I can try to hunt where the functionality has gone to in the next few days |
| 02:43:07 | <boblo2> | can someone pls tell me what the | does in haskell? |
| 02:43:10 | <conal> | solidsnack: IO is such a mess that it doesn't even have a denotational model, afaik. |
| 02:43:37 | <copumpkin> | boblo2: it's like an if in disguise! |
| 02:43:42 | <Peaker> | boblo2: Type sums -- declares that a type can have alternate data constructors (with their alternate fields) as potential values |
| 02:43:53 | <conal> | Peaker: thanks. i sure appreciate it. i'm lousy at tracking a lot of details, so i probably can't do a good job of library maintenance on my own. |
| 02:43:58 | <Peaker> | boblo2: do you know structs, unions and enums? |
| 02:44:18 | <conal> | Peaker: which is also why i like genuinely functional programming (tm) |
| 02:44:26 | <boblo2> | i am playing a "hacker" game that used haskell iv been learneing it the past few days |
| 02:44:28 | <Peaker> | conal: Looking through your code, it certainly doesn't seem that you have trouble tracking details.. |
| 02:44:37 | <conal> | hah! |
| 02:44:54 | <conal> | Peaker: i wonder if you mean the code that works or the code that doesn't. |
| 02:45:08 | <Peaker> | conal: even your imperative implementation of Reactive is pretty nifty :) btw: I wondered why you put the action serializer into Clock? |
| 02:45:17 | <Peaker> | conal: I mean that its beautifully structured |
| 02:45:40 | <conal> | Peaker: if it were's beautifully structured, i wouldn't have a chance. |
| 02:45:46 | <solidsnack> | conal: My suspicion is that anything better depends on a transactional store at some level. |
| 02:46:32 | <conal> | solidsnack: wow. that would an amazing result if one could prove it. |
| 02:47:03 | <solidsnack> | conal: Well, schema consistency would be the canonical example. |
| 02:47:14 | <solidsnack> | As long as you have named (instead of structural) types. |
| 02:47:18 | <conal> | solidsnack: there are non-existence proofs. hard for me to imagine how we'd come up with one. |
| 02:47:31 | <Peaker> | conal: Ugly code does not work very well in general, not just for you. Its just that most people have a lower standard of quality that they expect from their code |
| 02:47:36 | <conal> | solidsnack: i.e., prove the nonexistence of a solution that doesn't use transactional store. |
| 02:47:49 | <boblo2> | can someone tell me what +++ does in ParseLib? forgive me for the noobness |
| 02:48:03 | <conal> | Peaker: yeah. i don't trust code that "works". i only trust code that's "true". |
| 02:48:17 | <solidsnack> | It depends very much on what you want to do, of course. |
| 02:48:24 | <conal> | Peaker: compellingly true. code that's "inevitable". |
| 02:48:39 | <Peaker> | conal: You might have encountered this mentality, its very common, that is saying that it is utterly impossible to write *anything* without bugs, and even security problems, no less. IMO: This mentality stems from the fact that most people don't know how to write beautiful/well-structured code, and that badly structured code is true to that mentality |
| 02:48:41 | <solidsnack> | If you don't want hot update and such, you don't worry about having a transactional store of modules. |
| 02:49:39 | <conal> | Peaker: indeed. and like most claims of "impossible", it's a self-fulfilling prophecy. |
| 02:50:04 | <conal> | in other words when someone says something is impossible, i conclude that it's indeed impossible for them. |
| 02:50:05 | <copumpkin> | boblo2: is it the same one as Arrow? I haven't used ParseLib |
| 02:50:13 | <copumpkin> | :t (+++) |
| 02:50:14 | <lambdabot> | forall (a :: * -> * -> *) b c b' c'. (ArrowChoice a) => a b c -> a b' c' -> a (Either b b') (Either c c') |
| 02:50:14 | <p_l> | Peaker: I wouldn't say impossible, but I'd classify some of it beyond scope of the program itself :) |
| 02:50:15 | <conal> | Peaker: in part because people like to be right. |
| 02:51:25 | <Peaker> | conal: I think many are also very unimaginative :( |
| 02:51:54 | <boblo2> | (+++) :: Parser a -> Parser a -> Parser a |
| 02:52:09 | <boblo2> | what does that mean |
| 02:52:30 | <QtPlaty[HireMe]> | boblo2: That means that the function is a parser combinator. |
| 02:52:35 | <solidsnack> | A statement about possibility is, unfortunately, no more trustworthy than one about impossibility. |
| 02:52:38 | <copumpkin> | boo, not (+++) :: Parser a -> Parser b -> Parser (Either a b) |
| 02:52:55 | <conal> | solidsnack: because it's a statement and not a proof. |
| 02:52:59 | <solidsnack> | I've had a lot more trouble with requirements dictated by "possible" people than by "impossible" people. |
| 02:53:09 | <solidsnack> | Indeed. |
| 02:53:46 | <Berengal> | solidsnack: Without the "possible" people you wouldn't have requirements at all |
| 02:54:09 | <conal> | solidsnack: proof of possibility/existence merely requires an example. proofs of impossibility/non-existence are deeper. |
| 02:54:11 | <solidsnack> | Probably shouldn't pay attention to either kind of person. |
| 02:54:18 | <copumpkin> | who's excited for GHC plugins? |
| 02:54:20 | <copumpkin> | ACTION raises hand |
| 02:54:27 | <conal> | ACTION too |
| 02:54:45 | <copumpkin> | there are so many things I want to fool around with in GHC's compilation pipeline |
| 02:55:00 | <sclv_> | ACTION wants units! |
| 02:55:08 | <Peaker> | conal: added a comment there |
| 02:55:15 | <copumpkin> | :t undefined :: 1 |
| 02:55:17 | <lambdabot> | Unit |
| 02:55:21 | <copumpkin> | there, a unit! |
| 02:56:01 | <solidsnack> | I look forward to a system better than IO. |
| 02:56:25 | <copumpkin> | I want IO à la carte as described in that paper (with inferrable type) |
| 02:56:43 | <solidsnack> | That's not the same thing, though. |
| 02:56:47 | <copumpkin> | still |
| 02:56:55 | <conal> | Peaker: thx for the comment. |
| 02:57:01 | <copumpkin> | it's better than the current situtation :) |
| 02:57:04 | <copumpkin> | -t |
| 02:57:14 | <boblo2> | what do you call the | (so i can look it up without typing | ) |
| 02:57:27 | <copumpkin> | boblo2: it does different things in different places |
| 02:57:33 | <copumpkin> | guard |
| 02:57:39 | <Peaker> | conal: have any idea if Lisp macros can define call-cc? |
| 02:57:52 | <boblo2> | | (length (check (x ++ "p")) == 49 = False = mzero |
| 02:57:52 | <boblo2> | | otherwise = result [papply code2 x] |
| 02:57:55 | <conal> | Peaker: sure they can. |
| 02:57:57 | <p_l> | Peaker: As for "impossibility", what about condition where you might arrive at bugfree and secure code, but can rest assured that universe will find a way to screw you over? |
| 02:58:02 | <copumpkin> | boblo2: that's a guard |
| 02:58:07 | <conal> | Peaker: i've messed *quite* a lot with lisp macros. |
| 02:58:08 | <p_l> | Peaker: Well, they are turing complete... |
| 02:58:10 | <boblo2> | guard thanks |
| 02:58:28 | <conal> | Peaker: you can write a whole syntactic CPS transformer as a lisp macro. |
| 02:58:43 | <Peaker> | conal, p_l: Assuming the call-cc itself is a macro, but it doesn't get to transform the code called by the call-cc? |
| 02:59:00 | <conal> | my first functional 3D & functional animation system was implemented as an optimizing compiler packaged as a lisp macro. |
| 02:59:09 | <Peaker> | I am no expert on the exact details of how Lisp macros typically work |
| 02:59:28 | <p_l> | Peaker: Why shouldn't it get possibility to transform that code? |
| 02:59:37 | <conal> | Peaker: lisp macros are functions from syntax to syntax. |
| 02:59:39 | <phenrique> | Algum brasileiro? |
| 02:59:57 | <Peaker> | conal, p_l: btw: Turing completeness of Lisp macros does not necessarily mean that they can perform all types of desirable transformations on code (Halting problem) |
| 03:00:09 | <Peaker> | conal: Yeah, but what subset of the program do they get to transform? |
| 03:00:32 | <phenrique> | Hi all, Any tutorial for GHC in Eclipse? |
| 03:00:34 | <conal> | Peaker: only the part they're explicitly wrapped around. |
| 03:00:39 | <Peaker> | conal: I guess if you define the functions call-cc calls with some defun-callable-by-call-cc instead of defun, then you get syntax transform of the called function as well? |
| 03:01:01 | <conal> | Peaker: right. you make your own defun replacement and use it. |
| 03:01:04 | <p_l> | Peaker: If you want, you can make macro that emits code that will transform code that you call |
| 03:01:16 | <conal> | that's what i did. i *loved* lisp macro hacking. |
| 03:01:29 | <conal> | it's compiler writing with no syntactic hassle at all. |
| 03:01:36 | <Peaker> | conal: ah - that's almost as good as ContT, but ContT allows keeping your existing code.. |
| 03:01:36 | <Berengal> | macro hacking does have a certain allure to it... |
| 03:01:53 | <p_l> | I'm pretty sure that with cl-walker, one can modify whatever the hell one wants... |
| 03:01:56 | <Peaker> | conal: (no need to change all existing defuns or such) |
| 03:02:02 | <phenrique> | Hi all, Any tutorial for GHC in Eclipse? |
| 03:02:06 | <conal> | Peaker: yeah. macros allow for non-semantic transformation. haskell doesn't. |
| 03:02:22 | <conal> | that's the beauty of each. |
| 03:02:46 | <p_l> | Berengal: there are reasons, though, why CL community (outside of the cll trolls) has a certain dislike of macros... |
| 03:02:52 | <conal> | lisp is awesome for what it permits; haskell for what it forbids. |
| 03:03:05 | <Peaker> | conal: In a non-textual language, you might want to have ASG transformers, which are probably much easier to work with than textual AST transformers |
| 03:03:41 | <p_l> | Peaker: Well, Lisp is not textual (unless you count having identifiers in form of strings as textual...) |
| 03:03:47 | <conal> | Peaker: could be, though AST transformers are super easy in lisp |
| 03:03:54 | <Berengal> | To be fair, lisp doesn't have an AST |
| 03:04:03 | <conal> | yeah. |
| 03:04:10 | <Peaker> | p_l: it is limited to text, e.g: Cannot have references to other positions in the tree, so need namespaces and text atoms, instead |
| 03:04:17 | <Berengal> | There's syntax, and it forms a tree, but it's in no way abstract |
| 03:04:29 | <conal> | Berengal: agreed |
| 03:04:41 | <solidsnack> | conal: So perhaps I have misunderstood you all this time. |
| 03:04:50 | <conal> | ACTION is getting misty eyed about his old lisp days |
| 03:04:50 | <Berengal> | That's one part of what makes lisp so fun, and macros so easy |
| 03:05:16 | <solidsnack> | What you mean by all this is you think there might be something better than IO? |
| 03:05:26 | <solidsnack> | That you're not sure but you have a strong hunch? |
| 03:05:27 | <Peaker> | p_l: all code is truly a graph, not a tree.. Since encoding trees in text is easy, and encoding graphs is hard, most programming languages devise the concept of text namespaces and text names to form links in the tree, so the ASG (graph) is encoded in two steps, ASG+namespaces+names -> AST+syntax -> Text |
| 03:05:37 | <p_l> | Peaker: Well, you can have those references... let's say that what you write is "textual represantation that is somewhat limited compared to language" |
| 03:05:41 | <solidsnack> | In that case, it's much hard to take offense. |
| 03:05:48 | <solidsnack> | s/hard/harder/ |
| 03:06:26 | <p_l> | Peaker: In fact, there were graph-based editors for Lisp, though I don't know how far did they went with the concept |
| 03:07:06 | <Peaker> | p_l: Are there dynamic Lisp namespaces, or are namespaces basically static like Haskell, except for allowing some late binding to allow redefinitions? |
| 03:07:08 | <dibblego> | what is the join . intersperse function? |
| 03:07:12 | <conal> | solidsnack: what part of my perceived message changes, and how it easier to hear without taking offense? |
| 03:07:23 | <Peaker> | @type intersperse |
| 03:07:24 | <lambdabot> | forall a. a -> [a] -> [a] |
| 03:07:37 | <Peaker> | dibblego: (join . intersperse) literally? |
| 03:07:44 | <Peaker> | @type (join . intersperse) |
| 03:07:46 | <lambdabot> | Couldn't match expected type `(->) [a]' against inferred type `[]' |
| 03:07:46 | <lambdabot> | Probable cause: `intersperse' is applied to too many arguments |
| 03:07:46 | <lambdabot> | In the second argument of `(.)', namely `intersperse' |
| 03:07:50 | <solidsnack> | Well, it's one thing to suggest there is definitely something better than IO and quite another to suggest that you're actively looking for it. |
| 03:07:54 | <dibblego> | (join .) . intersperse |
| 03:07:58 | <aavogt> | @type intercalate |
| 03:08:00 | <lambdabot> | forall a. [a] -> [[a]] -> [a] |
| 03:08:01 | <p_l> | Peaker: namespaces in CL are fully dynamic |
| 03:08:09 | <conal> | hurray for fmap.fmap |
| 03:08:10 | <dibblego> | ah cheers |
| 03:08:13 | <Peaker> | solidsnack: conal is probably at the forefront of researchers looking for that better IO alternative |
| 03:08:23 | <conal> | @type (fmap.fmap) join intersperse |
| 03:08:24 | <lambdabot> | forall a. [a] -> [[a]] -> [a] |
| 03:08:31 | <conal> | semantic editor combinators rule! |
| 03:08:32 | <RyanT5000> | is there an ubuntu repository that has GHC 6.10.3? |
| 03:08:34 | <Peaker> | p_l: so how can a static editor for a dynamic-namespace graph work? |
| 03:08:40 | <Peaker> | p_l: the graph links are only known at runtime? |
| 03:08:42 | <solidsnack> | Peaker: I think Lynch and Lamport are, as well. |
| 03:08:50 | <solidsnack> | You guys just don't know it, yet. |
| 03:08:57 | <p_l> | Peaker: Who said it was static? It ran *inside* a lisp system |
| 03:09:03 | <copumpkin> | RyanT5000: most linux distros have ghc packages made by people who don't use haskell enough to understand it right :P |
| 03:09:45 | <p_l> | and thus could edit data that made up a lisp program (which might be separated from the editor itself) directly, instead of feeding it a textual representation |
| 03:09:49 | <RyanT5000> | copumpkin: :( |
| 03:10:00 | <solidsnack> | (And neither do they, to be fair.) |
| 03:10:08 | <Peaker> | p_l: well, every time a function is invoked, only then the links are formed. In that case, how do you edit that function as a graph? Do you snapshot its last execution and assume the links will remain that way? |
| 03:10:15 | <copumpkin> | RyanT5000: download the binary distribution from the ghc site |
| 03:10:34 | <RyanT5000> | copumpkin: yeah, i was just hoping to "set it and forget it" :P |
| 03:10:48 | <p_l> | Peaker: Why links should be made only during execution? Shouldn't compiler find connections? |
| 03:11:06 | <Peaker> | Someone asked me to give a Haskell presentation in a local university. I've never lectured about Haskell before. Anyone have any ideas/slides/etc? |
| 03:11:24 | <Berengal> | Peaker: What kind of an audience are we talking? |
| 03:11:25 | <conal> | solidsnack: i ask because i'd love to learn to express my message in a way that captures it without people taking offense (getting defensive). |
| 03:11:35 | <RyanT5000> | Peaker: are you supposed to be teaching Haskell or talking about it? |
| 03:11:35 | <Peaker> | p_l: if the connections can be found at compile-time, it means the connections are static. But if all the names/namespaces are dynamic, then they are not known by the compiler |
| 03:11:36 | <wli> | Peaker: monad transformer stacks |
| 03:11:36 | <roconnor> | Peaker: I have |
| 03:12:18 | <Peaker> | Berengal, RyanT5000: Just talking about it, to a bunch of literate CS people who probably know SML, but not referential transparency, type-classes, etc |
| 03:12:18 | <roconnor> | Peaker: I was asked to give a one of the talks for the last class for a programming languages course. |
| 03:12:23 | <p_l> | Peaker: Have you ever worked with a modern Common Lisp system? |
| 03:12:30 | <Peaker> | p_l: nope |
| 03:12:36 | <roconnor> | Peaker: I have slides, but the are a little personal. You can have/look at them if you want. |
| 03:12:38 | <Berengal> | Peaker: Needs more fix, possibly even mfix :) |
| 03:12:43 | <solidsnack> | conal: Maybe describe everything that would lead one not to believe what you are saying. |
| 03:12:51 | <Peaker> | roconnor: sure, thanks |
| 03:12:54 | <solidsnack> | Instead of leaving it to the imagination. |
| 03:13:14 | <RyanT5000> | Peaker: well, you might want to mention Tim Sweeney's presentation on Haskell as a language for writing video games (Tim wrote Unreal) |
| 03:13:40 | <Peaker> | wli: is that an idea of something to talk about, to people who don't know laziness, type-classes, or how to combine side-effects with referential transparency? :) |
| 03:13:46 | <RyanT5000> | Peaker: that's if you have a "why haskell?" slide or something |
| 03:13:55 | <roconnor> | Peaker: http://r6.ca/Haskell/FPI.pdf and http://r6.ca/Haskell/FPI.odp -- Functional Programming Imperative |
| 03:14:00 | <p_l> | Peaker: I guessed so :D Basically imagine that introducing code into system in many of them includes compilation, and (compile a) requires that a is of apriopriate type... |
| 03:14:14 | <Peaker> | RyanT5000: thanks |
| 03:14:31 | <conal> | solidsnack: parsing trouble on my end. would you mind trying again. |
| 03:14:38 | <RyanT5000> | Peaker: www.st.cs.uni-saarland.de/edu/seminare/2005/advanced-fp/docs/sweeny.pdf |
| 03:14:51 | <RyanT5000> | Peaker: summary here: beust.com/weblog/archives/000375.html |
| 03:14:57 | <Peaker> | p_l: well, if the dynamism you talked about is pre-compilation dynamism, then its quite different to what I meant :) |
| 03:15:00 | <solidsnack> | conal: Sure. There are sound reasons to reject the notion of something better than IO. |
| 03:15:09 | <solidsnack> | Network communication is IO. |
| 03:15:18 | <solidsnack> | There's one. |
| 03:15:23 | <wli> | Peaker: More of an overarching direction. Since monad transformer stacks are what you ultimately do then the idea is to lead them toward that. |
| 03:15:27 | <p_l> | Peaker: In Lisp there's no write-compile-run cycle, so it's a little weird :P |
| 03:15:28 | <conal> | solidsnack: so are register and stack frame munging. |
| 03:15:35 | <solidsnack> | conal: Right. |
| 03:15:48 | <solidsnack> | With memory protection you have something, though. |
| 03:15:48 | <conal> | solidsnack: that's why i distinguish between a language's implementation and it's semantic model. |
| 03:16:04 | <solidsnack> | There's no such thing as network memory protection, though. |
| 03:16:07 | <Peaker> | wli: It would probably take more than a 2 hour talk or so to get so far |
| 03:16:16 | <conal> | solidsnack: no such thing ever? or just not yet? |
| 03:16:18 | <solidsnack> | At least, not unless you introduce it as an assumption. |
| 03:16:26 | <solidsnack> | conal: Must at least be mentioned. |
| 03:16:30 | <Peaker> | p_l: by the time a function exists, all of the links from that function are static, right? |
| 03:16:55 | <Berengal> | Peaker: The type system and how it enforces discipline without the usual S&M |
| 03:17:07 | <Peaker> | Berengal: any good examples of such discipline? |
| 03:17:26 | <Berengal> | Peaker: No casts |
| 03:17:33 | <conal> | solidsnack: sigh. i figure if people want to open their minds they will, and if not, then no amount of spoon-feeding will get them past their attachment to impossibility. |
| 03:17:51 | <solidsnack> | conal: Well, you don't understand teaching, then. |
| 03:17:52 | <p_l> | Peaker: You could say that link goes through resolver, and said resolver maps Keyword -> Location |
| 03:17:59 | <wli> | Peaker: Well, one may have to stop short. |
| 03:18:11 | <solidsnack> | conal: If you have a message, you have to deliver it; it's not something you can do at your liesure. |
| 03:18:11 | <conal> | solidsnack: sure i do. differently from you, though, it seems. :) |
| 03:18:21 | <solidsnack> | No one's going to crawl to you. |
| 03:18:30 | <conal> | solidsnack: those who have ears, let them hear. |
| 03:18:40 | <conal> | that's why i don't push. |
| 03:18:52 | <conal> | it's not worth the effort. others hear easily. |
| 03:18:55 | <conal> | that's enough for me. |
| 03:19:01 | <p_l> | Peaker: Believe me, incremental programming is probably the single weirdest thing for many newbies to Lisp |
| 03:19:12 | <solidsnack> | There's a difference between not pushing and not addressing reasonable concerns. |
| 03:19:28 | <conal> | solidsnack: when people ask me i tell them. |
| 03:19:42 | <Berengal> | p_l: I found it to be perhaps the most sexy thing about it |
| 03:19:52 | <Berengal> | Macros came close |
| 03:19:55 | <solidsnack> | You also tell them when they don't ask you. |
| 03:20:34 | <conal> | solidsnack: sure i can. i prefer letting people have their own curiosity or lack of it. not everyone is interested in everything. |
| 03:20:35 | <p_l> | And I won't claim that I've mastered macros, which requires understanding of all stages code goes through, so that you know how to properly call (eval-when) |
| 03:20:49 | <Peaker> | solidsnack: when you say "<solidsnack> Network communication is IO. There's one" -- you're basically saying: "I know you're saying there might be an abstraction more fitting than IO, but you are wrong, and I am right, as this *must* be IO. I won't consider any alternative." -- That's closed-minded |
| 03:21:08 | <solidsnack> | Peaker: No, I am not saying that. |
| 03:21:11 | <Peaker> | p_l: Well, you could say Python is pretty similar to CL? I've got extensive Python experience |
| 03:21:21 | <solidsnack> | Peaker: I am saying, that's something about which I have real misgivings. |
| 03:21:31 | <solidsnack> | Peaker: What you said is, basically, I don't have them. |
| 03:21:41 | <conal> | solidsnack: it's my approach. i get it that it's not yours, and it doesn't appeal to. so i wouldn't try to persuade you. |
| 03:21:43 | <solidsnack> | In other words, you have decided to blow me off. |
| 03:22:06 | <Peaker> | solidsnack: if you have doubts about *how* it can be abstracted in different ways than IO, then say those. What you said is not that you have doubts, but that it couldn't be done, it must be IO |
| 03:22:24 | <conal> | solidsnack: that's a rather loaded way to put it. you don't have to take it personally. |
| 03:22:24 | <p_l> | Peaker: Not really, though it helps that it's interactive interpreter is nearly fully capable... but it doesn't have the same "image" model, I guess |
| 03:22:32 | <p_l> | Peaker: And well, it uses duck-typing :P |
| 03:22:54 | <Peaker> | p_l: Doesn't Lisp also use duck-typing? At least with CLOS? |
| 03:23:05 | <conal> | Peaker: yeah. that's my take also. |
| 03:23:09 | <p_l> | Peaker: No. And double NO for CLOS |
| 03:23:40 | <solidsnack> | conal: Thank you for your advice. |
| 03:23:52 | <solidsnack> | conal: Would you care to say, what made you think I took it personally. |
| 03:23:53 | <p_l> | If what CLOS does was duck typing, then Haskell would be duck-typed too |
| 03:23:53 | <solidsnack> | ? |
| 03:23:57 | <Peaker> | p_l: well, I see untyped+duck-typing as nicer than untyped without duck typing, as at least you can get something like type-classes that way.. |
| 03:23:58 | <conal> | solidsnack: because i interpret you as certain (non-curious), i don't invest much time in trying to convey my ideas to you. you could call that "blowing you off". i call it investing wisely. |
| 03:24:14 | <solidsnack> | conal: Oh, I was responding to Peaker. |
| 03:24:26 | <conal> | solidsnack: sure. using the loaded phrase "blow me off". |
| 03:24:39 | <conal> | solidsnack: instead of a neutral/objective phrase. |
| 03:24:58 | <p_l> | Peaker: CL is implicitly typed, Scheme is less strict |
| 03:25:11 | <solidsnack> | conal: Well, you might have criticized my language instead of my state of mind. |
| 03:25:22 | <Peaker> | p_l: CL is typed at all? Do expressions/terms/variables have types? |
| 03:25:38 | <dons> | heya conal |
| 03:25:44 | <p_l> | Peaker: yes, they have. It's just that by default it will try inferring types, but you can define them |
| 03:25:51 | <conal> | dons: hi there. |
| 03:26:00 | <Peaker> | dons: hey -- we were having trouble figuring out where the quickcheck Utils module stuff disappeared to in Quickcheck 2, do you have any idea? |
| 03:26:06 | <dons> | Utils? |
| 03:26:07 | <p_l> | Peaker: That's how CLOs does its method matching, after all |
| 03:26:09 | <dons> | never used it. |
| 03:26:18 | <solidsnack> | conal: Should you be inclined to present a model of something, which I could consider in light of my interest in networked systems, I would of course be very interested. |
| 03:26:19 | <dons> | probably it got merged, or ditched |
| 03:26:31 | <Peaker> | dons: Test.QuickCheck.Utils existed in qc<2 and had some interesting functions (iirc e.g testAssocativeBy) |
| 03:26:40 | <lolo> | can someone tell me why the line: code2 :: Parser gives me "Illegal type in type expression" |
| 03:26:42 | <solidsnack> | conal: I fear that things are not that definite in your work, at present. |
| 03:27:08 | <dons> | solidsnack: (reading the scrollback) this might be a good time to let it go? |
| 03:27:09 | <mauke> | lolo: what is Parser? |
| 03:27:10 | <solidsnack> | conal: If you are measure interest by investment, however, then I have expressed interest. |
| 03:27:25 | <solidsnack> | dons: Sure, I will set the matter aside. |
| 03:27:26 | <lolo> | ParseLib i belive |
| 03:27:29 | <dons> | thanks. |
| 03:27:44 | <Peaker> | dons: who maintains quickcheck on hackage? |
| 03:27:52 | <dons> | btw, in case people didn't see it: US-based hackathon coming up: http://www.haskell.org/haskellwiki/Hac_φ |
| 03:27:57 | <p_l> | Peaker: (type-of "test") => (SIMPLE-ARRAY CHARACTER (4)) (car (type-of "test"))=>SIMPLE-ARRAY |
| 03:27:57 | <Peaker> | (#haskell persona, that is) |
| 03:27:58 | <dons> | i'm kinda excited to have a 'local' hackathon! |
| 03:28:11 | <dons> | Peaker: koen claessen |
| 03:28:17 | <Peaker> | p_l: that's as much typed as Python is? |
| 03:28:52 | <p_l> | Peaker: More - Python doesn't check types, it checks if there's an entry in objects dictionary with apriopriate name and calls it |
| 03:29:14 | <p_l> | Peaker: typing in Python afaik doesn't get outside primitive types |
| 03:29:32 | <lolo> | newtype Parser a = P (String -> [(a,String)]) |
| 03:29:50 | <mauke> | lolo: yeah, then Parser is not a type |
| 03:29:53 | <p_l> | if an object has, I think it was "__call__" entry, you can do obj() and it will call method defined as obj.__call__() |
| 03:30:17 | <Peaker> | p_l: can you come to #haskell-blah ? |
| 03:30:24 | <p_l> | sure |
| 03:30:26 | <Peaker> | dons: not in #haskell ? |
| 03:30:31 | <lolo> | how would i go about fixing this? |
| 03:30:35 | <lolo> | im very new |
| 03:30:54 | <dons> | Peaker: nope. sadly. |
| 03:31:04 | <mauke> | lolo: supply the missing type argument |
| 03:31:14 | <dons> | lolo: i think we'll need to see a paste (e.g. hpaste.org) |
| 03:31:38 | <SamB_XP> | Paste: http://moonpatio.com ? |
| 03:31:55 | <lolo> | http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5111#a5111 |
| 03:32:40 | <mauke> | lolo: Parser String |
| 03:33:04 | <mauke> | what's with the nested 'do's? |
| 03:33:09 | <lolo> | idk |
| 03:33:27 | <lolo> | its a game im trying to solve |
| 03:33:32 | <lolo> | you see any more mistakes? |
| 03:34:39 | <mauke> | not mistakes per se, but there are a few unused variables |
| 03:35:27 | <lolo> | so were it says codex :: parser it needs to be codex :: Parser String? |
| 03:36:53 | <mauke> | yes |
| 03:38:18 | <lolo> | how does line 58 look? |
| 03:38:38 | <lolo> | opps i mean 111 |
| 03:39:09 | <mauke> | hmm... wrong :-) |
| 03:39:19 | <mauke> | what is it supposed to do? |
| 03:39:29 | <lolo> | what needs to be done there |
| 03:39:46 | <mauke> | what is it supposed to do? |
| 03:39:50 | <lolo> | it takes in a huge list of strings and finds one that matches the critera in the code |
| 03:40:22 | <mauke> | ... |
| 03:41:24 | <jimmyjazz14> | lambdabot: help |
| 03:43:07 | <roconnor> | @help |
| 03:43:08 | <lambdabot> | help <command>. Ask for help for <command>. Try 'list' for all commands |
| 03:44:49 | <jimmyjazz14> | how does one convert a Int type to a Integer type? |
| 03:45:07 | <copumpkin> | fromIntegral |
| 03:45:09 | <Peaker> | @hoogle Int -> Integer |
| 03:45:09 | <lambdabot> | Prelude toEnum :: Enum a => Int -> a |
| 03:45:09 | <lambdabot> | Data.Bits bit :: Bits a => Int -> a |
| 03:45:09 | <lambdabot> | Prelude toInteger :: Integral a => a -> Integer |
| 03:45:37 | <jimmyjazz14> | cool! |
| 03:45:39 | <lolo> | mauke: still looking at it? |
| 03:45:51 | <mauke> | no, why? |
| 03:46:13 | <lolo> | just wondering thanks for the help. really want to get this lol |
| 03:46:16 | <copumpkin> | lolo: you basically told him "here's my input and it does what the code says it does" |
| 03:48:21 | <lolo> | It does not do what the code says it does (right now) becouse it wont run :( it is supposed to take in a list of strings and print success along with the correct string when it is done (i think) |
| 04:01:18 | <RyanT5000> | has anyone here thought about simulation datastructures in Haskell? |
| 04:01:33 | <RyanT5000> | i'm writing a video game, and i'm trying to figure out how to model the simulation state |
| 04:01:48 | <wli> | Priority queue? |
| 04:02:00 | <RyanT5000> | i don't quite follow |
| 04:02:05 | <roconnor> | State ? |
| 04:02:12 | <RyanT5000> | roconnor: the state of the world |
| 04:02:19 | <roconnor> | @hoogle State |
| 04:02:20 | <lambdabot> | module Control.Monad.State |
| 04:02:20 | <lambdabot> | Test.HUnit.Base data State |
| 04:02:20 | <lambdabot> | Test.HUnit.Base State :: Path -> Counts -> State |
| 04:02:28 | <RyanT5000> | i.e.: "the hero's over here, there's some stuff over there" |
| 04:02:35 | <roconnor> | Control.Monad.State |
| 04:02:43 | <roconnor> | oh |
| 04:02:51 | <roconnor> | you mean the state structure itself? |
| 04:02:55 | <RyanT5000> | roconnor: i understand the State monad, but i'm referring to the states it represents |
| 04:02:56 | <RyanT5000> | yeah |
| 04:03:02 | <roconnor> | ah |
| 04:03:10 | <roconnor> | I was thinking about it. |
| 04:03:11 | <hatds> | mutable references in IO or ST ;-) |
| 04:03:20 | <RyanT5000> | hatds: haha good joke |
| 04:03:23 | <RyanT5000> | :P |
| 04:03:32 | <roconnor> | the issues of identity were difficult. |
| 04:03:42 | <RyanT5000> | roconnor: yeah, that's what i'm thinking about right now |
| 04:04:01 | <RyanT5000> | right now i'm thinking that the general principle will be something like |
| 04:04:15 | <RyanT5000> | getObjectState :: ID a -> StateDomain -> Maybe a |
| 04:04:18 | <RyanT5000> | kinda like a Map |
| 04:04:20 | <roconnor> | I was trying to think if two ships in identical conditions at the same place were interchangable. |
| 04:04:33 | <roconnor> | and if so, how would I target one of them but not the other. |
| 04:04:52 | <hatds> | I'm suspect of any Id type thing |
| 04:04:54 | <RyanT5000> | roconnor: well, i think some explicit concept of identity is necessary |
| 04:05:03 | <RyanT5000> | hatds: me too :-/ |
| 04:05:15 | <roconnor> | me too |
| 04:05:15 | <RyanT5000> | but i can't figure out a different way to do it |
| 04:05:33 | <hatds> | Once I was about to write a function GenEq :: a -> b -> Bool (i.e. false if they have different types, otherwise true) before I learned the error of my ways |
| 04:05:42 | <hatds> | *er otherwise using (==) |
| 04:05:52 | <RyanT5000> | yeah |
| 04:06:26 | <hatds> | what do you mean by 'simulation state'? Is it the state of the game world or is this a meta simulation an AI is running or something? |
| 04:06:29 | <RyanT5000> | but here we're talking about an explicit identity system, which is not really any different from a Map whose RHS types are not all the same |
| 04:06:39 | <RyanT5000> | it's the state of the game simulation at a particular point in time |
| 04:06:49 | <RyanT5000> | the simulation being the logical world |
| 04:06:56 | <RyanT5000> | (as opposed to the rendering engine, or whatever) |
| 04:06:59 | <hatds> | I think mutable references are a proper solution |
| 04:07:06 | <hatds> | but you could use the State Monad |
| 04:07:10 | <mauke> | hatds: I've written and actually used that |
| 04:07:38 | <RyanT5000> | hatds: mutable references tie me to a mutable paradigm; it would probably be easier to write in C# if i were going to do that |
| 04:07:45 | <hatds> | mauke: did it work well? in my situation I realized it was suboptimal but I didn't discount it completely |
| 04:08:02 | <mauke> | yes, it worked fine |
| 04:08:10 | <mauke> | @let a === b = cast a == Just b |
| 04:08:11 | <lambdabot> | Defined. |
| 04:08:15 | <mauke> | :t (===) |
| 04:08:15 | <lambdabot> | forall a b. (Typeable a, Typeable b, Eq b) => a -> b -> Bool |
| 04:08:34 | <RyanT5000> | ah, Typeable :) |
| 04:08:44 | <copumpkin> | eww |
| 04:08:51 | <copumpkin> | Eq on b? |
| 04:08:59 | <mauke> | > "yes" == "yes" |
| 04:09:00 | <lambdabot> | True |
| 04:09:02 | <mauke> | > "yes" === "yes" |
| 04:09:03 | <lambdabot> | True |
| 04:09:09 | <mauke> | > "yes" === 'n' |
| 04:09:10 | <lambdabot> | False |
| 04:09:17 | <hatds> | RyanT5000: mutable references don't preclude pure code from doing the heavy lifting |
| 04:09:24 | <copumpkin> | > "yes" === id |
| 04:09:25 | <lambdabot> | No instance for (Eq (a -> a)) |
| 04:09:25 | <lambdabot> | arising from a use of `===' at <in... |
| 04:09:34 | <RyanT5000> | hatds: that's true, but they do preclude forking your simulation |
| 04:09:47 | <RyanT5000> | hatds: or saving its state to disk in any reasonable way |
| 04:09:55 | <Peaker> | > id === "yes" |
| 04:09:56 | <lambdabot> | Add a type signature |
| 04:10:02 | <Peaker> | > (id::a->a) === "yes" |
| 04:10:03 | <lambdabot> | Add a type signature |
| 04:10:07 | <hatds> | forking the simulation for what end? |
| 04:10:13 | <RyanT5000> | hatds: who knows |
| 04:10:16 | <roconnor> | RyanT5000: I was thinking the "identity" of something being a big list of states that the object goes through. |
| 04:10:24 | <roconnor> | RyanT5000: and tying one big knot |
| 04:10:29 | <roconnor> | for the game |
| 04:10:36 | <roconnor> | But I never got far enough to actually try it. |
| 04:10:46 | <RyanT5000> | roconnor: how do you mean? the identity would be the history of the object? |
| 04:11:02 | <erikc> | whats the topic, it sounds like i have an opinion on this!! |
| 04:11:10 | <hatds> | handling the state of a game |
| 04:11:21 | <erikc> | o i sure do |
| 04:11:22 | <roconnor> | RyanT5000: well, you want to disgrard the notion of past and present a funture |
| 04:11:30 | <roconnor> | RyanT5000: think space-time :) |
| 04:11:40 | <RyanT5000> | roconnor: hmmm |
| 04:11:48 | <roconnor> | the whole game is laid out at once, just existing |
| 04:12:18 | <RyanT5000> | roconnor: what about player input? |
| 04:12:32 | <RyanT5000> | suppose it's a real-time game |
| 04:12:37 | <roconnor> | RyanT5000: it is represented by a oracle |
| 04:12:39 | <roconnor> | oh |
| 04:12:41 | <roconnor> | real time |
| 04:12:42 | <roconnor> | sorry |
| 04:12:46 | <RyanT5000> | hm |
| 04:12:50 | <roconnor> | my game was turn based. |
| 04:13:03 | <RyanT5000> | well, you can treat any game as turn-based internally |
| 04:13:05 | <RyanT5000> | just with lots of turns |
| 04:13:06 | <RyanT5000> | so that's fine |
| 04:13:07 | <roconnor> | the same ideas might work but it would be harder. |
| 04:13:17 | <roconnor> | again, just ideas of mine |
| 04:13:53 | <RyanT5000> | it sounds interesting |
| 04:13:54 | <roconnor> | each "object" is represented by a list of states it is in (or maybe state, time pairs in your case) |
| 04:14:09 | <roconnor> | and the entire "game" is a set of these "objects" |
| 04:14:11 | <RyanT5000> | well, i was thinking that an object could be approximately a function from time to state |
| 04:14:17 | <roconnor> | and the game is parameterized by the input. |
| 04:14:34 | <RyanT5000> | (i guess that's more or less equivalent to a list of pairs) |
| 04:14:45 | <roconnor> | the value of various elements of the "object" depends the value of other "objects" |
| 04:14:58 | <RyanT5000> | right |
| 04:15:09 | <RyanT5000> | that sounds hard to debug, though |
| 04:15:10 | <roconnor> | so you just write this down |
| 04:15:25 | <hatds> | RyanT5000: I think the best solution is more imperative than what you are thinking |
| 04:15:47 | <RyanT5000> | very easy to write noncausal code or code with circular dependencies |
| 04:16:10 | <roconnor> | object1 = initalstate:(some_interaction allObject) |
| 04:16:16 | <roconnor> | er |
| 04:16:18 | <roconnor> | not quite that |
| 04:16:19 | <roconnor> | anyhow |
| 04:16:23 | <roconnor> | I haven't really tried it. |
| 04:16:36 | <roconnor> | RyanT5000: true what you say about circular dependencies |
| 04:16:40 | <RyanT5000> | yeah |
| 04:16:51 | <roconnor> | like I said, I never tried it. Just an idea. |
| 04:16:54 | <roconnor> | could be crap |
| 04:17:08 | <RyanT5000> | well, it sounds reasonable to me |
| 04:17:11 | <roconnor> | and probably would be easier for a turn based game. |
| 04:17:39 | <RyanT5000> | hatds: well, one way to make it more imperative might be to write a custom monad |
| 04:17:46 | <halberd> | in programming we can't symbolically differentiate a function--what we symbolically differentiate is a data structure which represents a function but is not one |
| 04:17:47 | <RyanT5000> | hatds: that's basically the other solution i've been thinking about |
| 04:18:09 | <hatds> | RyanT5000: ultimately I think any game idea is most easily expressed as State -> State |
| 04:18:22 | <RyanT5000> | hatds: that's true |
| 04:18:32 | <RyanT5000> | hatds: but i need to impose a little more structure than that |
| 04:18:32 | <halberd> | you can't write a higher order function that will exactly differentiate an expression like \x -> x*x + 2*x |
| 04:18:46 | <hatds> | RyanT5000: sure |
| 04:19:11 | <copumpkin> | :t deriv |
| 04:19:12 | <lambdabot> | forall a b. (Num b, Num a) => (Dif a -> Dif b) -> a -> b |
| 04:19:14 | <mmorrow_> | halberd: differentiation is a map of functions though |
| 04:19:21 | <RyanT5000> | hatds: in particular, the structure needs to allow independent object to be written and debugged relatively independently |
| 04:19:24 | <halberd> | because you don't have access to the syntax tree of that function, you can only evaluate it at a value |
| 04:19:25 | <copumpkin> | > deriv (\x -> x*x + 2*x) x |
| 04:19:26 | <lambdabot> | 1 * x + x * 1 + 2 * 1 |
| 04:19:38 | <RyanT5000> | because i need to hire people and let them work independently |
| 04:19:48 | <mmorrow_> | halberd: separate of whether you want to evaluate the resulting function on any vals |
| 04:20:21 | <copumpkin> | you can cheat |
| 04:20:26 | <mauke> | > deriv sin x |
| 04:20:27 | <lambdabot> | 1 * cos x |
| 04:20:41 | <copumpkin> | halberd: it ain't pretty, but it's symbolic ;) |
| 04:20:53 | <mmorrow_> | (differentiation being a linear map of vector spaces (where the "vectors" are functions)) |
| 04:21:39 | <halberd> | copumpkin: I'm unfamiliar with how deriv works, what's the hook? could you write the function deriv in haskell? |
| 04:21:46 | <copumpkin> | that is written in haskell :) |
| 04:21:53 | <copumpkin> | without even looking at the AST |
| 04:22:02 | <hatds> | RyanT5000: disconnected myself for a sec there :) |
| 04:22:08 | <RyanT5000> | hatds: np |
| 04:22:13 | <halberd> | could I see the source for deriv then? |
| 04:22:15 | <roconnor> | > reduce $ deriv sin x |
| 04:22:16 | <lambdabot> | 1 * cos x |
| 04:22:17 | <copumpkin> | > scanl1 (+) [1..5] :: [Expr] |
| 04:22:18 | <lambdabot> | [1,1 + 2,1 + 2 + 3,1 + 2 + 3 + 4,1 + 2 + 3 + 4 + 5] |
| 04:22:26 | <copumpkin> | @hackage numbers |
| 04:22:27 | <lambdabot> | http://hackage.haskell.org/cgi-bin/hackage-scripts/package/numbers |
| 04:22:32 | <RyanT5000> | all i said was that i need to ensure that several people on a team can work within this object system without getting in each others' ways |
| 04:22:52 | <copumpkin> | http://hackage.haskell.org/packages/archive/numbers/2009.5.20.1/doc/html/src/Data-Number-Dif.html#deriv |
| 04:23:02 | <mmorrow> | halberd: it uses typeclass trickery, but the basic concept could be implemented separately (w/ a parser for function reps given as String) |
| 04:23:06 | <erikc> | is it singlethreaded? |
| 04:23:22 | <mauke> | http://okmij.org/ftp/Computation/Generative.html#diff-typeclass |
| 04:24:13 | <halberd> | but that's my point mmorrow that you have to access the representation, which is usualy not available... unles you use typeclass trickery |
| 04:24:19 | <hatds> | RyanT5000: I think all that is needed is an abstraction type |
| 04:24:25 | <halberd> | I'm reading how that works |
| 04:24:32 | <RyanT5000> | hatds: what do you mean by that? |
| 04:24:43 | <copumpkin> | halberd: it's not using anything untypical though :) |
| 04:24:54 | <halberd> | differentiating the _function_ \x -> x*x + 2*x is not the same as differentiating the _string_ "\x -> x*x + 2*x" |
| 04:25:04 | <halberd> | someone differentiating the string has an easier time |
| 04:25:19 | <hatds> | RyanT5000: each disparate object of type a has to provide a function a -> B. Where B is a type representing abstract game objects |
| 04:25:19 | <copumpkin> | yeah, and that package isn't touching the string |
| 04:25:27 | <halberd> | yes I know, it's doing something funny |
| 04:25:29 | <mmorrow> | halberd: well without being able to view what you're working with, it would have to be magic? |
| 04:25:46 | <mmorrow> | oh, i see how you mean |
| 04:25:50 | <hatds> | RyanT5000: the abstractions could be hierarchical as well |
| 04:26:20 | <mmorrow> | halberd: sure, given the haskell lang (as implem by ghc) you need typeclass hackery or a QuasiQuoter |
| 04:26:23 | <RyanT5000> | hatds: yeah, that's true |
| 04:26:35 | <RyanT5000> | hatds: but i think we need some way of figuring out how objects will interact |
| 04:26:37 | <mmorrow> | (if you don't want to just give the rep a a String that is) |
| 04:26:43 | <mmorrow> | *..as a .. |
| 04:26:51 | <roconnor> | RyanT5000: be sure to check out data.accessor |
| 04:27:01 | <hatds> | RyanT5000: that is specified by the type B |
| 04:27:33 | <erikc> | RyanT5000: the approach i've taken (in C and Lua) is single-threaded message passing concurrency between actors (win32-style), the game state is stored relationally, separately from the actor code |
| 04:27:59 | <RyanT5000> | erikc: what sorts of things were actors? |
| 04:28:08 | <RyanT5000> | roconnor: that looks really useful |
| 04:28:12 | <erikc> | after the messages for a frame have been dispatched, bulk updating (physics, collisions, etc) and rendering runs in parallel using the frozen data |
| 04:28:26 | <erikc> | those systems generate messages on lockless queues that feed back to the actors |
| 04:28:31 | <roconnor> | RyanT5000: it is so useful, it should be part of the Haskell Platform! :) |
| 04:28:40 | <RyanT5000> | roconnor: perhaps :) |
| 04:28:47 | <hatds> | I think the interaction between actors will be dependent on what game it is |
| 04:28:48 | <mmorrow> | but i think the particulars (wrt how it's accomplished in whatever language) of any given implem of symbolic differentiation is a separate topic than what you're really doing |
| 04:29:03 | <roconnor> | RyanT5000: the State Monad stuff is exactly what you want. |
| 04:29:05 | <hatds> | interaction in an RTS would be much different than a 2D sidescroller |
| 04:29:13 | <copumpkin> | @let loeb x = fmap ($ loeb x) x |
| 04:29:15 | <lambdabot> | Defined. |
| 04:29:17 | <copumpkin> | :t loeb |
| 04:29:19 | <lambdabot> | forall (f :: * -> *) b. (Functor f) => f (f b -> b) -> f b |
| 04:29:49 | <copumpkin> | > loeb [ (!!5), const 3, liftM2 (+) (!!0) (!!1), (*2) . (!!2), length, const 17] |
| 04:29:51 | <lambdabot> | [17,3,20,40,6,17] |
| 04:30:03 | <erikc> | RyanT5000: characters, enemies, props, cameras, grenades, rockets, dropped weapons |
| 04:30:12 | <RyanT5000> | erikc: ah, i see |
| 04:30:20 | <RyanT5000> | erikc: and what sorts of things were considered "state"? |
| 04:30:41 | <roconnor> | copumpkin: who was that comment for? |
| 04:30:45 | <erikc> | the rule of thumb is anything that we want scripters/designers to be able to create/manipulate is done in actor land |
| 04:30:48 | <copumpkin> | no one |
| 04:30:58 | <copumpkin> | myself/anyone? |
| 04:30:59 | <erikc> | once we have a feature figured out, we can always move it to the backend for perf |
| 04:31:03 | <RyanT5000> | erikc: ah, makes sense |
| 04:31:13 | <erikc> | the tough part is you want it dynamic early on to experiment with the gameplay |
| 04:31:15 | <RyanT5000> | erikc: did you do this in Haskell? |
| 04:31:18 | <mmorrow> | @pl fix (\k a -> fmap ($ k a) a) |
| 04:31:19 | <lambdabot> | fix (fmap . flip id =<<) |
| 04:31:27 | <mmorrow> | @type fix (fmap . flip id =<<) |
| 04:31:29 | <lambdabot> | forall (f :: * -> *) c. (Functor f) => f (f c -> c) -> f c |
| 04:31:35 | <copumpkin> | lol |
| 04:31:42 | <mmorrow> | weee |
| 04:31:45 | <copumpkin> | it definitely fels like a fixworthy function |
| 04:31:52 | <erikc> | implemented a poor man's erlang in C++ and then message pumps run in either C++ or Lua |
| 04:32:27 | <RyanT5000> | erikc: ah, right, i remember you telling me about that the other day |
| 04:33:07 | <RyanT5000> | erikc: so the way you prevented cyclic references was by means of a delay in message passing? |
| 04:33:22 | <erikc> | the actors are stateless, all the data is stored separately so it can be serialized/snapshotted |
| 04:33:39 | <erikc> | data hiding is a headache in games cause there's a tendency to come up with game rules that throw the data hiding out the window anyway |
| 04:33:54 | <erikc> | you want to be able to crawl all over the data |
| 04:33:59 | <RyanT5000> | erikc: hm, yeah |
| 04:34:08 | <erikc> | yea, frame delay for cyclic dependencies |
| 04:35:18 | <RyanT5000> | erikc: the data hiding thing is an interesting problem |
| 04:35:30 | <roconnor> | ACTION not sure I understand what erikc is describing, but it sounds cool. |
| 04:35:46 | <wy> | aloha |
| 04:35:51 | <erikc> | im also storing the data relationally so that data that is operated on together is packed together in memory |
| 04:36:04 | <erikc> | so i can dma it to the spus/gpu |
| 04:36:05 | <RyanT5000> | erikc: right, that seems good for performance |
| 04:36:36 | <wy> | Is there a reason the patterns in function definitions are not lazy? |
| 04:37:09 | <hatds> | essentially, the haskell core has two constructions: let (lazy) and case (strict) |
| 04:37:29 | <erikc> | game logic shouldnt affect performance much and is hopelessly intractable to parallelize (short of STM, but then you lose deterniminism) |
| 04:37:59 | <RyanT5000> | erikc: what about parallel Haskell? |
| 04:38:14 | <wy> | @users |
| 04:38:15 | <lambdabot> | Maximum users seen in #haskell: 658, currently: 581 (88.3%), active: 14 (2.4%) |
| 04:38:18 | <RyanT5000> | erikc: i guess spark creation costs could outstrip their value |
| 04:38:21 | <erikc> | you mean software transactional memory? |
| 04:38:30 | <hatds> | wy: function definitions are considered part of case expressions, you can use "~" to make a pattern lazy and irrefutable |
| 04:38:32 | <copumpkin> | or DPH? |
| 04:38:32 | <erikc> | or data parallelism? |
| 04:38:44 | <RyanT5000> | erikc: data parallelism |
| 04:39:00 | <erikc> | you definitely want data parallelism on the engine backend for the number crunching |
| 04:39:23 | <wy> | hatds: but why haven't them an implicit "~" prepended? |
| 04:39:25 | <RyanT5000> | http://www.macs.hw.ac.uk/~dsg/gph/ |
| 04:39:34 | <RyanT5000> | erikc: i'm talking about that; par and seq |
| 04:39:40 | <hatds> | are game algorithms outside of GPU stuff really suitable for data parallelism? |
| 04:39:53 | <RyanT5000> | i'm not sure; i haven't really looked into it |
| 04:40:13 | <erikc> | hatds: yup, physics, fluid simulation ai pathfinding |
| 04:40:29 | <copumpkin> | RyanT5000: that's all built into GHC now |
| 04:40:38 | <xmonader> | I'm facing this http://pastebin.com/d64840d0a while installing hs-twitter any help ? |
| 04:40:44 | <erikc> | generating potential fields for crowd pathfinding (though thats more wavefront parallelism) |
| 04:40:45 | <RyanT5000> | copumpkin: that's cool |
| 04:40:57 | <copumpkin> | RyanT5000: Control.Parallel(.Strategies) |
| 04:41:03 | <hatds> | wy: ~ also makes the pattern irrefutable, so f ~[x] = ... will result in a error if the input was actually [] *and* f tries to use the value x |
| 04:41:50 | <copumpkin> | just need to compile with the threaded RTS and run your program with +RTS -N x where x is the number of native threads you want to run (assuming you aren't running the latest development GHC, which can detect automatically) |
| 04:41:55 | <hatds> | wy: there was discussion in the creation of haskell whether to make patterns like tuples strict or lazy, since pattern matching on a tuple type cannot every fail in a well typed program (it can only succeed or diverge) |
| 04:42:04 | <RyanT5000> | erikc: so, if we're only concerned with usability and maintainability, how would it be to have public and private state for objects? |
| 04:42:51 | <RyanT5000> | erikc: if data needed to be looked through, it would need to be explicitly publicized |
| 04:42:55 | <erikc> | right |
| 04:43:23 | <erikc> | that works, you'll need a way to get at all the objects of particular classes |
| 04:43:42 | <RyanT5000> | erikc: we could potentially have index actors, which are basically actors who can access other actors' state |
| 04:43:55 | <RyanT5000> | erikc: the index dependency graph would have to be acyclic |
| 04:44:09 | <RyanT5000> | erikc: which would allow indices to be updated in the same frame as their constituents are changed |
| 04:44:28 | <halberd> | okay |
| 04:44:44 | <wy> | > let [x] = [] in x |
| 04:44:46 | <lambdabot> | * Exception: /tmp/8222069046277313235:71:36-43: Irrefutable pattern failed ... |
| 04:44:48 | <RyanT5000> | erikc: and, of course, actors would have access to *all* actors' states from the previous frame |
| 04:44:56 | <halberd> | so the deriv implementation is amazing, but it still doesn't exactly take the derivative of a function |
| 04:44:57 | <erikc> | right |
| 04:45:05 | <RyanT5000> | public state, that is |
| 04:45:10 | <halberd> | well, not of a function from real numbers to real numbers |
| 04:45:14 | <RyanT5000> | is a messaging system still needed? |
| 04:45:20 | <copumpkin> | halberd: why not? |
| 04:45:26 | <RyanT5000> | might be convenient, i suppose |
| 04:45:44 | <copumpkin> | halberd: http://conal.net/blog/posts/what-is-automatic-differentiation-and-why-does-it-work/ |
| 04:45:46 | <Ycros> | RyanT5000: I'm interested in what you come up with |
| 04:45:51 | <copumpkin> | http://en.wikipedia.org/wiki/Automatic_differentiation |
| 04:45:58 | <erikc> | i went with messaging cause a major goal is to add gameplay behaviors/entities at runtime |
| 04:46:02 | <halberd> | it takes the derivative of a function (Num a, Num b) => Dif a -> Dif b, not of a function (Num a, Num b) => a -> b |
| 04:46:10 | <RyanT5000> | Ycros: cool; i'm sure i'll be in this chatroom asking questions frequently |
| 04:46:26 | <halberd> | it only appears to do the latter, because of the operator overloading |
| 04:46:40 | <RyanT5000> | erikc: ah, that makes sense |
| 04:46:49 | <halberd> | so you get fooled into thinking the function is a -> b, when it's really Dif a -> Dif b |
| 04:46:54 | <copumpkin> | halberd: sure, but you can lift a function into Dif easily |
| 04:46:55 | <RyanT5000> | erikc: erlang is on the very-short-list of languages i need to learn |
| 04:47:13 | <mauke> | halberd: Dif a is a number |
| 04:47:19 | <copumpkin> | any function |
| 04:47:22 | <wy> | have you heard of a lecture topic "Lambda the Ultimate Calculus"? |
| 04:47:23 | <copumpkin> | involving nubmers :) |
| 04:47:35 | <copumpkin> | halberd: check out conal's blog post I linked to |
| 04:47:55 | <gwern> | > 1130 - 434 |
| 04:47:56 | <lambdabot> | 696 |
| 04:48:06 | <roconnor> | copumpkin: you can't lift any function in any meaningful way |
| 04:48:13 | <erikc> | RyanT5000: i view a game engine as the operating system and actors as the processes |
| 04:48:28 | <halberd> | copumpkin you could write a function converttoDif (Num a, Num b) => (a -> b) -> Dif a -> Dif b? |
| 04:48:38 | <RyanT5000> | erikc: yeah, that sounds about right |
| 04:48:40 | <copumpkin> | roconnor: well, any function of one numerical parameter to the same numerical type, I guess |
| 04:48:45 | <copumpkin> | or is that not true either? |
| 04:48:48 | <roconnor> | no |
| 04:49:02 | <roconnor> | you can only properly lift it by providing the derivative as well |
| 04:49:17 | <mauke> | halberd: that looks like id |
| 04:49:18 | <roconnor> | you can improperly lift it, and get a function with a meaningless deriviatve |
| 04:49:36 | <erikc> | if i had the freedom to use any language, i'd likely combine C for numerical kernels, Haskell for the engine and Factor for the gameplay level code |
| 04:49:50 | <roconnor> | halberd: no such function exists that does what you would want it to do. |
| 04:50:00 | <roconnor> | for converttoDif |
| 04:50:17 | <mauke> | oh yeah, wrong rank |
| 04:50:23 | <wy> | hatds: got it. So case is the place we do branching, so we actually need the values there |
| 04:50:26 | <copumpkin> | what kind of meaninglessness would the derivative be? |
| 04:50:34 | <RyanT5000> | erikc: how do you deal with things like collision detection, which create mutual restrictions on actors' behavior? |
| 04:50:37 | <hatds> | wy: yea |
| 04:50:40 | <roconnor> | copumpkin: always 0 or always 1, or really anything you want. |
| 04:50:48 | <copumpkin> | ah |
| 04:51:16 | <hatds> | collision detection has well studied algorithms |
| 04:51:16 | <copumpkin> | I guess I can definitely think of functions that wouldn't work |
| 04:51:19 | <halberd> | so even though converttoDif is a function that doesn't do anything clearly uncomputable, it can't be written in Haskell |
| 04:51:27 | <wy> | hatds: What's the other places where we need to force the values? |
| 04:51:43 | <hatds> | you just need to specify what you want to do when certain collisions are found |
| 04:51:45 | <copumpkin> | halberd: well, a -> b might not be continuous, for example |
| 04:51:52 | <roconnor> | halberd: um, what would you expect if you give it a function that is differentiable nowhere? |
| 04:52:05 | <roconnor> | halberd: does that count as uncomputable? |
| 04:52:34 | <halberd> | well you could probably extend Dif to deal with that case with Nan |
| 04:52:40 | <copumpkin> | it's not just NaN |
| 04:53:03 | <copumpkin> | just a curve with a kink in it, for example |
| 04:53:12 | <hatds> | wy: I'm not familiar with the fine differences between haskell core and regular haskell, but I think case expressions are where either all or most evaluation happens |
| 04:53:24 | <hatds> | maybe someone else knows for sure? |
| 04:53:30 | <mauke> | > deriv tan x |
| 04:53:32 | <lambdabot> | 1 * cos x * recip (cos x) + sin x * negate (1 * negate (sin x) * recip (cos... |
| 04:53:34 | <halberd> | the derivative could still be reported as NaN in that case, or Undefined if you prefer |
| 04:54:02 | <halberd> | that would be acceptable... but even with that you still couldn't write converttoDif |
| 04:54:04 | <roconnor> | ACTION tries to think of a case where the derivative is uncomputable. |
| 04:54:08 | <wy> | halberd: So I guess lazy patterns can be only used in definitions with only one pattern? |
| 04:54:09 | <roconnor> | I can probably make one up |
| 04:54:29 | <hatds> | not necessarily |
| 04:54:34 | <erikc> | RyanT5000: we use the rigid body step in the backend (we use havok) to solve collisions |
| 04:54:41 | <hatds> | the first pattern could be non-lazy |
| 04:54:56 | <hatds> | and to make things more complicated you can have a lazy pattern inside a non-lazy one |
| 04:55:13 | <Saizan> | but then the costructor will be forced (assuming we're not talking about nested patterns) |
| 04:55:22 | <halberd> | that's a point roconnor |
| 04:55:50 | <erikc> | RyanT5000: for things like actors making choices based on nearby items, (or say, oncoming objects), we have a hidden physics volume that detects intersections on the physics step and pipes messages back to the actors |
| 04:56:06 | <copumpkin> | http://pirate.shu.edu/projects/reals/cont/fp_weier.html |
| 04:56:10 | <RyanT5000> | erikc: ah, interesting |
| 04:56:25 | <RyanT5000> | erikc: so, i take it the physics system is more or less independent of the actors |
| 04:56:40 | <halberd> | nowhere differentiable != derivative is uncomputable... there are functions whose derivative is uncomputable though |
| 04:56:41 | <RyanT5000> | erikc: "of the actor system", i should say |
| 04:56:53 | <erikc> | yep, totally separate |
| 04:57:29 | <copumpkin> | halberd: well, functions that are computable but their derivatives aren't? |
| 04:58:04 | <halberd> | I don't know about that copumpkin but if a function is uncomputable yet differentiable, its derivative is uncomputable |
| 04:58:08 | <mgsloan> | the derivative of id is 1! |
| 04:58:18 | <RyanT5000> | erikc: that's good; have you ever worked with Unity? it's the only engine i've ever had the 'pleasure' of dealing with |
| 04:58:42 | <sjanssen> | hmm, hackathon in Philadelphia, eh? |
| 04:58:45 | <sjanssen> | ACTION google maps it |
| 04:59:08 | <erikc> | i've played with it, but not enough to comment, i bet it demos well to C-level executives though with the lovely gui :) |
| 04:59:24 | <erikc> | their use of Mono is smart |
| 04:59:34 | <hatds> | I'll be there, anyone else going? |
| 04:59:38 | <copumpkin> | I'd stay away, don't wanna catch that shit |
| 04:59:46 | <wy> | got it |
| 04:59:48 | <RyanT5000> | erikc: it demos *quite* well to C-levels |
| 05:00:00 | <halberd> | Properties of Functions are in general uncomputable |
| 05:00:13 | <wy> | still haven't figured out why strict flag can be only used on data constructors and not functions |
| 05:00:42 | <halberd> | i.e. there is no function which can take the source code listing for another function f, and determine whether or not f is a differentiable function with derivative f'(x)=2x |
| 05:01:00 | <copumpkin> | well yeah |
| 05:01:02 | <halberd> | as long as f is written in a Turing-complete language |
| 05:01:04 | <RyanT5000> | erikc: (one of) the problem(s) is, Unity conflates actors with physics objects |
| 05:01:08 | <sjanssen> | hatds: I'm considering it, depends on whether I want to travel halfway across the country |
| 05:01:11 | <hatds> | wy: perhaps you want bang patterns? http://www.haskell.org/ghc/docs/latest/html/users_guide/bang-patterns.html |
| 05:03:05 | <RyanT5000> | erikc: i really like this idea of conceiving of the game engine as an OS; i've put a lot of thought into OS design before, so it's a really useful analogy |
| 05:03:10 | <wy> | yes. |
| 05:03:45 | <wy> | It looks good and makes the things more uniform. Just hope we never need to write those bang flags |
| 05:03:49 | <RyanT5000> | erikc: what makes Factor a particularly good language for level scripting? |
| 05:04:10 | <wy> | Can the strict flags all be inserted by strictness analysis? |
| 05:04:20 | <twb> | Zow, someone's recommending Factor? |
| 05:04:50 | <gwern> | wy: well, someone wrote a GA program for inserting bangs and seeing whether performance was helped... |
| 05:05:35 | <erikc> | RyanT5000: runtime code replacement with jitting, incremental generational gc, image-based (save/restoring the gc heap, including closures), code-as-data/data-as-code dualism (your 'data files' can just be scripts, no need for data file parsers, use the language) |
| 05:06:07 | <twb> | I suppose it depends what your "primary" language is. |
| 05:06:16 | <hatds> | imho problems with strictness are problems with not understand your code, and it is always worth it to understand it clearly and have very few annotations than to use scatter shot annotations. |
| 05:06:19 | <RyanT5000> | erikc: ah, the image-based part is something i've been wanting for quite some time |
| 05:07:00 | <erikc> | RyanT5000: naughty dog games uses plt scheme for their scripting and state machine descriptions in uncharted, they had a good gdc presentation this year |
| 05:07:45 | <twb> | Naught Dog used an in-houes CL dialect called something like "GOAL" for a while... |
| 05:07:50 | <conal> | halberd: if you do check out & like my auto-diff posts, you might also like my upcoming icfp paper, "beautiful differentiation" |
| 05:08:12 | <erikc> | twb: yep, till sony bought them and 'asked' them to build more tech in c++ to share with other studios |
| 05:08:26 | <twb> | erikc: ah, is that how it went? |
| 05:08:45 | <twb> | erikc: I heard that the one guy who understood how the GOAL compiler worked, quit/died/something |
| 05:08:46 | <erikc> | yea, their data build pipeline is still largely lisp |
| 05:09:01 | <QtPlaty[HireMe]> | IndentParser doesn't seem to be compatable with the current parsec |
| 05:09:14 | <erikc> | he left as well, yes (Andy Gavin), he went and made a webapp and sold it so fox interactive for 100 mil :) |
| 05:09:16 | <halberd> | conal: I got the idea of how it works |
| 05:09:50 | <halberd> | I did check out your posts and also the source code for the module |
| 05:10:08 | <RyanT5000> | are there any other languages that let you serialize closures? |
| 05:10:20 | <wy> | gwern: does GA mean genetic algorithm? |
| 05:10:26 | <erikc> | image-based ones, the lisps, smalltalks |
| 05:10:36 | <RyanT5000> | i'm specifically interested in serializing closures *without* serializing an image of the entire program |
| 05:11:31 | <halberd> | python does it ryan |
| 05:12:06 | <hatds> | I don't think the question makes sense with first class closures |
| 05:12:08 | <RyanT5000> | halberd: interesting; how would i search for that? pickling? |
| 05:12:16 | <RyanT5000> | hatds: really? |
| 05:12:33 | <hatds> | you could always have an abstract function type that is serializable |
| 05:12:58 | <halberd> | oh wait |
| 05:13:02 | <halberd> | maybe Python doesn't do it |
| 05:13:04 | <erikc> | you'd have to investigate but if you can serialize the image, you can probably serialize a portion of the object graph, provided the roots, it's the same problem |
| 05:13:18 | <RyanT5000> | erikc: that's what i was thinking |
| 05:13:33 | <RyanT5000> | erikc: but i've yet to see a language that does it |
| 05:13:55 | <hatds> | probably because doing so isn't getting you very much benefit |
| 05:14:19 | <halberd> | yeah python only does it for named functions |
| 05:14:57 | <RyanT5000> | hatds: i think there might be some benefits |
| 05:15:13 | <RyanT5000> | hatds: at the very least, it means that you can make *every* value in your language serializable |
| 05:15:31 | <RyanT5000> | hatds: which means that the programmer no longer has to worry about whether a particular datastructure is serializable or not |
| 05:15:42 | <hatds> | RyanT5000: I doubt it would be suitable for game saves |
| 05:15:48 | <hatds> | RyanT5000: in a game you really want your serialization to be aware of what it is serializing |
| 05:15:59 | <RyanT5000> | hatds: why's that? versioning? |
| 05:16:27 | <wy> | maybe using a "deriving Serializable" ? |
| 05:16:30 | <hatds> | RyanT5000: I just mean for efficiency |
| 05:16:47 | <wy> | otherwise roll you own ... |
| 05:17:00 | <erikc> | RyanT5000: factor has the serialize vocabulary which works on everything except continuations |
| 05:17:02 | <RyanT5000> | wy: but that means all the datastructures you *use* also need to derive it |
| 05:17:31 | <RyanT5000> | hatds: yeah, although i think it could be made reasonably efficient |
| 05:17:33 | <wy> | RyanT5000: As long as they have the method serialize ? |
| 05:18:04 | <RyanT5000> | wy: right, but i want something lower-level than that, so that we can guarantee, upfront, that *everything* is serializable in the whole language |
| 05:18:12 | <Saizan> | ocaml can serialize continuations, iirc |
| 05:18:55 | <wy> | RyanT5000: You could ask the compiler to add Serializable to every data type |
| 05:19:09 | <RyanT5000> | wy: how? |
| 05:19:17 | <wy> | then override it only for efficiency purposes |
| 05:19:27 | <RyanT5000> | Saizan: very interesting |
| 05:19:46 | <wy> | RyanT5000: use metaprogramming probably |
| 05:19:57 | <Saizan> | wy: yes, the point is finding a compiler/runtime system that supports it for everything, like functions etc. |
| 05:19:59 | <erikc> | but yea, where i work we just manually serialize a subset of the relations (and since the relations are packed in memory, it is quite fast) |
| 05:20:36 | <RyanT5000> | erikc: yeah, that makes sense; i'm not at the stage of worrying about efficiency yet, though |
| 05:20:53 | <RyanT5000> | erikc: right now, i'm trying to figure out what my ideal semantic model would be |
| 05:20:57 | <wy> | Saizan: even for monadic ones ? |
| 05:20:58 | <erikc> | yea |
| 05:21:04 | <RyanT5000> | later, i'll make concessions for efficiency purposes |
| 05:21:16 | <Saizan> | wy: yeah |
| 05:21:27 | <RyanT5000> | (your approach does sound like a good one, though) |
| 05:21:43 | <Peaker> | Its weird that a game developer (Sweenie) is a theory guy. In my mind, they're all C++-heads :) |
| 05:21:46 | <Saizan> | wy: i guess some things like file handles are not serializable by principle, however |
| 05:21:54 | <wy> | Saizan: good idea |
| 05:22:25 | <wy> | Saizan: why aren't they? |
| 05:22:41 | <RyanT5000> | Peaker: well, i'm a theory guy becoming a game guy :) |
| 05:23:01 | <RyanT5000> | really, very few people bridge the theory/practice gap |
| 05:23:12 | <Peaker> | RyanT5000: he's not just an ordinary game guy, he supposedly wrote Unreal Tournament |
| 05:23:16 | <Saizan> | wy: an Handle is a bit like a pointer to a resource, and that resource might not be there or be in the same state when you deserialize |
| 05:23:21 | <erikc> | Peaker: christer ericcson (tech director on god of war) is also a big theory guy (he has a big 'please game devs learn some other languages' rant) http://realtimecollisiondetection.net/blog/ |
| 05:23:22 | <Jedai> | Didn't Clean allows to serialize closures ? |
| 05:23:24 | <twb> | academic vs. pragmatic? |
| 05:23:57 | <wy> | Saizan: we could force them to be in the same state |
| 05:24:06 | <IntergalacticOwl> | i'm a game kid becoming a theory kid |
| 05:24:57 | <wy> | Saizan: Once the object with the reference to the resource is sent, we must establish the channel over the network to synchronize the state |
| 05:25:08 | <Jedai> | wy: To do that we could need a copy of the file itself though... |
| 05:25:30 | <Jedai> | In case it changed or something |
| 05:25:47 | <Saizan> | wy: suppose the Handle comes from a Socket, how do you restore that if the other machine is down? |
| 05:25:47 | <wy> | Jedai: you just need to send the changes |
| 05:26:12 | <Saizan> | wy: or if it was a Socket initiated by the other machine (i.e. we're the server) |
| 05:26:43 | <Jedai> | wy: I'm not following you, how do you serialize changes that aren't there when you do the serializing ? |
| 05:27:01 | <wy> | Saizan: it can't be absolutely safe. Just let's pray for the best ;) |
| 05:27:30 | <Saizan> | yeah, just don't serialize Handle's :) |
| 05:27:35 | <erikc> | RyanT5000: the os analogy is good cause it separates the engineers (engine) from the creatives (game), the engine is a fairly known quantity, the game is going to change a lot, so theres a static/dynamic fight there |
| 05:27:37 | <Peaker> | erikc: cool |
| 05:27:43 | <Saizan> | at best you can serialize a FilePath, or an URI |
| 05:27:51 | <erikc> | a coworker of mine is a big propodent of gradual typing cause of that |
| 05:28:08 | <erikc> | as the game settles, you want to type it and push it down into higher perf land |
| 05:28:23 | <RyanT5000> | erikc: yeah, that makes a lot of sense |
| 05:28:28 | <Jedai> | Saizan: eventually with a position in the file and a signature for coherence purpose |
| 05:28:37 | <RyanT5000> | erikc: i don't think there's any language out there whose type system supports such a thing very well |
| 05:29:36 | <mmorrow> | Saizan: with pointers though, you can just map them to unique ids you create as you serialize (i'm assuming you are slurping the targets of all pointers here too) |
| 05:30:07 | <Saizan> | mmorrow: "pointer" was only an analogy |
| 05:30:09 | <mmorrow> | Saizan: with handles/scokets it's a whole different story, since you don't have access (probably) to the other side |
| 05:30:23 | <Saizan> | the problem is internal vs. external resource |
| 05:30:33 | <mmorrow> | that's a good way to say it |
| 05:32:42 | <xmonader> | I'm facing this http://pastebin.com/d64840d0a while installing hs-twitter any help ? |
| 05:34:11 | <wy> | The implementation of IO monad is a little weird, why are there two other things beside the value? GHC.Prim.State# and GHC.Prim.RealWorld? |
| 05:34:50 | <hatds> | I only care to understand it from the high level |
| 05:34:54 | <Peaker> | erikc: I disagree with his rant (and tone!) about patterns for example.. |
| 05:35:19 | <Peaker> | Some patterns are language-specific, some are almost as "eternal" as data structures/algorithms |
| 05:35:45 | <hatds> | the implementation as I understand it is basically a trick to project sequencing (in the sense of time and effects) into a data dependency |
| 05:36:08 | <Peaker> | (The "Strategy" pattern, of passing dictionaries/functions around to higher-order functions, where the passed functions represent what you want to do. This "pattern" is implemented by type-classes or function records in Haskell, but its the same pattern from the group of four (?) book |
| 05:36:30 | <Saizan> | xmonader: it probably depends on an older version of json without specifying it in the .cabal file |
| 05:37:00 | <Peaker> | what's gradual typing? |
| 05:37:14 | <erikc> | http://www.google.ca/url?sa=t&source=web&ct=res&cd=3&url=http%3A%2F%2Fwww.cs.colorado.edu%2F~siek%2Fpubs%2Fpubs%2F2006%2Fsiek06%3A_gradual.pdf&ei=QzgWSrWPAov8swPhtOHZCA&usg=AFQjCNF2DrY06Kdj0CMQYstzgygzkhrLSw&sig2=1ZZQN68JBkkL67oy3GVTBw |
| 05:37:19 | <erikc> | err |
| 05:37:22 | <mmorrow> | Jedai: yeah, clean's rts supports both serializing and reading back in polymorphically typed data |
| 05:37:31 | <erikc> | google 'gradual typing' there are some papers |
| 05:37:43 | <xmonader> | Saizan, sounds so ''build-depends: base, HTTP >= 4000.0.1, network, utf8-string, json, mime, old-locale, old-time, random |
| 05:38:22 | <wy> | Peaker: the only way to learn the patterns is to forget them :) |
| 05:38:35 | <Peaker> | wli: what do you mean? |
| 05:39:51 | <Jedai> | wy: It seems to me that GHC.Prim.State# is the primitive type constructor behind both IO and ST and GHC.Prim.RealWorld is the state type for IO |
| 05:40:35 | <Peaker> | I remember reading luqui's realization a few months ago -- "why not put partiality in a monad?" or something like that. I see now that it was in Sweene's slides that I saw so long ago (but did not have enough background to understand): "Why not go one step further and define partiality as an effect, ..." |
| 05:40:50 | <Jedai> | mmorrow: Thanks, I vaguely remembered that that was one of the real advantage of Clean over Haskell |
| 05:41:35 | <mmorrow> | Jedai: this can totally be done in haskell too.. it's one thing i want to do actually |
| 05:42:12 | <Saizan> | xmonader: use --constraint="json < 4", as it appears JSRational constructor changed in version 0.4.1 |
| 05:42:14 | <mmorrow> | you just just serialize closures directly like vacuum does, and then going the other way you'd need to write some rts C code |
| 05:42:19 | <Jedai> | mmorrow: Well theoretically, but Clean had a clean implementation integrated with its compuler |
| 05:42:28 | <mmorrow> | Jedai: http://moonpatio.com/repos/MISC/Clean/dynamic/dynamics/ |
| 05:42:30 | <PetRat> | What's the precedence of the double colon operator relative to $? at gchi typing :info :: gives an error? |
| 05:42:32 | <wy> | Peaker: were you reading some blog post? I don't know the context |
| 05:42:41 | <mmorrow> | Jedai: a lot of it is hand-coded asm |
| 05:42:42 | <PetRat> | :info :: |
| 05:43:08 | <hatds> | :: extends as far as possible iirc |
| 05:43:14 | <xmonader> | Saizan, many thanks i'll try it and get back to you |
| 05:43:28 | <Peaker> | wy: reading www.st.cs.uni-saarland.de/edu/seminare/2005/advanced-fp/docs/sweeny.pdf |
| 05:43:30 | <PetRat> | hatds: I don't understand what you mean |
| 05:43:39 | <Peaker> | hmm -- what's "lenient evaluation"? |
| 05:44:05 | <PetRat> | let me give an example: can I write "fromSql $ head $ head $ rows :: Int" ? |
| 05:44:11 | <Peaker> | erikc: can you describe gradual typing in a sentence? :) |
| 05:44:21 | <mmorrow> | Jedai: err, well i should have said that clean has _dynamics_ where you can put in and recover polymorphically-typed data |
| 05:44:25 | <Peaker> | PetRat: fromSql . head . head $ rows |
| 05:44:37 | <mmorrow> | you can probably serialize that too i'd guess |
| 05:44:47 | <Jedai> | Peaker: wasn't it some kind of speculative evaluation ? |
| 05:44:49 | <Cale> | PetRat: More toward your question, :: gobbles up as much as it can |
| 05:45:04 | <wy> | :info ($!) |
| 05:45:20 | <Cale> | That is, it scopes over the largest possible preceding subexpression. |
| 05:45:37 | <mmorrow> | (i'm more interested in serializing/reifying which retains all sharing, and so can serialize "infinite data" which is just some tiny graph that's only infinite if you consider it to be a tree) |
| 05:45:45 | <PetRat> | Cale: I think my expression was fine then, in terms of getting the right type out of fromSql. |
| 05:45:49 | <Cale> | yeah |
| 05:45:58 | <mmorrow> | , vaccum (fix (\x -> 0 : x)) {- my favorite -} |
| 05:46:00 | <lunabot> | luna: Not in scope: `vaccum' |
| 05:46:05 | <mmorrow> | , vacuum (fix (\x -> 0 : x)) {- my favorite -} |
| 05:46:08 | <lunabot> | [(0,[1,0]),(1,[])] |
| 05:46:18 | <Cale> | (though take Peaker's advice and try to get in the habit of using composition instead of repeated $) |
| 05:46:59 | <mmorrow> | and also, the serialization function can be (forall a. a -> ByteString) |
| 05:47:06 | <Cale> | The current associativity of $ is a bit of a mistake, and there are benefits to using composition instead. |
| 05:47:28 | <xmonader> | Saizan, http://pastebin.com/d4c53c5b4 |
| 05:47:38 | <PetRat> | Cale: OK |
| 05:47:38 | <Jedai> | mmorrow: That would be awesome to have anyway :) |
| 05:47:43 | <mmorrow> | (deserialization though is more limited since haskell doesn't have dependent types or a polymorphic typeable) |
| 05:47:47 | <erikc> | Peaker: basically, dynamic by default, and then allow type annotations / infer types as much as possible |
| 05:47:51 | <mmorrow> | Jedai: totally! |
| 05:48:20 | <erikc> | and where there are type annotations, provide type checking |
| 05:48:31 | <Peaker> | erikc: ah, limited type inference, I hate that.. |
| 05:49:09 | <Cale> | PetRat: Since f . (g . h) = (f . g) . h, expressions written with composition are easier to take apart and transform, and it tends to help one's thought process to be thinking of functions being stuck end to end. |
| 05:49:26 | <RyanT5000> | Peaker: just think of it like a type system that doesn't finish its work until runtime :P |
| 05:49:48 | <wy> | Peaker: haven't found anything to do with strategy pattern |
| 05:49:48 | <erikc> | it is ideal for situations where you dont know what you want to make, but when you figure it out, you clamp it down by adding type annotations (and get the associated safety/perf benefits) |
| 05:49:49 | <Jedai> | RyanT5000: You mean a dynamic type system... |
| 05:49:51 | <Cale> | (whereas if you have something like f $ g $ h $ x, the substrings like g $ h probably aren't even well-typed) |
| 05:49:52 | <Peaker> | RyanT5000: its still untyped, afaic |
| 05:50:03 | <RyanT5000> | Jedai: no, it *starts* its work at compile-time :P |
| 05:50:06 | <Peaker> | wy: no no, that was from the rant of this other dude |
| 05:50:25 | <erikc> | which describes game dev in a nutshell, since 'a fun program' is a nebulous requirement :) |
| 05:50:32 | <Peaker> | erikc: its weird to me -- I think its easier to specify the type of what you want to make, than the implementaiton |
| 05:50:54 | <RyanT5000> | well, there's a lot of info in the type system that isn't necessary |
| 05:51:24 | <wy> | Peaker: I'd like to see the rant |
| 05:51:25 | <RyanT5000> | for example, i frequently know that i want a datatype that associates keys with values |
| 05:51:45 | <RyanT5000> | but i don't always know whether i want it to be a function, a list of pairs, a Map, or an array |
| 05:52:02 | <Peaker> | wy: http://realtimecollisiondetection.net/blog/?p=81 |
| 05:52:03 | <Jedai> | RyanT5000: You could use the typeclasses from Edison |
| 05:52:28 | <RyanT5000> | Jedai: i have; they're pretty good, but they're nowhere near pervasive enough to deal with this problem in general |
| 05:52:37 | <RyanT5000> | (nor comprehensive enough) |
| 05:52:41 | <Peaker> | RyanT5000: there are semantic differences between those types |
| 05:52:51 | <Peaker> | RyanT5000: if you don't know what semantics you want - how can you implement anything? |
| 05:53:15 | <erikc> | Peaker: and thats why game dev is such a mess :) |
| 05:53:17 | <RyanT5000> | Peaker: that's true; i chose a bad example |
| 05:53:49 | <RyanT5000> | Peaker: suppose i add a default response to the last 3 of them |
| 05:54:07 | <RyanT5000> | Peaker: (used for not-present or out-of-bounds inputs) |
| 05:54:32 | <Peaker> | ACTION has to go |
| 05:54:42 | <mmorrow> | mempty? |
| 05:54:47 | <RyanT5000> | alright; good luck on the presentation |
| 05:55:13 | <RyanT5000> | mmorrow: sure, the point is that i have 4 ways of representing the same idea |
| 05:55:26 | <RyanT5000> | mmorrow: and i have no way to tell the compiler that i don't care which of those it is |
| 05:55:46 | <RyanT5000> | mmorrow: (except in certain very limited circumstances like if i use Edison pervasively) |
| 05:55:51 | <mmorrow> | but why not rep write code that work with the general concept, that you can use regardless of implem? |
| 05:56:03 | <mmorrow> | (oh, i'm not talking about edison in particular) |
| 05:56:27 | <wy> | Peaker: I used to have list of all the design patterns and their corresponding terms in functional programming |
| 05:56:32 | <RyanT5000> | is there a general way to do that? |
| 05:56:36 | <mmorrow> | yeah, not having control over typeclass instance selection is a pain when there's more than one instance possible |
| 05:57:20 | <RyanT5000> | mmorrow: exactly |
| 05:57:26 | <mmorrow> | RyanT5000: if only we had ML structures/signatures/(recursive)-functors/etc |
| 05:58:12 | <RyanT5000> | mmorrow: especially if you have, e.g.: f :: C c => a -> c ; g :: C c => c -> b ; g (f x) |
| 05:58:28 | <wy> | My conclusion is that most "novel" design patterns arise from the lack of first-class functions. The others are old and used by people for decades. They just didn't have a name |
| 05:58:31 | <RyanT5000> | if i don't care at all which implementation of C gets used for c, i still have to pick one |
| 05:59:04 | <RyanT5000> | and since i'm virtually certain to never look at that code again, whatever i pick randomly the first time through will be there forever |
| 05:59:31 | <PetRat> | Suppose I'm mapping a function with > 2 arguments, and the argument to be supplied by the list is somewhere in the middle. is there something more idiomatic than map (\a -> g arg1 arg2 a arg3 arg4) xs ? |
| 05:59:32 | <mmorrow> | RyanT5000: yeah, i think that typeclasses are good for what they're good for, but suck as a substitute for signatures/functors |
| 05:59:52 | <RyanT5000> | mmorrow: yeah; what i really want in that particular example is for the compiler to just pick an instance for me |
| 05:59:52 | <wy> | Just like the way JUnit works. People have been using this kind of thing for decades, but some guy gave it a name and made fame from it... |
| 06:00:13 | <RyanT5000> | mmorrow: although that requires that classes can completely specify program behavior... |
| 06:00:25 | <RyanT5000> | (or at least specify it arbitrarily well) |
| 06:00:40 | <mmorrow> | RyanT5000: the signatures only specify abstract units |
| 06:00:46 | <mmorrow> | Map |
| 06:00:50 | <mmorrow> | Queue |
| 06:00:55 | <mmorrow> | Scheduler |
| 06:00:57 | <mmorrow> | etc |
| 06:01:00 | <RyanT5000> | yeah |
| 06:01:24 | <RyanT5000> | ideally, i'd like the programmer to usually not need to get below that level of abstraction |
| 06:03:11 | <PetRat> | bump Suppose I'm mapping a function with > 2 arguments, and the argument to be supplied by the list is somewhere in the middle. is there something more idiomatic than map (\a -> g arg1 arg2 a arg3 arg4) xs ? |
| 06:05:06 | <dibblego> | is the Test.QuickCheck module installed by default with most GHC distributions? |
| 06:05:46 | <mib_t968yv> | > pl (\l i k -> take (k-i+1) $ drop (i-1) l) |
| 06:05:48 | <lambdabot> | Not in scope: `pl' |
| 06:06:07 | <mib_t968yv> | ...d'oh >< |
| 06:06:45 | <mmorrow> | RyanT5000: totally, you could code whole programs as a bunch of (ML) functors, then stick in any structures with correct signatures later |
| 06:08:13 | <erikc> | PetRat: none i know of |
| 06:08:26 | <RyanT5000> | mmorrow: or better yet, have the compiler stick them in for you, and you only change them if you need to to improve efficiency |
| 06:08:44 | <RyanT5000> | mmorrow: i'm going to have to learn ML now :P |
| 06:08:52 | <mmorrow> | RyanT5000: well sure, assuming that all the structures you need are in the stdlib |
| 06:08:52 | <PetRat> | erikc: thanks |
| 06:09:05 | <RyanT5000> | (i've used it in some classes, but not seriously) |
| 06:09:20 | <mmorrow> | (structure in the ML sense) |
| 06:09:40 | <mib_t968yv> | @pl \l i k -> take (k-i+1) $ drop (i-1) l |
| 06:09:40 | <lambdabot> | ap (flip . (take .) . flip flip 1 . ((+) .) . subtract) . flip (drop . subtract 1) |
| 06:09:45 | <RyanT5000> | mmorrow: it seems like that could be made extensible relatively easily |
| 06:10:03 | <RyanT5000> | mmorrow: obviously you'd have to tell the compiler which structures you trust to implement a given interface |
| 06:10:20 | <RyanT5000> | (especially if a library has one good structure and another that's not so good) |
| 06:10:38 | <sjanssen> | this seems related to defaulting |
| 06:10:42 | <mmorrow> | RyanT5000: exactly, you then write code that hand-picks what structures you want |
| 06:11:07 | <mmorrow> | (and you can swap a choice out for another easily) |
| 06:11:08 | <RyanT5000> | yup |
| 06:11:31 | <RyanT5000> | i think that would eliminate a *huge* class of decisions from the average programmer's day |
| 06:11:56 | <mmorrow> | me too |
| 06:12:12 | <sjanssen> | perhaps some sort of extension to defaulting is in order: default(Integral Integer, MapClass Data.Map.Map) |
| 06:12:51 | <RyanT5000> | sjanssen: i think it would take some serious changes to haskell's class system to make such a thing workable |
| 06:13:03 | <RyanT5000> | sjanssen: at the very least, transitive mandatory inclusion of instances would need to be dealt with |
| 06:13:10 | <sjanssen> | RyanT5000: I'm not sure, defaulting already exists |
| 06:13:15 | <sjanssen> | RyanT5000: that issues seems orthogonal |
| 06:14:00 | <RyanT5000> | sjanssen: well, i suppose if the default were specified with the class, that would be fine enough |
| 06:14:20 | <sjanssen> | RyanT5000: hmm, is that really what we want to do? |
| 06:14:26 | <sjanssen> | seems like defaults should be module local |
| 06:14:48 | <RyanT5000> | sjanssen: they need to be importable, otherwise we're adding a ton of module overhead |
| 06:15:00 | <RyanT5000> | sjanssen: they could be imported from non-modules though, i guess |
| 06:15:07 | <sjanssen> | RyanT5000: yeah, importable would be good |
| 06:15:09 | <RyanT5000> | C preprocessor? :P |
| 06:15:34 | <erikc> | RyanT5000: logic programming with a good metalanguage for the logical->physical mapping would fit well too |
| 06:16:02 | <RyanT5000> | erikc: i'm confused as to what you're talking about |
| 06:16:04 | <erikc> | there is some interesting work here http://www.cs.cornell.edu/bigreddata/games/ in that area, applying relational algebra to game scripting |
| 06:16:17 | <RyanT5000> | ah, i see |
| 06:16:32 | <RyanT5000> | "Computer games are becoming the next frontier for social interaction between humans." |
| 06:16:41 | <RyanT5000> | a scary - and dubious - statement :P |
| 06:16:42 | <erikc> | grant-speak |
| 06:16:48 | <erikc> | :P |
| 06:17:35 | <twb> | I bet dijkstra would have something scathing to say about that |
| 06:17:40 | <RyanT5000> | yeah... i got plenty of that kind of crap in law school :P |
| 06:18:13 | <RyanT5000> | (lawyers don't reserve that kind of language for court - or even for work :P) |
| 06:18:36 | <erikc> | twb: about the 'computer games are ...' ? |
| 06:18:59 | <RyanT5000> | "computer gaming is" might be a more grammatical choice :P |
| 06:18:59 | <twb> | erikc: about grant-speak being about a reality that doesn't exactly match the real world. |
| 06:19:09 | <erikc> | ah hehe |
| 06:19:16 | <RyanT5000> | lol |
| 06:20:03 | <twb> | I recently read an essay (read: rant) of his about the lack of, as it were, plain speaking in papers. |
| 06:21:01 | <twb> | http://www.cs.utexas.edu/users/EWD/transcriptions/EWD04xx/EWD498.html This one? |
| 06:21:44 | <twb> | No, not that one. |
| 06:25:11 | <RyanT5000> | thanks for all the input on the game engine stuff, everyone |
| 06:25:23 | <RyanT5000> | i'm out |
| 06:33:04 | <twb> | Ah, here it is |
| 06:33:05 | <twb> | http://www.cs.utexas.edu/users/EWD/transcriptions/EWD13xx/EWD1300.html |
| 06:33:41 | <twb> | Not really as relevant as I remembered. |
| 06:33:43 | <dibblego> | when did GHC discontinue splitting into core and extra-libs (I'm going to give a presentation and I want all users to have Test.QuickCheck available) |
| 06:40:09 | <wy> | weilawei|away: hi |
| 06:47:36 | <xmonader> | Saizan, i tried your note but still not working |
| 06:49:56 | <mykelyk> | how can i call f :: Value -> [Value] -> IO (State Stack Value) |
| 06:50:38 | <mykelyk> | with a :: IO (State Stack Value) |
| 06:50:54 | <mykelyk> | and b :: IO (State Stack [Value]) ? |
| 06:52:11 | <mmorrow> | @type \f a b -> do sv <- a; svs <- b; return (do v <- sv; vs <- svs; f v vs) |
| 06:52:13 | <lambdabot> | forall (t :: * -> *) (t1 :: * -> *) t2 t3 t4. (Monad t, Monad t1) => (t2 -> t3 -> t1 t4) -> t (t1 t2) -> t (t1 t3) -> t (t1 t4) |
| 06:52:30 | <mmorrow> | although that seems like it's not what you want. where did you get those types from? |
| 06:52:59 | <mykelyk> | an intepreter |
| 06:53:36 | <mmorrow> | @type \f a b -> do sv <- a; svs <- b; return (do v <- sv; vs <- svs; f v vs) :: (a -> [a] -> IO (State s a)) -> IO (State s a) -> IO (State s [a]) -> IO (State s a) |
| 06:53:39 | <lambdabot> | Inferred type is less polymorphic than expected |
| 06:53:39 | <lambdabot> | Quantified type variable `s' is mentioned in the environment: |
| 06:53:39 | <lambdabot> | svs :: IO (State s a) -> t1 (bound at <interactive>:1:22) |
| 06:53:54 | <mmorrow> | @type (\f a b -> do sv <- a; svs <- b; return (do v <- sv; vs <- svs; f v vs)) :: forall s. (a -> [a] -> IO (State s a)) -> IO (State s a) -> IO (State s [a]) -> IO (State s a) |
| 06:53:56 | <lambdabot> | Not in scope: type variable `a' |
| 06:53:56 | <lambdabot> | Not in scope: type variable `a' |
| 06:53:56 | <lambdabot> | Not in scope: type variable `a' |
| 06:54:02 | <mmorrow> | @type (\f a b -> do sv <- a; svs <- b; return (do v <- sv; vs <- svs; f v vs)) :: forall s a. (a -> [a] -> IO (State s a)) -> IO (State s a) -> IO (State s [a]) -> IO (State s a) |
| 06:54:04 | <lambdabot> | Couldn't match expected type `State s a' |
| 06:54:04 | <lambdabot> | against inferred type `IO (State s a)' |
| 06:54:04 | <lambdabot> | In the expression: f v vs |
| 06:54:47 | <mmorrow> | hmm |
| 06:55:15 | <mykelyk> | any idea? |
| 06:55:31 | <mmorrow> | so you need to get the IO out of the State, run the State, and put the IO back (or something) |
| 06:56:52 | <mmorrow> | (\f a b -> do sv <- a; svs <- b; return (do v <- sv; vs <- svs; unsafePerformIO (f v vs))) :: forall s a. (a -> [a] -> IO (State s a)) -> IO (State s a) -> IO (State s [a]) -> IO (State s a) |
| 06:56:57 | <mmorrow> | is what i'd do |
| 06:57:19 | <mykelyk> | unsafePerformIO ? |
| 06:57:40 | <mmorrow> | are you sure that calling f on these inputs makes sense though? because if so then it's a design error on the part of the interpreter |
| 06:58:28 | <mmorrow> | mykelyk: you're putting it back into IO though, so it's fine here (in fact, it's really you're only choice to do what you want to do here) |
| 06:58:40 | <mmorrow> | s/you're/your/ |
| 06:59:06 | <Jedai> | mykelyk: Are you sure the type of f is correct ? Rather than IO (State...) an interpreter should rather produce something like StateT Stack IO a |
| 06:59:12 | <mmorrow> | yeah |
| 06:59:58 | <mykelyk> | probably you're right |
| 07:15:03 | <mmorrow> | @type \ios -> do State f <- liftIO ios; StateT (return . f) |
| 07:15:05 | <lambdabot> | forall t t1 (m :: * -> *). (MonadIO m) => IO (State t t1) -> StateT t m t1 |
| 07:15:48 | <mmorrow> | mykelyk: if the code for that interp isn't at-hand, you could do that to convert stuff |
| 07:16:57 | <mmorrow> | (so unsafePerformIO isn't actually the only way after-all, but the fact that this is equivalent to unsafePerformIO'ing that stuff above shows that that use of unsafePerformIO was "safe") |
| 07:17:18 | <mmorrow> | unsafePerformIO (return 42) == 42 |
| 07:17:47 | <mykelyk> | works like a charms thanks!!! |
| 07:17:47 | <mmorrow> | liftM2 (==) (return (unsafePerformIO (return 42))) (return 42) |
| 07:17:54 | <mmorrow> | mykelyk: cool :) |
| 07:36:39 | <dancor> | how do you decide between (a, b) and State a b |
| 07:39:48 | <Cale> | dancor: How do you mean? |
| 07:40:06 | <Cale> | dancor: You mean between using the state monad, and parameter passing? |
| 07:41:12 | <dibblego> | does anyone happen to know when GHC discontinued splitting into core and extra-libs? |
| 07:41:51 | <Cale> | dibblego: hmm... well, there's the Haskell Platform thing now, yeah? |
| 07:42:01 | <dibblego> | yeah |
| 07:42:16 | <dibblego> | I'm just going to give a tutoral to about 20 people and I want to make sure they all have QC |
| 07:43:14 | <Cale> | If they install it from Debian/Ubuntu, they'll have almost nothing, because the Debian guys split everything up on their own. |
| 07:43:26 | <dibblego> | bugger |
| 07:45:20 | <Cale> | Hmm, actually even the source for 6.10.3 comes in two packages. |
| 07:45:42 | <Cale> | But the binaries provided seem to be the big package. |
| 07:46:06 | <Cale> | Myself, I just install the binaries from the GHC website. |
| 07:48:16 | <twb> | Actually just ghc6 in Debian is like 100MB of stuff |
| 07:48:34 | <Cale> | hmm |
| 07:48:57 | <twb> | Uncompressed Size: 130M |
| 07:49:19 | <twb> | Make that 148 for 6.10.3 |
| 07:49:39 | <twb> | This is why I don't install GHC on my laptop; it only has 4GiB of storage. |
| 07:50:06 | <bnijk> | how do i access system beep in haskell |
| 07:50:18 | <twb> | bnijk: you could try echoing ^G to standard output |
| 07:50:22 | <Cale> | hmm, it seems my current /usr/local/lib/ghc-6.10.3 is 347M |
| 07:50:25 | <bnijk> | how do i do that :O |
| 07:50:32 | <bnijk> | wait no |
| 07:50:39 | <bnijk> | it needs to actually control the hz and duration |
| 07:50:41 | <Cale> | and I only install packages as user |
| 07:50:52 | <bnijk> | i already wrote a bash script to do this, to play a major scale and a chromatic scale on the system beep |
| 07:51:11 | <Cale> | So that's all stuff which came with GHC for me. |
| 07:51:25 | <Cale> | bnijk: hmm |
| 07:51:31 | <mmorrow> | [m@monire ~]$ du -h /usr/local/ghc/ghc-6.10.1 | tail -1 |
| 07:51:32 | <mmorrow> | 981M /usr/local/ghc/ghc-6.10.1 |
| 07:51:33 | <mmorrow> | :o |
| 07:51:39 | <dmwit> | bnijk: Call out to your bash script. ;-) |
| 07:51:56 | <bnijk> | that seems wasteful |
| 07:52:01 | <Cale> | bnijk: What does the bash script do, I wonder? |
| 07:52:12 | <bnijk> | well there's an accompanying c program |
| 07:52:19 | <bnijk> | which does beep (hz) (duration) |
| 07:52:38 | <mmorrow> | bnijk: i think it's a sysctl or something |
| 07:52:41 | <dmwit> | bnijk: If I were you, I'd call the same functions in Haskell as I did in C. |
| 07:52:55 | <bnijk> | let's see here |
| 07:53:00 | <mmorrow> | there's a program that plays songs my messing with the sysbeep, maybe that'll help |
| 07:53:00 | <twb> | mmorrow: -c (or -s? I forget). |
| 07:53:05 | <Cale> | I don't know of an existing Haskell API for that, but you can FFI import any C library calls you used. |
| 07:53:11 | <mmorrow> | twb: i can't remember |
| 07:53:19 | <bnijk> | int main(int argc, char *argv[]) { |
| 07:53:20 | <bnijk> | int fd = open("/dev/tty10", O_RDONLY); |
| 07:53:20 | <bnijk> | if (fd == -1 || argc != 3) return -1; |
| 07:53:20 | <bnijk> | return ioctl(fd, KDMKTONE, (atoi(argv[2])<<16)+(1193180/atoi(argv[1]))); |
| 07:53:20 | <twb> | I always use -sch, but it prints two lines |
| 07:53:22 | <bnijk> | } |
| 07:53:25 | <bnijk> | =-o |
| 07:53:27 | <bnijk> | how do i do that in haskell |
| 07:53:53 | <dmwit> | bnijk: Same way, call ioctl. |
| 07:54:04 | <bnijk> | there's some strange monadic way of doing that i assume |
| 07:54:07 | <dmwit> | ?where FFI |
| 07:54:07 | <lambdabot> | http://www.cse.unsw.edu.au/~chak/haskell/ffi/ |
| 07:55:03 | <xmonader> | while installing hs-twitter i get this message http://pastebin.com/d222fb15a i tried to set json package to < 4 as Saizan said but still no good |
| 07:55:37 | <Cale> | bnijk: Unless you mean the IO monad, no. (The fact that IO is a monad is irrelevant to this.) |
| 07:56:44 | <bnijk> | haskell is hard |
| 07:57:07 | <twb> | bnijk: if you try to write procedures in it, sure. |
| 07:57:08 | <dmwit> | I still don't understand why you're not just calling out to that C program. |
| 07:57:08 | <Cale> | ... I was just thinking that comprehending the meaning of that C program is pretty hard. |
| 07:57:09 | <xmonader> | bnijk, not true it's simply different |
| 07:57:25 | <dmwit> | It's the same thing you did with bash, why not do it with Haskell? |
| 07:57:48 | <dmwit> | And yeah, I'm with Cale here. That C doesn't look particularly easy to me. |
| 07:58:17 | <dmwit> | (Did you write the C code, or grab it from somewhere else?) |
| 07:58:22 | <bnijk> | (i grabbed it) |
| 07:58:29 | <twb> | Cale: looks like it converts the first command-line argument into a very stange number, then sends an ioctl to /dev/tty10 with that number as the argument to the call "KDMKTONE". |
| 07:58:38 | <Cale> | Why tty10? |
| 07:58:38 | <bnijk> | hell if i know how to do things like that |
| 07:58:48 | <twb> | Presumably because it's not allocated by default |
| 07:59:12 | <dmwit> | bnijk: So, what you're saying is, it's hard to do something in Haskell that you don't understand how to do in any other language, either. |
| 07:59:28 | <dmwit> | Forgive me if I'm not convinced of Haskell's shortcomings from that. =) |
| 07:59:30 | <bnijk> | lol |
| 07:59:35 | <bnijk> | i understand the c |
| 07:59:41 | <bnijk> | i just really don't know how to do it in haskell |
| 07:59:47 | <twb> | bnijk: then what is the purpose of the right shifting? |
| 07:59:53 | <twb> | Sorry, left shifting |
| 07:59:55 | <bnijk> | what? |
| 08:00:08 | <bnijk> | > map (\x->440*2**(x/12)) [0..12] |
| 08:00:08 | <twb> | bnijk: what is the purpose of the left shift used in that C code? |
| 08:00:09 | <lambdabot> | [440.0,466.1637615180899,493.8833012561241,523.2511306011972,554.3652619537... |
| 08:00:16 | <Cale> | atoi(argv[2])<<16 -- why shift left by 16 bits? |
| 08:00:43 | <twb> | bnijk: if you understand the C code, you should be able to explain it to us. |
| 08:00:52 | <Cale> | (1193180/atoi(argv[1])) -- and wtf is this? |
| 08:00:54 | <bnijk> | ah |
| 08:01:05 | <bnijk> | well the second part isj ust translating the hz |
| 08:01:11 | <twb> | Cale: I assume 11938180 is a frequency in Hz |
| 08:01:14 | <mmorrow> | twb: http://www.frank-buss.de/beep/index.html |
| 08:01:15 | <bnijk> | to what the beeping thing recognizes |
| 08:01:21 | <twb> | Bleh, too many beers |
| 08:01:22 | <bnijk> | whatever you call it |
| 08:01:29 | <Cale> | mm... |
| 08:01:34 | <bnijk> | and the first part... |
| 08:01:41 | <bnijk> | well |
| 08:01:48 | <twb> | mmorrow: yes, ^G is character 07 in ASCII |
| 08:01:50 | <bnijk> | that i don't get |
| 08:01:53 | <yowgi> | “KDMKTONE Generate tone of specified length. The lower 16 bits of argp specify the period in clock cycles, and the upper 16 bits give the duration in msec.” |
| 08:02:05 | <bnijk> | ah |
| 08:02:07 | <twb> | O dpm |
| 08:02:15 | <twb> | I don't think tty10 is special |
| 08:02:26 | <twb> | Other than being in /etc/securetty. |
| 08:02:32 | <yowgi> | 1193180 = CLOCK_TICK_RATE |
| 08:02:36 | <dmwit> | Ah, so the code is actually wrong. |
| 08:02:47 | <mmorrow> | nice, http://tldp.org/LDP/lpg/node83.html |
| 08:02:57 | <twb> | Runtime.getRuntime().exec("beep.exe 440 1000"); |
| 08:02:58 | <dmwit> | Although I suppose for something like this, sanitizing input is pretty low priority. |
| 08:02:58 | <twb> | Ahahahaha |
| 08:03:07 | <bnijk> | yes dmwit ;) |
| 08:03:20 | <mmorrow> | ACTION wraps messing with the pc speaker with the ffi |
| 08:03:26 | <twb> | yowgi: surely the tick rate depends on the architecture and the kernel version? |
| 08:03:34 | <dmwit> | Anyway, maybe we should be asking why you're not just using a real sound library. |
| 08:03:45 | <yowgi> | twb, probably, yes. |
| 08:03:46 | <bnijk> | dmwit: i wanted to do it with the system beep, of course |
| 08:03:48 | <twb> | dmwit: because this is "easier" man! |
| 08:03:53 | <bnijk> | anybody can write a PCM |
| 08:04:12 | <dmwit> | ACTION blinks |
| 08:04:20 | <dmwit> | And not anybody can copy and paste code from the web? |
| 08:04:35 | <twb> | Description: beep does what you'd expect: it beeps. But unlike printf "\a" beep allows you to control pitch, duration, and repetitions. Its job is to live inside shell/perl scripts and allow more granularity than one has otherwise. It is controlled completely through command line options. It's not supposed to be complex, and it isn't - but it makes system monitoring (or whatever else it gets hacked into) much more |
| 08:04:35 | <twb> | Homepage: http://johnath.com/beep/ |
| 08:04:36 | <bnijk> | that code only does the beeping itself |
| 08:04:44 | <bnijk> | how do i explain |
| 08:05:15 | <twb> | Package: cw |
| 08:05:15 | <twb> | Description: This package contains a simple command line client called cw, which sounds characters as Morse code on the console speaker. The included cwgen binary can generate groups of random characters for Morse code practice.[...] |
| 08:05:27 | <bnijk> | i don'tw ant to manipulate timbre |
| 08:05:38 | <bnijk> | if i wanted to do that i'd open up a whole wave to do it |
| 08:05:43 | <bnijk> | just wanted the tones |
| 08:06:58 | <bnijk> | speaking of which - how do i work in hex in haskell |
| 08:07:07 | <twb> | I actually looked into using cw for a headless server SOE, so it could report panics to untrained end users |
| 08:07:15 | <dmwit> | > 0x20 -- the usual way |
| 08:07:17 | <lambdabot> | 32 |
| 08:07:19 | <bnijk> | k |
| 08:07:39 | <bnijk> | so let's see here |
| 08:07:40 | <twb> | dmwit: what, no DEFAULT-READ-BASE ? ;-) |
| 08:08:16 | <bnijk> | a wav header would be formatted like... |
| 08:08:16 | <twb> | (SETQ DEFAULT-WRITE-BASE 2 DEFAULT-READ-BASE 36) (PRINC aaz3) ;-) |
| 08:08:35 | <twb> | ...not bothering to check CLHS, because I'm lazy. |
| 08:09:28 | <p_l> | twb: why setq and not setf? |
| 08:09:33 | <dmwit> | :t readIntAtBase |
| 08:09:34 | <lambdabot> | Not in scope: `readIntAtBase' |
| 08:09:35 | <twb> | p_l: haha |
| 08:09:40 | <dmwit> | :t readInt |
| 08:09:40 | <lambdabot> | forall a. (Num a) => a -> (Char -> Bool) -> (Char -> Int) -> String -> [(a, String)] |
| 08:10:17 | <bnijk> | > 0x524946462408000057415645666d74201000000001000200 ++ sampleRate ++ byteRate ++ 0x0400 ++ bps ++ firstSubchunk ++ 0x000800000000 ++ wav |
| 08:10:18 | <lambdabot> | Not in scope: `sampleRate'Not in scope: `byteRate'Not in scope: `bps'Not in... |
| 08:10:24 | <bnijk> | you're not in scope~! |
| 08:10:27 | <bnijk> | ACTION slaps lambdabot |
| 08:10:28 | <dmwit> | :t index |
| 08:10:29 | <lambdabot> | forall a. (Ix a) => (a, a) -> a -> Int |
| 08:10:35 | <dmwit> | :t findIndex |
| 08:10:36 | <lambdabot> | forall a. (a -> Bool) -> [a] -> Maybe Int |
| 08:10:49 | <bnijk> | although the > will turn it to int i assume |
| 08:11:21 | <bnijk> | bah |
| 08:11:45 | <dmwit> | > readInt 36 (const True) (\c -> fromJust $ lookup c (zip (['0'..'9'] ++ ['a'..'z']) [0..])) "aaz3" |
| 08:11:47 | <lambdabot> | [(480783,"")] |
| 08:12:14 | <bnijk> | back to LYAH |
| 08:12:40 | <dmwit> | bnijk: uh... (++) operates on String |
| 08:12:47 | <bnijk> | meh |
| 08:12:49 | <bnijk> | make those :'s |
| 08:13:01 | <bnijk> | and then mush it all together |
| 08:13:19 | <bnijk> | i don't remember |
| 08:13:31 | <dmwit> | You have a bit to learn, yet, it seems. =) |
| 08:13:40 | <bnijk> | i would imagine |
| 08:13:53 | <bnijk> | so tell me then, how do i just mash together data |
| 08:14:10 | <dmwit> | Also, what the hell am I thinking? (++) works on [a] for any a, not just Char. |
| 08:14:20 | <dmwit> | bnijk: You don't just mash together data. |
| 08:14:25 | <dmwit> | That's the beauty of Haskell. |
| 08:14:33 | <Cale> | Well, I wrote a program which I imagine does the same thing as your C code, but I have no idea how to test it. |
| 08:14:36 | <mmorrow> | heh http://www.scylla-charybdis.com/tool.php/kdmktone |
| 08:14:46 | <mmorrow> | for a in 1 2 3; do ./kdmktone 1000 1500 2000 0; done |
| 08:15:12 | <bnijk> | what do you mean |
| 08:15:17 | <bnijk> | @dmwit |
| 08:15:18 | <lambdabot> | Unknown command, try @list |
| 08:15:52 | <Cale> | actually it does seem to work. |
| 08:16:13 | <Cale> | ACTION goes to hpaste |
| 08:16:23 | <bnijk> | if i'm trying to make a pcm wav, dmwit, they have to be aligned in that order |
| 08:16:34 | <bnijk> | how else do you propose to do that |
| 08:16:40 | <Cale> | http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5112#a5112 |
| 08:16:49 | <Cale> | bnijk: ^^ |
| 08:17:08 | <bnijk> | nifty |
| 08:17:16 | <dmwit> | bnijk: We have Data.Word for that, but your first guy there will almost certainly overflow a Word. |
| 08:17:21 | <bnijk> | so cale |
| 08:17:23 | <Cale> | It doesn't work without root permissions on my machine. |
| 08:17:34 | <bnijk> | plug in map (\x->440*2**(x/12)) [0..12] into that |
| 08:17:40 | <bnijk> | and you have a chromatic scale :) |
| 08:18:01 | <Cale> | Well, modulo the numeric conversion. |
| 08:18:10 | <bnijk> | ah, yes |
| 08:18:15 | <bnijk> | ...modulo? |
| 08:18:21 | <bnijk> | 0.0 |
| 08:18:33 | <Cale> | I just mean, ignoring that |
| 08:18:54 | <bnijk> | which part of your program is the frequency |
| 08:19:05 | <Cale> | I just stole the two parameters from your program |
| 08:19:23 | <bnijk> | ACTION looks again |
| 08:19:24 | <bnijk> | ok |
| 08:19:29 | <Cale> | I think a is the frequency |
| 08:19:35 | <Cale> | That's less convenient :) |
| 08:19:54 | <Cale> | ACTION swaps them |
| 08:19:57 | <danra> | any recommendations for a package to translate a tree of strings into an image of tree of strings? |
| 08:20:16 | <dmwit> | How about Data.Tree.showTree? |
| 08:20:19 | <Cale> | ah, it actually does beep, so it should probably just be called beep |
| 08:20:20 | <dmwit> | Or whatever that thing is called. |
| 08:20:20 | <bnijk> | ACTION frowns |
| 08:20:37 | <dmwit> | drawTree |
| 08:20:56 | <danra> | thanks dmwit, will look at that |
| 08:21:03 | <bnijk> | dmwit: i wrote a program once in quasi-functional php+postgres that just compiled little bits of text people wrote |
| 08:21:06 | <bnijk> | into a huge tree |
| 08:21:13 | <bnijk> | the things that came out of the project |
| 08:21:17 | <bnijk> | mmnph |
| 08:22:20 | <bnijk> | i want a big poster of that... |
| 08:22:24 | <bnijk> | ACTION fiddles some more |
| 08:24:02 | <bnijk> | cale you still there? |
| 08:24:19 | <Cale> | yeah |
| 08:24:36 | <Cale> | hmm, it doesn't seem to block until the beep is finished |
| 08:24:56 | <bnijk> | can you explain this 'foreign...' line to me |
| 08:25:44 | <bnijk> | or should i just look up these functions |
| 08:25:46 | <Cale> | sure |
| 08:25:53 | <Cale> | It's mostly special syntax. |
| 08:26:29 | <dmwit> | foreign isn't a function, actually. |
| 08:26:46 | <bnijk> | what is it then, a keyword |
| 08:26:52 | <dmwit> | A foreign import provides a bridge from Haskell to other languages. |
| 08:26:56 | <Cale> | yes |
| 08:27:06 | <dmwit> | Yep, "foreign", "import", and "ccall" are all keywords there. |
| 08:27:14 | <bnijk> | and these types |
| 08:27:16 | <Cale> | ccall is the calling convention |
| 08:27:20 | <bnijk> | Fd and CInt |
| 08:27:20 | <twb> | dmwit: language*s*? Other than C, what are they? |
| 08:27:40 | <dmwit> | twb: I believe there's a spec for Java, at least, though I doubt anyone implements it. |
| 08:27:43 | <dmwit> | ?where ffi |
| 08:27:43 | <lambdabot> | http://www.cse.unsw.edu.au/~chak/haskell/ffi/ |
| 08:27:46 | <twb> | Or are we talking about Haskell rather than GHC now? |
| 08:28:01 | <dmwit> | yep |
| 08:28:08 | <dmwit> | err: yep, Haskell, not GHC |
| 08:28:11 | <twb> | Righto |
| 08:28:14 | <Cale> | and then "sys/ioctl.h ioctl" describes that the function we want is described in the header <sys/ioctl.h> and is called ioctl |
| 08:28:23 | <twb> | I tend to conflate the two in implied contexts. |
| 08:28:30 | <Cale> | Then it's followed by the name of the Haskell function to import it as |
| 08:28:49 | <Cale> | and the Haskell type at which we want to import it... there has to be some agreement between this type and the C type |
| 08:28:58 | <Cale> | (but there's a bit of flexibility) |
| 08:29:00 | <dmwit> | twb: I guess "stdcall" is the only other required calling convention. |
| 08:29:11 | <quicksilver> | twb: other than C, any other language which supports the C ABI :) |
| 08:29:15 | <dmwit> | twb: (For win32 api calls.) |
| 08:29:31 | <bnijk> | hmph |
| 08:29:35 | <bnijk> | yours isn't in /usr/share? |
| 08:29:44 | <Cale> | what? |
| 08:29:46 | <bnijk> | er, /usr/include/sys/ |
| 08:29:50 | <bnijk> | ioctl.h |
| 08:29:51 | <Cale> | It is. |
| 08:30:31 | <bnijk> | ok so that reads in #define _SYS_IOCTL_H |
| 08:30:40 | <Cale> | hmm? |
| 08:30:45 | <bnijk> | in ioctl.h |
| 08:30:55 | <bnijk> | or #ifndef |
| 08:31:01 | <Cale> | Well, it reads the declaration of the ioctl function from there |
| 08:31:20 | <bnijk> | how did you know what type ioctl had O.O |
| 08:31:31 | <Cale> | I gave it by hand. |
| 08:31:43 | <Cale> | (I looked it up in the manpage and looked at what you wrote) |
| 08:31:47 | <dmwit> | From man ioctl |
| 08:32:11 | <Cale> | ioctl normally takes an arbitrary number of parameters, so I imported it at one specific type |
| 08:32:44 | <bnijk> | ACTION nods |
| 08:33:07 | <Cale> | The first parameter is a file descriptor, and the unix package for Haskell already has stuff for working with those |
| 08:33:21 | <Cale> | So I imported System.Posix.Types and System.Posix.IO |
| 08:33:29 | <Cale> | and used the Fd type defined there |
| 08:33:41 | <Cale> | CInt as you'd imagine is just a C int |
| 08:33:59 | <bnijk> | so those types are for the arguments passed to kdmktone |
| 08:34:10 | <Cale> | to ioctl rather |
| 08:34:15 | <bnijk> | er, yeah |
| 08:34:23 | <Cale> | kdmktone is a particular CInt constant |
| 08:34:24 | <danra> | any recommendations for a package to translate a tree of strings into an *pretty* image of tree of strings? (*not* Data.Tree.drawTree :-) |
| 08:34:47 | <bnijk> | danra: GIMP |
| 08:34:48 | <dmwit> | danra: As in, a non-text image? |
| 08:34:53 | <danra> | yup |
| 08:34:53 | <Cale> | by the way: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5112#a5113 |
| 08:34:58 | <dmwit> | danra: If so, you might be interested in the GraphViz library. |
| 08:35:19 | <danra> | dmwit: Thanks again, will look at that |
| 08:35:19 | <bnijk> | oo |
| 08:35:28 | <dmwit> | Have you ever used GraphViz? |
| 08:35:31 | <danra> | no |
| 08:35:33 | <Cale> | oh, you also need to import Control.Concurent |
| 08:35:38 | <Cale> | Concurrent* |
| 08:35:51 | <Cale> | (for threadDelay) |
| 08:35:59 | <bnijk> | wait what are the imports for the second one |
| 08:36:07 | <mmorrow> | ACTION just destroyed his eardrums with high pitched pure sinewaves of death |
| 08:36:13 | <Cale> | I'll add a revision with the whole file :) |
| 08:36:19 | <dmwit> | danra: Okay, well, there's lots of information on the web about it, anyway. |
| 08:36:33 | <dmwit> | danra: The GraphViz library just produces input for the various graphviz executables. |
| 08:36:55 | <bnijk> | and just out of curiosity (looking at haskell wiki) is there a haskell word processor? |
| 08:37:04 | <bnijk> | or just IDEs and text editors |
| 08:37:14 | <dmwit> | There's no word processor that I know of. |
| 08:37:23 | <dmwit> | Are you going to make one? |
| 08:37:35 | <bnijk> | i'm more eager about making my game, dmwit |
| 08:37:42 | <bnijk> | but haskell + opengl is a goddamn nightmare |
| 08:37:52 | <dmwit> | huh |
| 08:38:08 | <Cale> | http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5112#a5114 |
| 08:38:08 | <bnijk> | imagine marcel duchamp + escape velocity |
| 08:38:21 | <bnijk> | ah, ok |
| 08:38:31 | <Cale> | bnijk: I even added a function to play the scale for you :) |
| 08:38:42 | <bnijk> | ACTION compiles |
| 08:38:52 | <Cale> | I needed to run it as root. |
| 08:39:02 | <bnijk> | yeah |
| 08:39:22 | <Cale> | Oh, it doesn't have a main. |
| 08:39:38 | <Cale> | main = playScale 300 440 |
| 08:39:41 | <Cale> | will do |
| 08:40:04 | <bnijk> | how do you compile |
| 08:40:06 | <bnijk> | ghc -S? |
| 08:40:15 | <bnijk> | mmnn ok |
| 08:40:26 | <Cale> | ghc --make beep |
| 08:40:27 | <bnijk> | nope |
| 08:40:53 | <bnijk> | nifty |
| 08:41:02 | <bnijk> | now to make it play "three blind mice"... |
| 08:42:01 | <bnijk> | wow that did not work quite right |
| 08:42:06 | <bnijk> | forgot to use whole steps |
| 08:42:22 | <bnijk> | screw this, it's gonna play 'smoke on the water' |
| 08:42:26 | <mmorrow> | here's my Beep module http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5115#a5115 |
| 08:43:22 | <mmorrow> | gotta figure out now how to encode (Hz,Seconds) -> (?,?) that (ioctl_l 1 kDMKTONE) uses... then can make music |
| 08:44:14 | <bnijk> | cale: [floor (root * 2 ** (k/12)) | k <- [0,3,5,0,3,6,5,0,3,5,3,0]] |
| 08:44:14 | <Cale> | It doesn't need any permissions if you use /dev/tty, but then it only works for me in a real linux terminal, and not in gnome-terminal. :) |
| 08:44:26 | <mmorrow> | (warning: "toggle n", where n > ~70 or 80 is painful if the volume's up enough) |
| 08:44:39 | <bnijk> | :D |
| 08:45:02 | <mmorrow> | ah yeah, i had to switch to a no-X screen |
| 08:45:04 | <bnijk> | i've been thinking a lot over the past few months of what you can do with haskell and pcm |
| 08:45:16 | <bnijk> | i've been making music with granular synth for the past few years - but this has gotta be even better |
| 08:45:18 | <mmorrow> | bnijk: probably cool stuff |
| 08:45:40 | <mmorrow> | ooh, cool. yeah i bet haskell would be perfect |
| 08:47:24 | <mmorrow> | Cale: ooh, awesome just saw your paste |
| 08:47:40 | <bnijk> | oo even better |
| 08:47:41 | <bnijk> | floor (root * 2 ** (k/12)) | k <- [0,2,4,0,0,2,4,0,4,5,7,4,5,7,7,9,7,5,4,0,7,9,7,5,4,0,0,-5,0,0,-5,0]] |
| 08:47:54 | <Cale> | mmorrow: I just translated his C code |
| 08:48:23 | <mmorrow> | Cale: what C code? |
| 08:48:26 | <mmorrow> | oh, hah |
| 08:48:35 | <mmorrow> | i should've looked more closely |
| 08:49:10 | <dmwit> | Heh, I was wondering why I couldn't beep my speaker here, and then I remembered: I disabled those kernel modules. ^_^ |
| 08:49:17 | <mmorrow> | err, i think the timeval is supposed to be a CLong |
| 08:49:25 | <bnijk> | holy crap look at all the midi programs in the AUR |
| 08:49:48 | <Cale> | mmorrow: Quite possibly. I didn't think too hard about it. |
| 08:50:06 | <mmorrow> | Cale: wouldn't that mess with the bit shifting though? |
| 08:50:10 | <mmorrow> | ACTION is confused |
| 08:50:32 | <Cale> | Well, the shift is only 16 bits. |
| 08:51:02 | <Cale> | So the resulting constant is only 32 bits anyway. |
| 08:52:24 | <Cale> | mmorrow: How do I get your beep to work? |
| 08:52:29 | <mmorrow> | Cale: ahhh. i just realized why my version was crushing my ears.. |
| 08:52:42 | <mmorrow> | Cale: i didn't get that far :) |
| 08:52:54 | <mmorrow> | (although "toggle 50" e.g. works) |
| 08:53:06 | <mmorrow> | (don't go above 70 !!@) |
| 08:53:44 | <bnijk> | http://gimp.org/unix/howtos/gimp-midi.html :O :O :O |
| 08:53:53 | <bnijk> | and people tell me photoshop is better |
| 08:54:03 | <mmorrow> | Cale: oh, also it's outputing to stdout, so you need to be on a real tty |
| 08:54:10 | <Cale> | mmorrow: Ah... |
| 08:54:49 | <dmwit> | bnijk: Sounds... useful. |
| 08:55:48 | <bnijk> | it's clearly not |
| 08:55:51 | <bnijk> | what's important is that it's possible |
| 08:55:55 | <Cale> | bnijk: More than midi, it would be fun to have something to control the brush parameters in realtime using, say, the average amplitude and pitch of whatever's coming in on the microphone :) |
| 08:56:10 | <bnijk> | automatic livepainting cale :O |
| 08:56:33 | <bnijk> | although i suppose there's already visualization programs |
| 08:56:43 | <bnijk> | ACTION leaves to go make eggs |
| 08:56:45 | <dmwit> | There's tablets... |
| 08:56:55 | <bnijk> | that's not the point dmwit ;) |
| 08:57:04 | <dmwit> | ACTION is always missing the point |
| 08:57:25 | <Cale> | You could combine it with a tablet too :) |
| 08:57:58 | <Cale> | Hehe, have the pitch mapped to hue, with an octave producing the same colour. :) |
| 08:58:07 | <Cale> | (but maybe brighter? :) |
| 08:58:14 | <bnijk> | that reminds me |
| 08:58:19 | <bnijk> | i have to install that program... |
| 08:58:20 | <bnijk> | max |
| 08:58:20 | <bnijk> | ! |
| 08:59:57 | <bnijk> | or something like that - video + audio programming |
| 09:01:19 | <bnijk> | as technology advances - imagine even thinking of this stuff a century ago |
| 09:02:09 | <xmonader> | anyone got hs-twitter installed ? |
| 09:02:19 | <Axman6> | ChilliX: you around? |
| 09:02:46 | <Cale> | xmonader: Not me... |
| 09:02:56 | <xmonader> | Cale :( |
| 09:03:18 | <xmonader> | i reinstalled json, installed < 4 and still :@ |
| 09:07:29 | <Axman6> | ACTION thinks ChilliX should come down to canberra and give his data parallel haskell talk at ANU |
| 09:09:27 | <Cale> | 眠いんで、お休み、みんな！ |
| 09:11:30 | <xmonader> | Cale, :S |
| 09:23:31 | <jn> | does anyone have a more concise way of doing the following? http://paste.lisp.org/display/80642 |
| 09:25:58 | <skorpan> | jn: i don't know, but shouldn't maxBound for Hours be "Hours 23"? |
| 09:26:08 | <bnijk> | yeah..and 0..59 |
| 09:26:29 | <bnijk> | while you're at it |
| 09:26:38 | <bnijk> | ACTION coughs |
| 09:26:41 | <jn> | no. |
| 09:26:49 | <bnijk> | lol |
| 09:27:24 | <jn> | it's an inclusive/exclusive range |
| 09:27:37 | <quicksilver> | jn: well, instead of "ranged 0 24 Hours" you could do "ranged minBound maxBound Hours" |
| 09:28:03 | <jn> | aye, that removes the duplication, but it's still horribly long |
| 09:28:18 | <quicksilver> | then you can give 'ranged minBound maxBound' a name |
| 09:28:36 | <jn> | that's true |
| 09:29:09 | <quicksilver> | you'd need to change stuff slightly - maxBound would be (Hours 24) not just 24. |
| 09:29:42 | <LeCamarade> | What does it mean when a GHC program says "thread blocked indefinitely"? |
| 09:29:47 | <jn> | yeah. just would've thought there'd be some more concise way of making a bounded integral type. |
| 09:30:16 | <skorpan> | @hoogle a -> [a] -> [a] |
| 09:30:17 | <lambdabot> | Data.List intersperse :: a -> [a] -> [a] |
| 09:30:17 | <lambdabot> | Data.List insertBy :: (a -> a -> Ordering) -> a -> [a] -> [a] |
| 09:30:17 | <lambdabot> | Data.List deleteBy :: (a -> a -> Bool) -> a -> [a] -> [a] |
| 09:30:21 | <LeCamarade> | I only use MVar, SampleVars, and Chan from Control.Concurrent |
| 09:31:35 | <quicksilver> | LeCamarade: it means you have a deadlock. |
| 09:31:51 | <quicksilver> | LeCamarade: no thread can proceed because they are all waiting on another thread |
| 09:33:27 | <LeCamarade> | Hmm. |
| 09:33:58 | <LeCamarade> | I thought I had architected it to prevent that. But threads make quantum mechanics look all too deterministic. |
| 09:43:16 | <ChilliX> | Axman6: Hey |
| 09:43:16 | <lambdabot> | ChilliX: You have 1 new message. '/msg lambdabot @messages' to read it. |
| 09:51:08 | <jn> | http://paste.lisp.org/display/80643 think that's the best i can do |
| 09:53:19 | <ivanm> | What's the difference between isNumber and isDigit |
| 09:53:20 | <ivanm> | ? |
| 09:53:29 | <Beelsebob> | @type isNumber |
| 09:53:31 | <lambdabot> | Char -> Bool |
| 09:53:38 | <ivanm> | they have the same type ;-) |
| 09:53:41 | <Beelsebob> | > isNumber '.' |
| 09:53:42 | <lambdabot> | False |
| 09:54:00 | <Beelsebob> | > isNumber '-' |
| 09:54:02 | <lambdabot> | False |
| 09:54:03 | <Beelsebob> | no idea |
| 09:54:07 | <ivanm> | duh, I should probably have a look at the docs rather than trying to get lambdabot to work it out for me... |
| 09:54:17 | <Beelsebob> | what is the diff? |
| 09:54:48 | <ivanm> | isNumber: "Selects Unicode numeric characters, including digits from various scripts, Roman numerals, etc. " |
| 09:54:52 | <ivanm> | isDigit is for 0..9 |
| 09:55:05 | <Beelsebob> | oh, neat |
| 09:55:20 | <RayNbow> | > isNumber 'V' |
| 09:55:22 | <lambdabot> | False |
| 09:55:41 | <ivanm> | RayNbow: must be special unicode chars |
| 09:55:48 | <RayNbow> | yeah, I figured :p |
| 09:56:04 | <Beelsebob> | > isNumber '➁' |
| 09:56:05 | <trofi^w> | > isNumber "a number" |
| 09:56:07 | <lambdabot> | True |
| 09:56:07 | <lambdabot> | Couldn't match expected type `Char' against inferred type `[Char]' |
| 09:56:22 | <Beelsebob> | > isNumber '¾' |
| 09:56:23 | <lambdabot> | True |
| 09:59:16 | <ivanm> | Beelsebob: heh |
| 09:59:39 | <dobblego> | what is a good property for a beginner to automated testing to demonstrate ==> ? |
| 09:59:45 | <ivanm> | Beelsebob: what char is the first one you tried (before trofi^w's attempt)? |
| 10:00:12 | <trofi^w> | twho in a circle |
| 10:00:19 | <ivanm> | ahhh |
| 10:01:07 | <lilac> | > isNumber '²' |
| 10:01:09 | <lambdabot> | True |
| 10:03:02 | <lilac> | > isNumber '⁰' |
| 10:03:04 | <lambdabot> | True |
| 10:03:28 | <lilac> | > all isNumber "¹²³⁴⁵⁶⁷⁸⁹⁰" |
| 10:03:30 | <lambdabot> | True |
| 10:05:18 | <lilac> | @hoogle isNumber |
| 10:05:19 | <lambdabot> | Data.Char isNumber :: Char -> Bool |
| 10:05:20 | <lilac> | @hoogle whichNumberIsItThough |
| 10:05:21 | <lambdabot> | No results found |
| 10:08:30 | <Axman6> | > filter isNumber [chr 0 ..] |
| 10:08:32 | <lambdabot> | "0123456789\178\179\185\188\189\190\1632\1633\1634\1635\1636\1637\1638\1639... |
| 10:08:52 | <Axman6> | > text . drop 10 . filter isNumber $ [chr 0 ..] |
| 10:08:53 | <lambdabot> | ²³¹¼½¾٠١٢٣٤٥٦٧٨٩۰۱۲۳۴۵۶۷۸۹०१२३४५६à¥... |
| 10:09:30 | <RayNbow> | that doesn't look quite right :p |
| 10:09:37 | <Axman6> | no... |
| 10:09:42 | <Axman6> | > text . drop 30 . filter isNumber $ [chr 0 ..] |
| 10:09:44 | <lambdabot> | ۴۵۶۷۸۹०१२३४५६७८९০১২৩৪৫৬৭৮৯৴... |
| 10:10:05 | <Axman6> | > text . drop 20 . filter isNumber $ [chr 0 ..] |
| 10:10:07 | <lambdabot> | ٤٥٦٧٨٩۰۱۲۳۴۵۶۷۸۹०१२३४५६à¥à¥®à¥¯à§¦à§§à§¨à§©à... |
| 10:12:03 | <int-e> | > isNumber '¾' |
| 10:12:06 | <lambdabot> | True |
| 10:15:19 | <ivanm> | int-e: Beelsebob beat you to it ;-) |
| 10:15:50 | <ivanm> | Axman6: what does text do? display unicode chars? |
| 10:15:54 | <ivanm> | @type text |
| 10:15:56 | <lambdabot> | String -> Doc |
| 10:16:00 | <ivanm> | ahhh |
| 10:16:49 | <skorpan> | @pl (\x y -> x == 3) |
| 10:16:50 | <lambdabot> | const . (3 ==) |
| 10:17:13 | <skorpan> | isn't there any combinator for doing the same thing? |
| 10:18:15 | <Botje> | :t curry (first (==3)) |
| 10:18:16 | <lambdabot> | forall a b. (Num a) => a -> b -> (Bool, b) |
| 10:18:23 | <skorpan> | :t first |
| 10:18:25 | <lambdabot> | forall (a :: * -> * -> *) b c d. (Arrow a) => a b c -> a (b, d) (c, d) |
| 10:18:29 | <Botje> | hmm, nope |
| 10:18:43 | <Botje> | fst . curry . first (==3) isn't really very readable :P |
| 10:19:20 | <lilac> | @type curry ((==3) . fst) |
| 10:19:22 | <lambdabot> | forall a b. (Num a) => a -> b -> Bool |
| 10:29:13 | <pao> | hi all, what is the most efficient way to parse an iso format timestamp stored in a ByteString? |
| 10:30:34 | <opqdonut> | write a state machine |
| 10:32:44 | <ivanm> | if I want to span a String such that I take the first part is either numbers of '.' and at most one lower case letter on the end, what would be the best way of doing it (without regexes!)? |
| 10:33:23 | <dancor> | is there an easy way to make (data MyData a = MyData {info :: [String], res :: Maybe a}) have a Monad instance that acts like (type MyData a = StateT [String] Maybe a) |
| 10:33:25 | <ivanm> | span (\c -> isDIgit c || c == '.') str, and then check the first char of the second bit of the String to see if it's a lower case letter? |
| 10:34:07 | <mauke> | @unmtl StateT [String] Maybe a |
| 10:34:08 | <lambdabot> | [String] -> Maybe (a, [String]) |
| 10:34:39 | <mauke> | dancor: eh? |
| 10:34:41 | <dancor> | hm |
| 10:34:53 | <pao> | opqdonut: could you elaborate? :-) |
| 10:34:54 | <dancor> | i guess i don't understand StateT |
| 10:35:30 | <dobblego> | why do I get a warning and not an error for missing coarbitrary on Test.QuickCheck.Arbitrary instance declaration? |
| 10:35:31 | <dancor> | my thing should be like ([String], Maybe a) and bind should just concat the Strings and fail on any Nothing |
| 10:35:35 | <dancor> | i guess i have to write it out |
| 10:36:11 | <dancor> | @mtl [String] -> Maybe (a, [String]) |
| 10:36:11 | <lambdabot> | Maybe you meant: ft map msg pl unmtl url |
| 10:37:07 | <Lemmih> | ?unmtl MaybeT (Writer [String]) a |
| 10:37:07 | <lambdabot> | (Maybe a, [String]) |
| 10:39:03 | <lilac> | ivanm: i got a parse error on your question :) |
| 10:39:23 | <mmorrow_> | Cale: new and improved Beep: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5118#a5118 |
| 10:39:42 | <mmorrow_> | it's too bad you can only play one freq at a time.. |
| 10:40:12 | <dancor> | beepcluster |
| 10:40:26 | <mmorrow_> | zomg |
| 10:40:29 | <ivanm> | lilac: I need to parse something that is of the form [0-9]+(\.[0-9]+)* with an optional trailing lowercase letter |
| 10:40:50 | <ivanm> | and I want to do so without regexes :( (since I want to have no deps apart from GHC and cabal) |
| 10:43:05 | <Botje> | i thought Text.Regex was in base nowadays? |
| 10:44:06 | <dancor> | {- verify possibleFinalLetter and put into your data structure -} where (mainThing, possibleFinalLetter) = span (\ x -> isDigit x || x == '.') ? |
| 10:44:26 | <dancor> | ivanm seems to be way ahead of me |
| 10:44:56 | <dancor> | you don't want .. tho |
| 10:45:14 | <dancor> | anyway it's easy to split on . and then work with the resulting list |
| 10:45:24 | <ivanm> | recursively? |
| 10:45:30 | <ivanm> | Botje: it isn't, AFAICT |
| 10:45:44 | <yowgi> | ivanm, how about with pattern matching and Maybe? |
| 10:45:50 | <ivanm> | Botje: in my 6.10.3 haddock page, there's no regex listed |
| 10:45:56 | <Botje> | > fmap (head) $ listToMaybe (reads "12345.6789p") :: Maybe (Double, Char) |
| 10:45:58 | <lambdabot> | Couldn't match expected type `[a]' |
| 10:46:03 | <ivanm> | yowgi: pattern matching? on each individual possible lowercase letter? |
| 10:46:12 | <yowgi> | sorry i meant guards |
| 10:47:00 | <yowgi> | parse (c:cs) | isNumber c = parseSecondPart cs c |
| 10:47:11 | <yowgi> | parseSecondPart using the span |
| 10:47:38 | <yowgi> | all other non-matching cases = Nothing |
| 10:47:47 | <Botje> | > fmap (second (head)) $ listToMaybe $ reads "12345.6789p" :: Maybe (Double, Char) |
| 10:47:48 | <lambdabot> | Just (12345.6789,'p') |
| 10:47:50 | <ivanm> | ummmmm.... that sounds kinda dodgy |
| 10:47:54 | <Botje> | ivanm: ^^ |
| 10:48:15 | <Botje> | you probaby want to do something about the head and replace it with safehead, but that's one way. |
| 10:48:22 | <dancor> | what about 1.1.1p |
| 10:49:00 | <BONUS> | ?umtl WriterT w (ContT r []) a |
| 10:49:00 | <lambdabot> | (a -> w -> [r]) -> [r] |
| 10:49:07 | <BONUS> | uhh |
| 10:49:11 | <BONUS> | is this correct? |
| 10:50:08 | <ivanm> | > fmap (second (head)) $ listToMaybe $ reads "12345.6789pa" :: Maybe (Double, Char) |
| 10:50:09 | <lambdabot> | Just (12345.6789,'p') |
| 10:50:18 | <Botje> | don't ask me, Cont* is scary voodoo magic |
| 10:50:20 | <ivanm> | > fmap (second (head)) $ listToMaybe $ reads "12345.6789-pa" :: Maybe (Double, Char) |
| 10:50:22 | <lambdabot> | Just (12345.6789,'-') |
| 10:50:32 | <BONUS> | isnt WriterT w (ContT r []) a supposed to be ((a,w) -> [r]) -> [r] |
| 10:51:04 | <ivanm> | I'm already doing a case on span (can't use reads, as it might have more than one .num part) |
| 10:51:14 | <mauke> | @unmtl ContT r (WriterT w []) a |
| 10:51:14 | <lambdabot> | (a -> [(r, w)]) -> [(r, w)] |
| 10:51:49 | <BONUS> | that one looks ok |
| 10:51:52 | <BONUS> | but |
| 10:51:59 | <BONUS> | @umtl WriterT w (ContT r []) a |
| 10:51:59 | <lambdabot> | (a -> w -> [r]) -> [r] |
| 10:52:07 | <BONUS> | this looks kinda wrong |
| 10:52:56 | <mauke> | I have to agree |
| 10:55:43 | <BONUS> | @umtl WriterT w (ListT Maybe) a |
| 10:55:44 | <lambdabot> | Maybe [(a, w)] |
| 10:56:18 | <BONUS> | this looks ok yeah but it screws it up for WriterT w (ContT r []) a |
| 10:57:14 | <BONUS> | who could be notified of this bug, i wonder |
| 10:57:31 | <ivanm> | ACTION wonders if learning/using Text.ParserCombinators would be worth it... |
| 10:57:54 | <Botje> | parsec is dead easy and dead sexy |
| 11:00:30 | <mxc> | botje - it is sexy. |
| 11:01:13 | <lilac> | ivanm: you want parse :: String -> ([Int], Maybe Char) ? |
| 11:01:37 | <ivanm> | no, I want parse :: String -> Maybe (String, String) |
| 11:01:54 | <lilac> | what are the two strings? |
| 11:01:55 | <ivanm> | though it _could_ be using ([Int], Maybe Char) as the first STring... |
| 11:02:01 | <lilac> | oh, ok :) |
| 11:02:40 | <ivanm> | the second String is the rest of the initial String |
| 11:03:06 | <ivanm> | so Reads String, but replacing [] with Maybe (well, I'm still using [] at the moment, but will probably switch) |
| 11:09:24 | <Baughn> | @tell conal http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5121#a5121 <-- Here, have another bug. I'm going away for the weekend. |
| 11:09:25 | <lambdabot> | Consider it noted. |
| 11:09:59 | <raji> | how do we write the set of regular expression that doesn't contain 110 as substring , i know the answer as (0|10)*1* , how to derive the answer ? |
| 11:10:47 | <ivanm> | Baughn: how considerate of you ;-) |
| 11:11:34 | <kau> | hello! why join (++) $ "to" doesn't works? |
| 11:11:56 | <Baughn> | > join (++) "to" |
| 11:11:56 | <mauke> | > join (++) $ "to" |
| 11:11:58 | <lambdabot> | "toto" |
| 11:11:58 | <lambdabot> | "toto" |
| 11:12:02 | <kau> | says me No instance for (Monad ((->) [Char])) |
| 11:12:10 | <mauke> | kau: import Control.Monad.Instances |
| 11:12:16 | <kau> | ok thanks! |
| 11:13:52 | <Baughn> | ivanm: My motto is "To serve Man". |
| 11:13:54 | <kau> | is there a function to upper case? |
| 11:13:59 | <nominolo> | is there any way to wait on a file to be filled? I only get eof exceptions |
| 11:14:02 | <dobblego> | > toUpper 'a' |
| 11:14:04 | <lambdabot> | 'A' |
| 11:14:30 | <mauke> | nominolo: manually |
| 11:14:32 | <nominolo> | without polling, of course |
| 11:14:42 | <nominolo> | mauke: ? |
| 11:14:44 | <mauke> | depends on the platform |
| 11:14:49 | <mauke> | linux has inotify |
| 11:14:49 | <Baughn> | nominolo: popen "tail -f" .. |
| 11:14:53 | <mauke> | tail -f pools |
| 11:14:55 | <mauke> | er, polls |
| 11:15:06 | <Baughn> | Hm? That's.. |
| 11:15:22 | <nominolo> | all i want is to open a named pipe for clients to talk to my deamon |
| 11:15:23 | <mmorrow_> | inotify is pretty nice |
| 11:15:42 | <Baughn> | So it does. Huh, I figured it'd use whatever the OS had available. |
| 11:15:45 | <nominolo> | (I hate Vim for not having process state) |
| 11:15:46 | <mauke> | oh, so this is a pipe, not a normal file |
| 11:15:52 | <nominolo> | mauke: right |
| 11:16:16 | <nominolo> | mauke: though, that shouldn't make a difference for the client, does it? |
| 11:16:30 | <nominolo> | (And I don't even use Vim!) |
| 11:16:38 | <mauke> | opening a fifo blocks until the other end is opened |
| 11:16:41 | <nominolo> | @seen MarcWeber |
| 11:16:41 | <lambdabot> | I saw MarcWeber leaving #haskell 4m 4d 2h 3s ago, and . |
| 11:16:59 | <mauke> | but what do you mean by "filled"? |
| 11:17:01 | <Baughn> | *Closing* a FIFO produces sigpipe on the other end |
| 11:17:15 | <mauke> | incomplete |
| 11:17:32 | <nominolo> | Ok, let's start with the problem I'm trying to solve: |
| 11:17:37 | <mauke> | closing the read end of a fifo produces SIGPIPE if someone tries to write to it |
| 11:17:45 | <lilac> | > map reads ["123","123.","123.4"] :: [[(Int,String)]] |
| 11:17:47 | <lambdabot> | [[(123,"")],[(123,".")],[]] |
| 11:17:47 | <lilac> | ^^ WTF? |
| 11:18:10 | <nominolo> | I'm trying to get Vim to communicate with a Scion server (which needs to remember some state) |
| 11:18:19 | <lilac> | is that really the correct behaviour for Read Int? |
| 11:18:32 | <nominolo> | But Vim (without using Python) cannot keep a reference to a process |
| 11:18:35 | <mmorrow> | nominolo: could you just use a tcp socket and listen on loopback? |
| 11:18:52 | <Baughn> | nominolo: Or a unix domain socket |
| 11:19:07 | <dobblego> | @type isPrefixOf |
| 11:19:08 | <nominolo> | so I want to have a small script that connects to the server whenever the vim user wants to do something that needs communicating with the server |
| 11:19:09 | <lambdabot> | forall a. (Eq a) => [a] -> [a] -> Bool |
| 11:19:18 | <lilac> | > let parse xs = do (v, ys) <- reads xs; case ys of { '!':zs -> do { (vs,t,rest) <- parse zs; return (v:vs,t,rest) }; z:zs | isLower z -> return (v:[],Just z,zs); _ -> return (v:[],Nothing,ys) } in parse "123!4456lxyq" :: [([Int],Maybe Char,String)] |
| 11:19:20 | <lambdabot> | [([123,4456],Just 'l',"xyq")] |
| 11:19:20 | <nominolo> | Baughn: right, domain sockets are secure |
| 11:19:34 | <nominolo> | but I had some problems getting quick startup times on Windows at least |
| 11:19:36 | <Baughn> | nominolo: ..well, depending on the OS |
| 11:19:41 | <nominolo> | it should connect within milliseconds |
| 11:19:47 | <Baughn> | nominolo: Domain sockets are *not* secure on all operating systems |
| 11:20:00 | <nominolo> | Baughn: right, but BSD sockets are even less so |
| 11:20:15 | <Baughn> | nominolo: Option 2: ~/.vim.cookie |
| 11:20:19 | <Eelis> | is there a reason why top-level if-blocks aren't supported in a cabal file? i want to make an optional executable that's only built/installed if a flag was passed. |
| 11:20:27 | <nominolo> | Baughn: hm, what is that? |
| 11:20:40 | <mmorrow> | nominolo: graphviz comes with a little handy script that starts vim with the given file, and opens a viewer window, then uses inotify to wait for changes to the file, and every time it sees one it updates the view of the graph |
| 11:20:43 | <Baughn> | nominolo: A bunch of random data your server writes that your client has to supply to be allowed in. |
| 11:21:08 | <ivanm> | Baughn: lol |
| 11:21:10 | <lilac> | @src Read Int |
| 11:21:10 | <lambdabot> | Source not found. My mind is going. I can feel it. |
| 11:21:24 | <nominolo> | Baughn: right, yeah, something like that. some sort of shared secret |
| 11:21:41 | <nominolo> | mmorrow: interesting. have a link? |
| 11:22:07 | <nominolo> | mmorrow: oh, inotify is Linux only? |
| 11:22:16 | <Baughn> | ivanm: Ah, I was wondering if you understood. |
| 11:22:25 | <mmorrow> | nominolo: the shell script is super simple (and you probably have it if you have graphviz), but it's http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5122#a5122 |
| 11:22:30 | <mmorrow> | "vimdot" |
| 11:22:38 | <mmorrow> | nominolo: yeah, linux only unfortunately |
| 11:23:38 | <ivanm> | Baughn: was afk for a bit |
| 11:24:10 | <ivanm> | mmorrow: presumably it would work in BSD if the BSD kernel had support for it... |
| 11:25:03 | <nominolo> | the other option is to require python support for Vim. But already MacVim doesn't include a recent enough version... :/ |
| 11:25:58 | <mauke> | require perl support instead!! |
| 11:27:16 | <nominolo> | mauke: is that more widespread? |
| 11:27:36 | <mauke> | no (I just like perl better) |
| 11:27:44 | <nominolo> | mauke: I know that in general perl is more widely available than python |
| 11:27:53 | <BONUS> | ?umtl StateT s (WriterT w []) a |
| 11:27:53 | <lambdabot> | s -> [(a, s, w)] |
| 11:28:07 | <BONUS> | isnt this also supposed to be s -> [((a,s),w)] |
| 11:28:18 | <nominolo> | BONUS: it's isomorphic |
| 11:28:29 | <BONUS> | yeah that's true |
| 11:28:36 | <mmorrow> | (modulo a few undefined's) |
| 11:28:44 | <nominolo> | mmorrow: right, as always |
| 11:28:55 | <BONUS> | ?umtl WriterT w (ContT r []) a |
| 11:28:56 | <lambdabot> | (a -> w -> [r]) -> [r] |
| 11:29:01 | <BONUS> | but is this isomorphic? |
| 11:29:04 | <mauke> | yes |
| 11:29:08 | <BONUS> | to ((a,w) -> [r]) -> [r] |
| 11:29:09 | <mauke> | see uncurry |
| 11:29:13 | <BONUS> | ah of course |
| 11:29:18 | <nominolo> | BONUS: module strictness |
| 11:29:26 | <nominolo> | modulo* |
| 11:29:48 | <Baughn> | ivanm: But you caught the reference. One of the greatest works of anti-rationalist propaganda of all time.. |
| 11:30:11 | <ivanm> | Baughn: ummm.... you were quoting something? |
| 11:30:19 | <ivanm> | if so, I didn't catch the reference... |
| 11:30:24 | <Baughn> | ivanm: The silver chair |
| 11:30:52 | <ivanm> | never heard of it |
| 11:31:07 | <Baughn> | ivanm: "To serve man" is written above the main entrance to a giant's mansion. The punchline is, of course, "for dinner". |
| 11:31:25 | <ivanm> | oh, right, Narnia |
| 11:31:35 | <ivanm> | didn't recall it being that exact phrase... |
| 11:32:32 | <Baughn> | I wonder how C.S. Lewis would handle being one of the most hated authors in europe? |
| 11:32:51 | <sioraiocht> | Baughn: self-righteously? |
| 11:32:59 | <Baughn> | sioraiocht: Ah. Probably. |
| 11:36:32 | <Adamant> | who hates C.S. Lewis? |
| 11:36:47 | <ivanm> | Adamant: atheists |
| 11:36:59 | <Baughn> | Adamant: And rationalists |
| 11:37:01 | <ivanm> | IIRC, the golden compass, etc. was written directly as an anti-Narnia series |
| 11:37:09 | <Adamant> | ivanm: lol yeah |
| 11:37:18 | <Adamant> | that's why I can't take it seriously |
| 11:37:30 | <Adamant> | nobody even reads C.S. Lewis in the first place |
| 11:37:46 | <dolio> | I read all but one. |
| 11:37:49 | <p_l> | narnia isn't that bad, I guess it's more of the hype that surrounds it... hype turning a self-fueling hate machine... |
| 11:37:53 | <Adamant> | to get fired up enough about his writing to make a series of crappy fantasy novels.... |
| 11:38:32 | <dolio> | That is, all but one Narnia book. I've never bothered reading his apologetics or anything, because I hear they're terrible. |
| 11:39:11 | <p_l> | ACTION hears about C.S.Lewis being so hated for the first time |
| 11:39:22 | <decaf> | p_l: me too |
| 11:39:31 | <Baughn> | Adamant: I read cs lewis. I didn't realize it was propaganda at the time, though. :P |
| 11:39:43 | <Adamant> | Baughn: everything's propoganda |
| 11:39:50 | <Baughn> | Adamant: That said, the golden compass is simply a lot better - as a book, not as propaganda. I think it fails in the latter sense.. |
| 11:39:56 | <Adamant> | people are trying to convince you of their view |
| 11:40:12 | <Adamant> | Baughn: I don't like Narnia but Golden Compass is even worse |
| 11:40:22 | <Baughn> | No. There are plenty of works of fiction that are nothing more than works of fiction, with no author tracts mixed in. |
| 11:40:47 | <Adamant> | Baughn: name one. they're influenced by an author's beliefs and worldview |
| 11:40:50 | <p_l> | ACTION liked Narnia even more when he noticed parallels with Bible, having read Revelation quite a lot due to NGE... |
| 11:41:20 | <O_4> | NGE? |
| 11:41:24 | <p_l> | Baughn: They can still function as propaganda |
| 11:41:28 | <Baughn> | Neon genesis evangelion |
| 11:41:35 | <O_4> | Ah |
| 11:42:04 | <Baughn> | p_l, adamant: I don't think a book like.. hm, let's say Horatio Hornblower |
| 11:42:18 | <Baughn> | It's written to /entertain/, that's all |
| 11:42:19 | <p_l> | Baughn: Oh, it works very very well as propaganda :D |
| 11:42:20 | <Adamant> | Baughn: lol |
| 11:42:28 | <Adamant> | as an American, bullshit |
| 11:42:42 | <decaf> | a fairy tale with satirs is christian propaganda? |
| 11:42:43 | <Adamant> | Horatio takes out a US ship during the War of 1812 |
| 11:42:43 | <Baughn> | p_l: How's that, anti-oldtimes propaganda? |
| 11:42:54 | <Baughn> | decaf: Aslan = Jesus. Take it from there. |
| 11:42:55 | <p_l> | Baughn: Actually books that aren't written directly as propaganda give me that effect the most :P |
| 11:43:12 | <decaf> | Baughn: aslan means lion in turkish, take this one :) |
| 11:43:12 | <Adamant> | in one of his books and it's portrayed as a good thing |
| 11:43:28 | <Baughn> | decaf: That doesn't surprise me a bit |
| 11:43:31 | <p_l> | Adamant: Of course it is :3 |
| 11:43:36 | <Baughn> | Adamant: He's /british/ |
| 11:43:53 | <Adamant> | Baughn: but it doesn't help your propoganda claim |
| 11:44:02 | <Adamant> | in fact it reinforces min |
| 11:44:04 | <Adamant> | e |
| 11:44:05 | <p_l> | decaf: I'd say... just last book of Narnia |
| 11:44:17 | <Baughn> | Adamant: Perhaps it'd work as propaganda if I were british myself |
| 11:44:53 | <Baughn> | (But probably not. I'm.. not anti-nation, but at least rather neutral towards the concept.) |
| 11:44:56 | <p_l> | Baughn: What about seeing the actions of characters and thinking to yourself "now that's something I'd like to see IRL"? |
| 11:45:24 | <Baughn> | p_l: Then that isn't changing your ideas, it's just highlighting an existing one |
| 11:46:00 | <Baughn> | decaf: Oh, and the silver chair - the book explicitly states hat wishful thinking is preferable to reality. |
| 11:46:17 | <p_l> | Baughn: Or changing the bias... quite possibly you won't meet anything new... |
| 11:48:35 | <decaf> | Baughn: I'm very sorry to tell this: if that tale is a christian propaganda, that proof that christians are as 'smart' as my people (muslum turks) |
| 11:49:14 | <Baughn> | decaf: Mm. Hate to break it to you, but theists are equally irrational regardless of religion. |
| 11:49:33 | <Baughn> | Some may make it more visible, but overall.. *pout* |
| 11:49:44 | <p_l> | afaik narnia was written for tolkien... wasn't it some kind of wager, btw? |
| 11:50:44 | <Baughn> | 'Tolkien disliked the first Narnia book, published in 1950, telling the author it had too many clashing elements and was pushing the Christianity "message" too far. He also apparently "hated" Lewis's allegorical fight between good and evil, with Jesus represented by Aslan the Lion.' <-- Not exactly. |
| 11:51:17 | <Baughn> | And yet, Tolkien was the one who converted Lewis from atheist to christian. :/ |
| 11:51:35 | <p_l> | Baughn: Lewis writing it with Tolkien in mind doesn't clash with Tolkien not liking end result :P |
| 11:51:36 | <Baughn> | Converts are always the most fervent, I guess |
| 11:56:07 | <Baughn> | decaf: http://www.guardian.co.uk/books/2005/dec/05/cslewis.booksforchildrenandteenagers <-- Yeah, that about covers it. |
| 11:58:06 | <quicksilver> | Baughn: ERROR : Article too long. |
| 11:58:06 | <quicksilver> | :) |
| 11:58:55 | <decaf> | Baughn: I didn't read anything about lewis until a few minutes ago. |
| 12:03:41 | <ivanm> | quicksilver: you mean tl;dr ? :p |
| 12:15:01 | <FalconNL> | Does anyone know a way to take the log of large numbers? logBase 10 (10^308) == 308.0, but logBase 10 (10^309) == Infinity |
| 12:15:20 | <Baughn> | > 10^309 |
| 12:15:22 | <lambdabot> | 100000000000000000000000000000000000000000000000000000000000000000000000000... |
| 12:15:25 | <Baughn> | > 10^309 :: Double |
| 12:15:27 | <lambdabot> | Infinity |
| 12:15:31 | <mauke> | > logBase 10 (10^309) :: CReal |
| 12:15:32 | <lambdabot> | 309.0 |
| 12:15:50 | <Baughn> | > 10^309 :: CReal |
| 12:15:52 | <lambdabot> | 100000000000000000000000000000000000000000000000000000000000000000000000000... |
| 12:16:05 | <Baughn> | ..what's going on here? |
| 12:17:37 | <ivanm> | Baughn: what do you mean? |
| 12:17:45 | <ivanm> | CReal is computable reals (by roconnor?) |
| 12:17:50 | <Baughn> | Well, CReal shouldn't have greater range than.. |
| 12:18:00 | <Baughn> | Oh. So, not a Foreign.C type |
| 12:18:05 | <ivanm> | no |
| 12:18:31 | <mauke> | there is no 'real' type in C |
| 12:18:37 | <ivanm> | is there any sample usages or tutorials on how to use Text.ParseCombinators.ReadP? |
| 12:23:41 | <augustss> | CReal is based on David Lester's code. |
| 12:27:43 | <FalconNL> | Hm. If I call intLength (10^20000), where intLength n = succ . floor . logBase 10 $ (fromIntegral n :: CReal), it takes 1.5 seconds. Surely CReal can't be that slow? |
| 12:27:51 | <ivanm> | it is |
| 12:27:53 | <ivanm> | it's all functions |
| 12:28:02 | <ivanm> | rather than actually "storing" the number |
| 12:28:24 | <EvilTerran> | > exp (sqrt (-1) * pi) :: Complex CReal -- it is pretty cool, tho, even if it is slow |
| 12:28:25 | <lambdabot> | (-1.0) :+ 0.0 |
| 12:28:50 | <ivanm> | > exp (sqrt (-1) * pi) :: Complex Double |
| 12:28:52 | <lambdabot> | (-1.0) :+ 1.2246467991473532e-16 |
| 12:28:54 | <ivanm> | heh |
| 12:29:44 | <FalconNL> | Ok, let's try this another way then. What I'm trying to do is to determine the length of the decimal representation of a number in a faster way than length . show. Since the log-based approach doesn't seem to be working very well, does anyone know another way? |
| 12:30:07 | <ivanm> | any particular reason why you want to do so? |
| 12:30:32 | <ivanm> | and of what datatype are you finding said length? |
| 12:30:40 | <FalconNL> | for an algorithm to work with palindrome numbers for http://programmingpraxis.wordpress.com/2009/05/22/the-next-palindrome/ |
| 12:30:45 | <ivanm> | (since for a CReal, you can have infinite length...) |
| 12:30:51 | <ivanm> | > pi :: CReal |
| 12:30:53 | <lambdabot> | 3.1415926535897932384626433832795028841972 |
| 12:31:01 | <FalconNL> | Since the algorithm needs to work up to 10^1000000, probably Integers |
| 12:31:05 | <ivanm> | that's right... defaults to 20 decimal places or something... |
| 12:31:33 | <ivanm> | FalconNL: so by "decimal", you didn't mean "number of places after the decimal point"? |
| 12:31:52 | <quicksilver> | "decimal" means "representation in base 10" :P |
| 12:32:05 | <FalconNL> | no, I meant the amount of digits when shown in base 10, so 1001 -> 4, 17 -> 2, etc. |
| 12:32:06 | <EvilTerran> | FalconNL, binary search to find i such that 10^i <= x < 10^(i+1)? |
| 12:32:19 | <ivanm> | quicksilver: true, but I'm so used to having it mean "decimal places" when dealing with students, that that's what I automatically assume |
| 12:32:26 | <ivanm> | FalconNL: divide it by a really big number |
| 12:32:29 | <ivanm> | then log it |
| 12:32:44 | <ivanm> | something like that |
| 12:32:51 | <ivanm> | (Basically, you scale it down) |
| 12:33:02 | <quicksilver> | FalconNL: does (floor . logBase 10) for doubles not work well? |
| 12:33:18 | <quicksilver> | oh, 10^(10^6) |
| 12:33:23 | <quicksilver> | yeah that's too big for that :) |
| 12:33:36 | <EvilTerran> | it seems ugly to get floating point involved in a problem on integers |
| 12:33:40 | <FalconNL> | quicksilver: If you try it with doubles it fails at 10^309. It works fine for 10^308 |
| 12:33:59 | <quicksilver> | EvilTerran: it is ugly, but it's sometimes faster. In this case it simply fails. |
| 12:34:07 | <quicksilver> | EvilTerran's binary search sounds good. |
| 12:34:20 | <quicksilver> | writing a fast version of length . show is another possibility. |
| 12:36:39 | <EvilTerran> | another possibility would be looking at the GHC internals of Integer |
| 12:37:01 | <EvilTerran> | you might even be able to do it O(1) that way with a bit of fudging |
| 12:37:27 | <lilac> | FalconNL: why not do it with a string? |
| 12:37:35 | <quicksilver> | lilac: he said it was too slow. |
| 12:38:15 | <lilac> | quicksilver: i meant do the whole problem in a String and never go to a number representation |
| 12:38:17 | <FalconNL> | lilac: because length $ show $ 10^1000000 takes about 800 ms |
| 12:39:14 | <quicksilver> | ah nod. |
| 12:39:43 | <Baughn> | @tell conal I've isolated the bug in my previous paste to filterE. I will now proceed to pine for my laptop all weekend. |
| 12:39:43 | <lambdabot> | Consider it noted. |
| 12:50:52 | <byorgey> | ACTION makes lambda-rhubarb muffins |
| 12:51:04 | <Badger> | :) |
| 12:51:06 | <opqdonut> | how about curry-howard muffins |
| 12:51:18 | <byorgey> | those are good too |
| 12:51:21 | <Badger> | curried howard? |
| 12:51:36 | <dev31212> | hi quicksilver |
| 12:51:38 | <dev31212> | hi all |
| 12:51:53 | <byorgey> | hi dev31212 |
| 12:52:01 | <dev31212> | hy byorgey |
| 12:52:03 | <dev31212> | how goes it |
| 12:52:29 | <byorgey> | quite well, how are you? |
| 12:52:55 | <Baughn> | @tell conal Compare the existing definition of filterE with this: "filterE p m = justE $ (\i -> if p i then Just i else Nothing) <$> m". Neither /works/, but this one works /better/ - I suppose some odd laziness issues there, since they look identical to me. |
| 12:52:56 | <lambdabot> | Consider it noted. |
| 12:52:59 | <dev31212> | Im good thanks...I negotiated an easier work load, so I have more time to devote to my haskell web project |
| 12:53:06 | <dev31212> | less pay, but less stress and more time..so Im happuy |
| 12:53:14 | <gio123> | is here any german speakers? |
| 12:53:23 | <dev31212> | Im evaluating a framework called "Kibro" at the moment.. |
| 12:53:26 | <dev31212> | know anything about it? |
| 12:53:56 | <ivanm> | gio123: I would say that Germany is full of German speakers |
| 12:54:02 | <ivanm> | here, though, is a different story ;-) |
| 12:54:05 | <dev31212> | hi ivanm |
| 12:54:10 | <quicksilver> | Baughn: I discussed an alternative implementation of filterE with conal. |
| 12:54:10 | <ivanm> | hey dev31212 |
| 12:54:20 | <dev31212> | Jr parle un pue de le feancais, si-tu-plait |
| 12:54:23 | <quicksilver> | Baughn: it doesn't work for his implementation although it does for mine. |
| 12:54:33 | <quicksilver> | Baughn: might be interesting to you. |
| 12:54:35 | <dev31212> | je* |
| 12:54:56 | <dev31212> | hayoo kibro |
| 12:55:01 | <dev31212> | > hayoo kibro |
| 12:55:02 | <lambdabot> | Not in scope: `hayoo'Not in scope: `kibro' |
| 12:55:15 | <frwmanners> | Baughn: so, I came to the conclusion that no version of filterE could work with reactive-10.5's structure of time |
| 12:55:27 | <opqdonut> | @hackage kibro |
| 12:55:28 | <lambdabot> | http://hackage.haskell.org/cgi-bin/hackage-scripts/package/kibro |
| 12:55:32 | <frwmanners> | Baughn: as in, AddBounds (Improving Stuff) |
| 12:55:33 | <dev31212> | Thanks :) |
| 12:55:45 | <opqdonut> | :) |
| 12:55:47 | <dev31212> | Anyone know about kibro? I cant find a web page or a tutorial for it |
| 12:56:27 | <dev31212> | I am trying to avoid building my own web framework...but on the other hand that might be the best way for me to learn the language.. |
| 12:56:52 | <dev31212> | hmm |
| 12:57:00 | <dev31212> | I think I just made an argument for making my own framework |
| 12:57:37 | <ivanm> | so we get Yet Another Haskellian Web Framework? |
| 12:57:45 | <dev31212> | Yeah :) |
| 12:57:51 | <dev31212> | From a former PHP guy |
| 12:57:57 | <dev31212> | well current...hope to be former one day |
| 12:58:03 | <mauke> | ACTION runs away |
| 12:58:05 | <EvilTerran> | better write a monad tutorial (and library!) as well ;P |
| 12:58:05 | <dev31212> | lol |
| 12:58:19 | <boml> | mauke: what does this part do "use List::Util qw(reduce)" |
| 12:58:41 | <ivanm> | EvilTerran: what kind of library? |
| 12:58:48 | <boml> | he's gone. I'll just send it to everyone |
| 12:58:57 | <boml> | what does this mean? use List::Util qw(reduce) |
| 12:59:03 | <quicksilver> | Baughn: it's perl. |
| 12:59:06 | <quicksilver> | oops |
| 12:59:08 | <quicksilver> | boml: it's perl. |
| 12:59:23 | <boml> | yepp |
| 12:59:26 | <ivanm> | I know it's compulsory to write a monad tutorial (something I've been remiss in), but it's compulsory to write a library as well? :s |
| 13:01:35 | <EvilTerran> | ivanm, well, judging by the number of 'em on hackage |
| 13:01:53 | <ivanm> | aren't half of them by dons and mmorrow? |
| 13:01:53 | <ivanm> | ;-) |
| 13:02:04 | <EvilTerran> | "... Middleware (1), Monadic Regions (1), Monads (24), Music (11) ..." |
| 13:03:08 | <decaf> | 'google suggest' helps to type "What the hell are monads?" |
| 13:04:02 | <ray> | :t (>>=) |
| 13:04:03 | <lambdabot> | forall (m :: * -> *) a b. (Monad m) => m a -> (a -> m b) -> m b |
| 13:04:08 | <ray> | ^^ monad tutorial |
| 13:04:26 | <ivanm> | @quote endofunctors |
| 13:04:27 | <lambdabot> | psykotic says: [monads aren't hard] they're just monoids on the category of endofunctors over some category, what's the big deal? |
| 13:04:34 | <ivanm> | ^^ that's what monads are! |
| 13:04:57 | <EvilTerran> | in terms of general monad libs, there's MonadLab, monadLib, and mtl, at least; not to mention category-extras & co's shennanigans |
| 13:04:57 | <ivanm> | I was actually after the supposed wadler quote, but that'll do (since it's basically the same thing) |
| 13:05:10 | <ivanm> | does that mean that psykotic came up with that first? |
| 13:05:18 | <mauke> | @quote chocolate |
| 13:05:18 | <lambdabot> | Duqicnk says: a monad is like a train that runs backwards in time, which is made of tiny chocolate robots |
| 13:05:37 | <ivanm> | EvilTerran: IIRC, don't most people adovcate using monadlab or monadlib rather than mtl as they're much better libraries? |
| 13:05:56 | <ivanm> | and monad transformers are better defined/thought up nowadays than when mtl was written? |
| 13:06:05 | <ivanm> | mauke: O...K... then... |
| 13:07:10 | <EvilTerran> | ivanm, yeah, i think that's the case; not sure how m'lab and m'lib compare, tho |
| 13:07:18 | <jonte> | Hello boys and girls! I have a problem. I have a function returning (Maybe a, Splay a),it's called 'insert', and I wish to assign this return value to two variables (pattern match). I'm doing this: where; (e, t) = insert elem tree. What am I doing wrong? :) Thanks! |
| 13:07:53 | <quicksilver> | jonte: that's fine except for the ; after where |
| 13:08:03 | <mauke> | jonte: what's the error message? |
| 13:08:42 | <EvilTerran> | quicksilver, isn't a ; valid there if you're using explicit layout? |
| 13:08:48 | <jonte> | quicksilver, That'äs just for line breaking in the chat.. hmm.. The error I get is: Type error in right hand side, (Maybe c) doesn't match (a,b) |
| 13:08:48 | <quicksilver> | I'm not sure. |
| 13:09:03 | <quicksilver> | then your insert doesn't return (Maybe a,Splay a) |
| 13:09:09 | <quicksilver> | even though you claim it does :P |
| 13:09:14 | <quicksilver> | it returna (Maybe c) |
| 13:09:24 | <mauke> | maybe it returns Maybe (a, Splay a) |
| 13:09:31 | <EvilTerran> | > let { foo = bar where; bar = () } in foo |
| 13:09:33 | <lambdabot> | () |
| 13:09:35 | <tetha> | hehe, reasons to add explicit type signatures |
| 13:10:01 | <EvilTerran> | ^ you're allowed blank lines in layout blocks, so it makes sense that that's valid, really. |
| 13:10:05 | <jonte> | Hmm! I'll have to look in the library. it SHOULD return that. :) |
| 13:10:35 | <jonte> | Oh. Darn. You're right. :P Sorry to bother you guys. :) |
| 13:11:05 | <EvilTerran> | > () where {;;} -- listening to sad music in oversized headphones? |
| 13:11:06 | <lambdabot> | () |
| 13:11:38 | <tetha> | heh |
| 13:11:56 | <tetha> | or cthuhlu with headphones |
| 13:13:04 | <Japsu> | Every time you misspell fhtagn, Cthulhu eats a cultist. |
| 13:13:12 | <Japsu> | Please, think about the cultists. |
| 13:13:57 | <ivanm> | @type where |
| 13:13:58 | <lambdabot> | parse error on input `where' |
| 13:13:59 | <tetha> | since I like him and dislike cultists, I guess misspelling is a good thing |
| 13:14:04 | <ivanm> | duh |
| 13:18:28 | <EvilTerran> | @type () where {} |
| 13:18:29 | <lambdabot> | parse error on input `where' |
| 13:18:36 | <EvilTerran> | > () where {} -- rather |
| 13:18:38 | <lambdabot> | () |
| 13:19:50 | <ray> | > () where () |
| 13:19:52 | <lambdabot> | <no location info>: parse error (possibly incorrect indentation) |
| 13:20:02 | <ray> | therefore, haskell sucks |
| 13:20:07 | <ivanm> | :o |
| 13:20:09 | <ivanm> | @slap ray |
| 13:20:09 | <lambdabot> | ACTION beats up ray |
| 13:20:12 | <opqdonut> | :D |
| 13:20:23 | <EvilTerran> | > ({-) where (-}) |
| 13:20:24 | <lambdabot> | () |
| 13:20:38 | <ivanm> | EvilTerran: heh |
| 13:20:43 | <mauke> | > () where () = () |
| 13:20:44 | <lambdabot> | () |
| 13:20:48 | <ivanm> | ACTION has no idea wtf it's meant to do though... |
| 13:21:03 | <ivanm> | mauke: you can set () to equal something? :o |
| 13:21:06 | <ivanm> | > () where () = 1 |
| 13:21:07 | <lambdabot> | No instance for (Num ()) |
| 13:21:08 | <lambdabot> | arising from the literal `1' at <intera... |
| 13:21:10 | <mauke> | > 1 where 2 = 3 |
| 13:21:10 | <ivanm> | heh |
| 13:21:12 | <lambdabot> | 1 |
| 13:21:21 | <kpreid> | ivanm: no, it's a pattern match with no variables |
| 13:21:22 | <ray> | making it symmetrical with comments is cheating |
| 13:21:27 | <ivanm> | mauke: that works because 2 isn't evaluated/defined? |
| 13:21:28 | <mauke> | > let 2 + 2 = 5 in 2 + 2 |
| 13:21:29 | <lambdabot> | 5 |
| 13:21:32 | <ivanm> | kpreid: oh, yeah |
| 13:21:38 | <ivanm> | mauke: :o |
| 13:21:42 | <ivanm> | and how does _that_ work? |
| 13:21:51 | <quicksilver> | it redefines (+) |
| 13:21:52 | <ray> | it's defining + |
| 13:21:54 | <kpreid> | since it's not in a case or lambda, it doesn't even care if it matches |
| 13:21:58 | <quicksilver> | > let 2 + 2 = 5 in 2 + 3 |
| 13:22:00 | <lambdabot> | * Exception: /tmp/7791359399523594690:71:56-64: Non-exhaustive patterns in ... |
| 13:22:00 | <kpreid> | > () where False = True |
| 13:22:01 | <lambdabot> | () |
| 13:22:16 | <ivanm> | quicksilver: you can do that? :o |
| 13:22:18 | <quicksilver> | it's a bug in lambdabot that it accepts 'where' on a top level expression, by the way. |
| 13:22:22 | <Peaker> | @type let 2+2=5 in (+) |
| 13:22:23 | <ivanm> | @type let 2 + 2 = 5 in 2 + 2 |
| 13:22:24 | <lambdabot> | forall t t1 t2. (Num t2, Num t1, Num t) => t -> t1 -> t2 |
| 13:22:24 | <lambdabot> | forall t. (Num t) => t |
| 13:22:31 | <quicksilver> | ivanm: you can define any operator. You can shadow any variable. |
| 13:22:34 | <kpreid> | > let 2 + k = succ (succ k) in 2 + 2 + 2 |
| 13:22:35 | <lambdabot> | * Exception: /tmp/7038665918903841508:71:52-72: Non-exhaustive patterns in ... |
| 13:22:35 | <quicksilver> | ivanm: (+) is not magic. |
| 13:22:41 | <kpreid> | > let 2 + k = succ (succ k) in 2 + (2 + 2) |
| 13:22:43 | <lambdabot> | 6 |
| 13:22:48 | <ivanm> | quicksilver: yes, but it's already defined for the numeric types... |
| 13:23:08 | <EvilTerran> | quicksilver, i thought it was a deliberate feature, tbh. it's kinda useful. |
| 13:23:27 | <ray> | > let id _ = error "i shadowed id" in id () |
| 13:23:29 | <lambdabot> | * Exception: i shadowed id |
| 13:23:59 | <ivanm> | ACTION didn't know you could override/shadow top-level values/functions... |
| 13:24:07 | <EvilTerran> | and it would suggest a point to the "let v = ... in v" which has lead to shennanigans in the past |
| 13:24:10 | <EvilTerran> | @v |
| 13:24:10 | <lambdabot> | "\"\\\"\\\\\\\"\\\\\\" |
| 13:24:48 | <quicksilver> | ivanm: you can override shadow anything. |
| 13:25:10 | <EvilTerran> | well, apart from reserved words/symbols |
| 13:25:36 | <mauke> | > let pi = text "π" in pi |
| 13:25:37 | <lambdabot> | π |
| 13:25:54 | <doserj> | > let qualified = 3 in qualified |
| 13:25:54 | <lambdabot> | 3 |
| 13:26:00 | <doserj> | ACTION is cheating |
| 13:26:34 | <EvilTerran> | ACTION notes that the newbie error of "case ... of ...; otherwise -> ..." doesn't get caught because of this very shadowing |
| 13:26:58 | <ivanm> | EvilTerran: because otherwise becomes a variable that matches anything? |
| 13:27:02 | <ivanm> | rather than an alias for True? |
| 13:27:04 | <ivanm> | > otherwise |
| 13:27:06 | <lambdabot> | True |
| 13:27:11 | <Badger> | > let everything = 42 in everything |
| 13:27:13 | <lambdabot> | 42 |
| 13:27:13 | <ivanm> | @index otherwise |
| 13:27:13 | <lambdabot> | Data.Bool, Prelude |
| 13:27:25 | <EvilTerran> | ivanm, exactly, and then things get very odd if you try to use it in a guard within that case branch |
| 13:27:34 | <ivanm> | EvilTerran: heh |
| 13:29:00 | <EvilTerran> | ?type \x -> case x of Nothing -> "foo"; otherwise | fromJust x == "bar" -> "baz" | otherwise -> "eek" -- silly contrived example |
| 13:29:01 | <lambdabot> | Couldn't match expected type `Bool' |
| 13:29:01 | <lambdabot> | against inferred type `Maybe [Char]' |
| 13:29:01 | <lambdabot> | In a stmt of a pattern guard for |
| 13:30:04 | <cnwdup> | > let braces = between (string "{") (string "}"); expr = braces expr <|> many (noneOf "{}") in parse expr "" "parse me!" |
| 13:30:06 | <lambdabot> | Not in scope: `between'Not in scope: `string'Not in scope: `string'Not in s... |
| 13:30:22 | <cnwdup> | Mh, parsec's not in scope? ): |
| 13:30:26 | <quicksilver> | nope. |
| 13:30:34 | <cnwdup> | Anyways, can anyone explain why this results in a stack overflow: |
| 13:30:42 | <cnwdup> | let braces = between (string "{") (string "}"); expr = braces expr <|> many (noneOf "{}") in parse expr "" "{ }" |
| 13:30:51 | <cnwdup> | I do not see why. |
| 13:31:57 | <quicksilver> | cnwdup: it yields 'Right " "' for me. |
| 13:32:25 | <cnwdup> | Ups. I meant parse (many expr "" "{ }". |
| 13:32:42 | <mauke> | many/many |
| 13:33:06 | <quicksilver> | (many expr) |
| 13:33:06 | <quicksilver> | ? |
| 13:33:07 | <mauke> | you now have an infinite sequence of zero-length matches |
| 13:33:13 | <cnwdup> | Ah. Okay. |
| 13:33:14 | <EvilTerran> | cnwdup, well, many can match the empty string |
| 13:33:28 | <EvilTerran> | ACTION pokes his lag-o-meter |
| 13:33:31 | <mauke> | ACTION recommends Mastering Regular Expressions :-) |
| 13:33:47 | <Berengal> | Empty strings are evil |
| 13:33:54 | <Berengal> | Don't match them |
| 13:34:32 | <maltem> | mauke, sure, and then they're going to match expressions with matching braces |
| 13:35:00 | <mauke> | not a problem |
| 13:35:44 | <EvilTerran> | cnwdup, i would suggest braces = between (char '{') (char '}'); expr = braces expr <|> many1 (noneOf "{}") |
| 13:35:48 | <EvilTerran> | (i think it's called many1) |
| 13:36:22 | <cnwdup> | EvilTerran, yeah. I ended up with expr = braces (many1 expr) <|> many1 (noneOf "{}"). Thank you guys. |
| 13:36:33 | <rubendv> | is it possible to implement fromIntegral in regular Haskell? |
| 13:36:43 | <mauke> | many1/many1 |
| 13:36:53 | <EvilTerran> | mauke, ? |
| 13:36:57 | <mauke> | infinite backtracking avoided thanks to parsec's early commit |
| 13:36:58 | <doserj> | @src fromIntegral |
| 13:36:59 | <lambdabot> | fromIntegral = fromInteger . toInteger |
| 13:37:30 | <EvilTerran> | <mauke> many/many <mauke> many1/many1 <- IDGI |
| 13:37:41 | <rubendv> | how about fromInteger and toInteger then? |
| 13:37:56 | <mauke> | my $expr = qr/\{(?0)\}|[^{}]*/; |
| 13:38:15 | <EvilTerran> | rubendv, if the type you're writing an instance for is in regular haskell, yes |
| 13:38:18 | <mauke> | EvilTerran: "many applied to many applied to something" |
| 13:38:24 | <mauke> | i.e. direct nesting |
| 13:38:41 | <rubendv> | EvilTerran: so it is not possible for builtin types like Int |
| 13:38:57 | <EvilTerran> | rubendv, well, seeing as they're not "regular haskell"... |
| 13:39:09 | <EvilTerran> | although they could be defined as such, seeing as they're abstract |
| 13:39:15 | <rubendv> | I see |
| 13:39:29 | <rubendv> | I was just wondering about that, thank you. |
| 13:40:33 | <EvilTerran> | ?src Bool |
| 13:40:33 | <lambdabot> | data Bool = False | True deriving (Eq, Ord) |
| 13:40:39 | <EvilTerran> | rubendv, note Bool is "regular haskell" |
| 13:41:20 | <EvilTerran> | so not all Prelude types are "built in" by this definition |
| 13:41:39 | <ivanm> | I notice here that regex-{base,posix,compat} are included in the ghc documentation: http://www.haskell.org/ghc/docs/latest/html/libraries/index.html |
| 13:41:40 | <maltem> | mauke, humph. ok, if you read "regular expression" as "not-so-regular regular expression", then sure, no problem |
| 13:41:50 | <ivanm> | is this part of extralibs or something, because on my local install they're not... |
| 13:42:04 | <rubendv> | So fromInteger and toInteger are functions you have to define when making an Integral instance? |
| 13:42:12 | <mauke> | maltem: pretty much all regexes are irregular |
| 13:42:15 | <ivanm> | @src Integral |
| 13:42:15 | <lambdabot> | class (Real a, Enum a) => Integral a where |
| 13:42:15 | <lambdabot> | quot, rem, div, mod :: a -> a -> a |
| 13:42:15 | <lambdabot> | quotRem, divMod :: a -> a -> (a,a) |
| 13:42:15 | <lambdabot> | toInteger :: a -> Integer |
| 13:42:23 | <ivanm> | rubendv: fromInteger is in Num |
| 13:42:30 | <ivanm> | toInteger is in INtegral |
| 13:43:03 | <maltem> | mauke, I hoped that if there were a place for straight terminology, it would be #haskell :)) |
| 13:43:22 | <rubendv> | I see |
| 13:44:05 | <rubendv> | Makes sense :), thank you. |
| 13:45:47 | <Spockz|lap> | test |
| 13:46:29 | <ivanm> | sorry, but your test failed due to insufficient input |
| 13:46:49 | <Spockz|lap> | darn, sorry. I thought it would only run on the active server. |
| 13:48:08 | <ivanm> | heh |
| 13:58:10 | <lilac> | FalconNL: how fast does this palindrome thing have to be? |
| 13:59:00 | <lilac> | FalconNL: is 1.066s for 1000000 digits quick enough? |
| 13:59:03 | <ivanm> | lilac: you do realise he was asking about 1.5 hrs ago, don't you? |
| 13:59:13 | <lilac> | ivanm: yes, but i've been at lunch :) |
| 13:59:17 | <FalconNL> | lilac: there's no real requirement, but the faster the better. I'm currently reworking my algorithm a bit to make it faster |
| 13:59:18 | <ivanm> | lilac: IIRC, he was after 10^309 digits... |
| 13:59:19 | <ivanm> | heh |
| 14:01:50 | <skorpan> | can a logical predicate take two arguments? |
| 14:01:57 | <lilac> | the universe doesn't have enough space to store a number with 10^309 digits... |
| 14:01:58 | <skorpan> | e.g. P(x, y) |
| 14:02:32 | <FalconNL> | I only need 10^6 digits, which is 10^(10^6). logBase fails at 10^309 |
| 14:03:50 | <lilac> | yep, my approach works for 10^6 digits |
| 14:04:08 | <opqdonut> | skorpan: yes, usually |
| 14:04:39 | <skorpan> | could this function be written any prettier? http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5126#a5126 |
| 14:05:03 | <skorpan> | i tried to rewrite it using a fold, but failed |
| 14:05:38 | <opqdonut> | skorpan: it's pretty nice as it is |
| 14:06:05 | <skorpan> | i usually try to avoid helper functions like that one and it sure feels pretty foldy |
| 14:06:05 | <opqdonut> | you could turn the case into guards for the duplicatesBy' definition |
| 14:06:13 | <FalconNL> | liliac: would you mind showing your solution? Obviously you're doing something faster than I currently am |
| 14:06:24 | <Twey> | skorpan: duplicatesBy f = groupBy f . sortBy (comparing f) |
| 14:06:26 | <opqdonut> | if you wrote that as a fold, it'd have to be a fold over the tails of a list |
| 14:06:27 | <Twey> | No? |
| 14:06:46 | <opqdonut> | Twey: yeah but that needs an ordering |
| 14:06:50 | <Twey> | Or, wait, probably not comparing f |
| 14:06:56 | <skorpan> | Twey: comparing puts Ord |
| 14:06:57 | <Twey> | Yeah |
| 14:06:58 | <skorpan> | yeah |
| 14:06:59 | <skorpan> | :) |
| 14:07:09 | <opqdonut> | and it'd also need a (head . filter (\x -> length x >= 1)) |
| 14:07:35 | <Twey> | a.k.a. (head . filter (not . null)) |
| 14:07:55 | <opqdonut> | gah, i meant length x > 1 |
| 14:08:05 | <Twey> | Ah |
| 14:08:06 | <Twey> | Yeah |
| 14:08:13 | <paper_cc> | well... nubBy f xs = map fst $ filter (not . uncurry f) $ zip xs (tail xs), but it fails on xs == [] |
| 14:08:15 | <skorpan> | any other suggestions? |
| 14:08:32 | <opqdonut> | hmm, duplicatesBy ys = [x | (x:xs) <- tails ys, isJust $ find (p x) xs] |
| 14:08:37 | <Twey> | @index [a] -> [[a]] |
| 14:08:37 | <lambdabot> | bzzt |
| 14:08:37 | <paper_cc> | skorpan: ^^above |
| 14:08:44 | <Twey> | @hoogle [a] -> [[a]] |
| 14:08:45 | <lambdabot> | Data.List inits :: [a] -> [[a]] |
| 14:08:45 | <lambdabot> | Data.List tails :: [a] -> [[a]] |
| 14:08:45 | <lambdabot> | Data.List groupBy :: (a -> a -> Bool) -> [a] -> [[a]] |
| 14:08:57 | <Twey> | Hmm |
| 14:09:01 | <skorpan> | > let duplicatesBy p ys = [x | (x:xs) <- tails ys, isJust $ find (p x) xs] in duplicatesBy (==) [1,2,3,1,2] |
| 14:09:03 | <lambdabot> | [1,2] |
| 14:09:08 | <skorpan> | > let duplicatesBy p ys = [x | (x:xs) <- tails ys, isJust $ find (p x) xs] in duplicatesBy (==) [1,2,3,1,2,4,2] |
| 14:09:10 | <lambdabot> | [1,2,2] |
| 14:09:18 | <skorpan> | hm, this is not right |
| 14:09:35 | <opqdonut> | skorpan: same behaviour as your function |
| 14:09:36 | <skorpan> | my original function does the wrong thing... |
| 14:09:42 | <opqdonut> | multpile duplicates... yeah |
| 14:09:54 | <skorpan> | no, my function doesn't give multiple duplicates |
| 14:09:55 | <paper_cc> | @ty \f xs -> map fst $ filter (uncurry f) $ zip xs (tail xs) -- this is wrong the same way |
| 14:09:57 | <lambdabot> | forall a. (a -> a -> Bool) -> [a] -> [a] |
| 14:10:02 | <skorpan> | the problem is that i want *all* duplicates! |
| 14:10:06 | <skorpan> | and yours doesn't do that either :) |
| 14:10:24 | <skorpan> | so dupsBy (==) [1,2,3,1,2] would give me [1,2,1,2] |
| 14:10:29 | <opqdonut> | yeah |
| 14:10:55 | <skorpan> | (this is for a little verifier returning any multiple function declarations) |
| 14:11:24 | <skorpan> | so i need to be able to provide any predicate and have all the duplicates |
| 14:11:38 | <FalconNL> | lilac: would you mind showing your solution? Obviously you're doing something faster than I currently am |
| 14:12:36 | <skorpan> | i think another "case" does it |
| 14:12:45 | <skorpan> | yes, it does. |
| 14:13:00 | <skorpan> | updated ugly version: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5127#a5127 |
| 14:13:37 | <paper_cc> | > inits [1..5] |
| 14:13:39 | <lambdabot> | [[],[1],[1,2],[1,2,3],[1,2,3,4],[1,2,3,4,5]] |
| 14:13:40 | <byorgey> | skorpan: doesn't that return the duplicates in reverse order? |
| 14:13:44 | <byorgey> | or maybe that doesn't matter |
| 14:13:45 | <paper_cc> | > tails [1..5] |
| 14:13:47 | <lambdabot> | [[1,2,3,4,5],[2,3,4,5],[3,4,5],[4,5],[5],[]] |
| 14:14:01 | <skorpan> | byorgey: doesn't really matter, but ideally i'd want them in the right order |
| 14:14:06 | <skorpan> | i guess reverse would do that |
| 14:14:14 | <byorgey> | yeah, just stick in a call to reverse |
| 14:14:34 | <rubendv> | skorpan: looks like you could use the Maybe monad |
| 14:14:42 | <skorpan> | rubendv: i thought so too, but can't figure out how |
| 14:14:50 | <skorpan> | Nothing >>= _ = Nothing |
| 14:14:56 | <doserj> | > let dupsBy p ys = [b | n <- [0..(length ys) -1], let (as,b:bs) = splitAt n ys, any (p b) (as++bs)] in dupsBy (==) [1,2,3,2,4,2,1] |
| 14:14:58 | <lambdabot> | [1,2,2,2,1] |
| 14:15:09 | <rubendv> | you could redefine it perhaps? |
| 14:15:23 | <byorgey> | off the top of my head, I don't think you can use the Maybe monad |
| 14:15:23 | <skorpan> | rubendv: what do you mean? |
| 14:15:31 | <byorgey> | because Nothing here doesn't indicate overall failure. |
| 14:15:36 | <skorpan> | indeed |
| 14:15:40 | <rubendv> | or is that something you can't do? |
| 14:15:40 | <byorgey> | it just means that element isn't a duplicate. |
| 14:16:23 | <byorgey> | rubendv: what, redefine >>= for Maybe? |
| 14:16:30 | <byorgey> | you can't really do that |
| 14:16:42 | <paper_cc> | > let splits xs = zipWith (++) (inits xs) (tail $ tails xs) in splits [1..4] |
| 14:16:43 | <byorgey> | you wouldn't really want to, anyway |
| 14:16:45 | <lambdabot> | [[2,3,4],[1,3,4],[1,2,4],[1,2,3]] |
| 14:17:51 | <rubendv> | and what if you define >>= for a Maybe SomeDataType? |
| 14:18:20 | <skorpan> | that doesn't make sense to me |
| 14:18:46 | <skorpan> | this function is intended to be somewhat generic |
| 14:18:55 | <quicksilver> | that is mplus |
| 14:18:58 | <quicksilver> | in the Maybe monad |
| 14:19:03 | <quicksilver> | > Nothing `mplus` Just 3 |
| 14:19:05 | <lambdabot> | Just 3 |
| 14:19:08 | <skorpan> | ah |
| 14:19:28 | <doserj> | > let dupsBy p ys = [b | (as,b:bs) <- init $ zip (inits ys) (tails ys), any (p b) (as++bs)] in dupsBy (==) [1,2,3,2,4,3,2,1] |
| 14:19:30 | <doserj> | > let dupsBy p ys = [b | (as,b:bs) <- init $ zip (inits ys) (tails ys), any (p b) (as++bs)] in dupsBy (==) [1,2,3,2,4,3,2,1] |
| 14:19:32 | <lambdabot> | [1,2,3,2,3,2,1] |
| 14:19:50 | <davidL> | does this data surprise anyone? http://web.cecs.pdx.edu/~lazard/chart.png Why would I use one over another for dumping a [Word8] to a file? |
| 14:20:21 | <skorpan> | > let dupsBy p ys = [b | (as,b:bs) <- init $ zip (inits ys) (tails ys), any (p b) (as++bs)] in dupsBy (==) [] |
| 14:20:22 | <lambdabot> | [] |
| 14:20:26 | <skorpan> | > let dupsBy p ys = [b | (as,b:bs) <- init $ zip (inits ys) (tails ys), any (p b) (as++bs)] in dupsBy (==) [1,2,1] |
| 14:20:27 | <lambdabot> | [1,1] |
| 14:20:30 | <skorpan> | interesting... |
| 14:21:02 | <quicksilver> | but it's not obvoious the best way to write it |
| 14:21:14 | <skorpan> | certainly not |
| 14:21:21 | <quicksilver> | since in the case that you find it as "y", you use (x:found) |
| 14:21:23 | <quicksilver> | not (z:found) |
| 14:21:30 | <quicksilver> | so the two branches don't really have the same pattern |
| 14:21:36 | <skorpan> | i'll try doing something with mplus... |
| 14:22:18 | <skorpan> | i can't seem to find the definition of Maybe's mplus |
| 14:22:29 | <byorgey> | @src Maybe mplus |
| 14:22:30 | <lambdabot> | Nothing `mplus` ys = ys |
| 14:22:30 | <lambdabot> | xs `mplus` ys = xs |
| 14:22:35 | <opqdonut> | skorpan: how about something like: dupsBy p xs = filter (\x -> count (p x) xs > 1) xs |
| 14:22:35 | <skorpan> | okay, found it :) |
| 14:22:40 | <lilac> | FalconNL: my solution is posted as a comment on that blog |
| 14:22:50 | <skorpan> | let dupsBy p xs = filter (\x -> count (p x) xs > 1) xs in [1,2,1] |
| 14:22:53 | <skorpan> | > let dupsBy p xs = filter (\x -> count (p x) xs > 1) xs in [1,2,1] |
| 14:22:54 | <lambdabot> | Not in scope: `count' |
| 14:23:06 | <skorpan> | > let dupsBy p xs = filter (\x -> (length . filter) (p x) xs > 1) xs in [1,2,1] |
| 14:23:07 | <lambdabot> | Couldn't match expected type `[a]' |
| 14:23:12 | <skorpan> | hurmm... |
| 14:23:20 | <doserj> | > let dupsBy p ys = [b | (as,b:bs) <- init $zip (inits ys) (tails ys), any (p b) (as++bs)] in dupsBy ((==)`on`(`mod`10)) [1,11,21,2,12,3,4,5] |
| 14:23:21 | <lambdabot> | [1,11,21,2,12] |
| 14:23:34 | <byorgey> | (length.) . filter I think |
| 14:23:42 | <skorpan> | > let dupsBy p xs = filter (\x -> ((length .) . filter) (p x) xs > 1) xs in [1,2,1] |
| 14:23:44 | <lambdabot> | [1,2,1] |
| 14:23:50 | <paper_cc> | let splits xs = zipWith (++) (inits xs) (tail $ tails xs); dupsBy f xs = catMaybes $ zipWith (\x other -> find (f x) other) xs (splits xs) in dupsBy (==) [1,2,3,1,2] |
| 14:23:51 | <skorpan> | no cigar! |
| 14:23:54 | <paper_cc> | oh |
| 14:23:56 | <paper_cc> | > let splits xs = zipWith (++) (inits xs) (tail $ tails xs); dupsBy f xs = catMaybes $ zipWith (\x other -> find (f x) other) xs (splits xs) in dupsBy (==) [1,2,3,1,2] |
| 14:23:58 | <lambdabot> | [1,2,1,2] |
| 14:24:10 | <byorgey> | skorpan: um, look carefully at what you typed there =) |
| 14:24:13 | <davidL> | > let dupsBy p xs = filter (\x -> length (filter p xs) > 1) xs in dupBy (==) [1,2,1] |
| 14:24:14 | <lambdabot> | Not in scope: `dupBy' |
| 14:24:18 | <davidL> | > let dupsBy p xs = filter (\x -> length (filter p xs) > 1) xs in dupsBy (==) [1,2,1] |
| 14:24:19 | <lambdabot> | Couldn't match expected type `Bool' |
| 14:24:28 | <skorpan> | oh byorgey :) |
| 14:24:28 | <wicket> | I'm not entirely sure what kind of results you want from [1,2,3,1,2] |
| 14:24:30 | <opqdonut> | davidL: (p x) ... |
| 14:24:35 | <skorpan> | wicket: [1,2,1,2] |
| 14:24:46 | <davidL> | > let dupsBy p xs = filter (\x -> length (filter (p x) xs) > 1) xs in dupsBy (==) [1,2,1] |
| 14:24:47 | <lambdabot> | [1,1] |
| 14:25:00 | <skorpan> | > let dupsBy p xs = filter (\x -> length (filter (p x) xs) > 1) xs in dupsBy (==) [1,2,3,1,2] |
| 14:25:02 | <lambdabot> | [1,2,1,2] |
| 14:25:05 | <skorpan> | this is beautiful |
| 14:25:07 | <byorgey> | brilliant! |
| 14:25:13 | <opqdonut> | thank you :) |
| 14:25:17 | <skorpan> | thanks davidL, mind if i use this code in yi? :) |
| 14:25:18 | <byorgey> | opqdonut++ |
| 14:25:23 | <davidL> | what |
| 14:25:24 | <skorpan> | oh, what |
| 14:25:27 | <skorpan> | was that opqdonut? |
| 14:25:33 | <opqdonut> | sure, go ahead |
| 14:25:44 | <skorpan> | thanks |
| 14:25:48 | <doserj> | > let dupsBy p ys = [y | y <- ys, any (p y) (ys\\[y])] in dupsBy ((==)`on`(`mod`10)) [1,11,21,2,12,3,4,5] -- cheating a bit |
| 14:25:49 | <lambdabot> | [1,11,21,2,12] |
| 14:26:05 | <opqdonut> | ys\\[y], hng |
| 14:27:22 | <paper_cc> | ys \\ [y] requires Eq |
| 14:27:22 | <byorgey> | doserj: I don't think that's quite the same, is it? |
| 14:27:33 | <skorpan> | it's not |
| 14:27:34 | <doserj> | the nested filters are nice |
| 14:27:48 | <doserj> | I did say I cheated :) |
| 14:27:52 | <byorgey> | hehe |
| 14:28:05 | <opqdonut> | doserj: they capture the intent quite nicely, right |
| 14:28:19 | <opqdonut> | i find the > 1 a bit ugly though |
| 14:28:30 | <byorgey> | even better would be to make a function to test whether the length of a list is > 1 |
| 14:28:39 | <byorgey> | calling length and checking if its > 1 is not lazy enough |
| 14:28:57 | <mauke> | not . null . drop 1 |
| 14:29:44 | <opqdonut> | mauke: better i guess |
| 14:29:45 | <gwern> | @quote drug.*addict |
| 14:29:45 | <lambdabot> | No quotes match. Just what do you think you're doing Dave? |
| 14:29:54 | <gwern> | @remember seydar what's the nick of the drug addict who wrote learn you a haskell? and i mean that in the best possible way. |
| 14:29:55 | <lambdabot> | I will remember. |
| 14:30:09 | <gwern> | @quote String.*poor |
| 14:30:10 | <lambdabot> | No quotes match. |
| 14:30:22 | <gwern> | @remember roconnor String is kinda a poor data type for strings. |
| 14:30:23 | <lambdabot> | I will remember. |
| 14:30:33 | <gwern> | @quote beaten.*up |
| 14:30:33 | <lambdabot> | No quotes match. There are some things that I just don't know. |
| 14:30:44 | <gwern> | @remember kyevan I had a haskell-related dream last night. Sorta. I was beaten up by some kids because I tried to go somewhere my type didn't match, apparently. |
| 14:30:44 | <lambdabot> | Done. |
| 14:30:56 | <gwern> | @quote Windows.*98 |
| 14:30:56 | <lambdabot> | No quotes match. Have you considered trying to match wits with a rutabaga? |
| 14:31:05 | <gwern> | @remember edwardk Haskell 98 is the Windows 98 of standards ;) |
| 14:31:06 | <lambdabot> | Okay. |
| 14:31:12 | <gwern> | @quote endofunctor |
| 14:31:12 | <lambdabot> | psykotic says: [monads aren't hard] they're just monoids on the category of endofunctors over some category, what's the big deal? |
| 14:31:16 | <gwern> | @quote endofunctor |
| 14:31:16 | <lambdabot> | PhilipWadler says: I'm delighted to learn that "a monad is a monoid in the category of endofunctors"---anyone know where I can find a good tutorial? |
| 14:31:32 | <gwern> | @quote play.*monads |
| 14:31:32 | <lambdabot> | No quotes match. Sorry about this, I know it's a bit silly. |
| 14:31:42 | <gwern> | @remember DavidLeimbach Don't play with your monads... eventually you'll go bind. |
| 14:31:42 | <lambdabot> | Okay. |
| 14:31:45 | <gwern> | @flush |
| 14:32:45 | <weilawei> | lambdabot is on the ball today ^__^ |
| 14:33:50 | <opqdonut> | @quote |
| 14:33:51 | <lambdabot> | robreim says: Great, now we've got to start an IRC channel for lojban-speaking, haskell-coding anarchists |
| 14:34:16 | <weilawei> | @quote |
| 14:34:17 | <lambdabot> | SamB_XP says: oh. does A/S/L mean Age/Sex/Programming Language? |
| 14:34:24 | <weilawei> | xD i wish |
| 14:36:22 | <gnuvince> | I have a dumb question: some programming languages (Smalltalk, Clojure, Factor) provide abstract interfaces to their collection types, making it easy to switch from one to another. Haskell doesn't seem to have that, though I guess it could with a typeclass. Is the reason for that purely historic, or am I missing an important technical point? |
| 14:37:16 | <opqdonut> | i'd say historic |
| 14:37:33 | <solrize_> | foldable ? |
| 14:37:44 | <byorgey> | gnuvince: note this isn't about the languages themselves but about the standard libraries. |
| 14:37:45 | <opqdonut> | Functor :) |
| 14:38:26 | <byorgey> | gnuvince: I agree this is one area where the standard libraries could be improved somewhat. Although there are libraries like Edison. |
| 14:38:29 | <opqdonut> | gnuvince: a partial reason is that the usual kind of collection "interface" would require multi param typeclasses, fundeps and possibly associated types |
| 14:38:31 | <byorgey> | dunno if that's still maintained. |
| 14:39:11 | <gwern> | edison is maintained; just no one uses it |
| 14:39:27 | <opqdonut> | a shame |
| 14:40:28 | <quicksilver> | gnuvince: is it clear whether it's worth it? |
| 14:40:40 | <quicksilver> | the difference between the collection types is normally such that if you need one, you need that one. |
| 14:40:52 | <quicksilver> | the "lowest common denominator" of their interfaces is very small |
| 14:41:00 | <FalconNL_> | I don't suppose there's a faster version of read? (read $ show $ 10^1000000 :: Integer) is very slow. |
| 14:41:15 | <quicksilver> | whilst the corresponding upper bound contains all sorts of necessarily inefficient things. |
| 14:41:21 | <opqdonut> | you're not really supposed to use read for anything real |
| 14:41:36 | <solrize_> | ??? |
| 14:41:50 | <FalconNL_> | opqdonut: it's fast enough up to 10^10000 or so, but after that it becomes slow |
| 14:41:58 | <byorgey> | FalconNL_: yeah, read is rather slow |
| 14:42:05 | <lilac> | i wouldn't think of any situation where you have to read a million digit number as 'real' anyway... |
| 14:42:11 | <Berengal> | Why would you want to read such a number? |
| 14:42:14 | <opqdonut> | lilac: hehe |
| 14:42:30 | <byorgey> | if you want something faster you can roll your own with a fold over a bytestring |
| 14:42:37 | <opqdonut> | indeed |
| 14:42:42 | <solrize_> | why don't they just fix read? |
| 14:42:49 | <pejo> | lilac, read is slow for other things as well. D. Swierstra and some others had a paper at the haskell symposium last year on it. |
| 14:42:51 | <gwern> | wonder if data.binary would read just Integers faster |
| 14:43:28 | <lilac> | Berengal: FalconNL is attacking this: http://programmingpraxis.wordpress.com/2009/05/22/the-next-palindrome/ |
| 14:43:29 | <byorgey> | solrize_: I think part of the problem is that read is so general. for reading any particular thing a custom solution is likely to be faster. |
| 14:43:40 | <Berengal> | lilac: ah |
| 14:43:48 | <solrize_> | i don't get it. other languages have read-like functions that are fast. |
| 14:44:17 | <quicksilver> | solrize_: part of the reason is the underlying slowness of String |
| 14:44:18 | <quicksilver> | I suspect. |
| 14:44:20 | <lilac> | FalconNL_: for what it's worth, i think you should consider solutions which don't involve converting your string to a number and back |
| 14:44:22 | <quicksilver> | I don't know how much of it. |
| 14:45:15 | <byorgey> | yeah, that's a fun problem |
| 14:45:20 | <FalconNL_> | lilac: The speed difference between your and my version is in read. My Integer -> String algorithm runs slightly faster than yours. But ideally I'd like an Integer -> Integer algorithm ,since it's cleaner |
| 14:45:24 | <byorgey> | I seem to recall solving it once upon a time =) |
| 14:45:26 | <mauke> | hah |
| 14:45:30 | <quicksilver> | solrize_: a 1-byte-per-character string rep is always going to be faster than a 24-byte-per-character string rep :) |
| 14:45:31 | <mauke> | I solved PALIN in Perl |
| 14:45:48 | <byorgey> | I think I solved it in C =) |
| 14:46:00 | <mauke> | yeah, I also have a C solution |
| 14:46:37 | <pejo> | quicksilver, read is sometimes exponential: http://portal.acm.org/citation.cfm?id=1411286.1411296 |
| 14:47:12 | <lilac> | FalconNL_: I think I'd do an Integer -> Integer algorithm somewhat differently (not via String) |
| 14:47:59 | <quicksilver> | pejo: if you're going to paste a link, paste one that I can download from :P |
| 14:48:31 | <pejo> | quicksilver, ok, let me retry. |
| 14:48:36 | <pejo> | quicksilver, read is sometimes exponential. |
| 14:48:45 | <pejo> | quicksilver, :P |
| 14:49:00 | <quicksilver> | pejo: but, not in the case we are discussing. |
| 14:49:26 | <quicksilver> | pejo: only in the case of infix constructors, as far as I can tell. |
| 14:49:30 | <quicksilver> | so, not relevant here. |
| 15:05:48 | <mmorrow> | i was just looking at the jhc-generated C code for |
| 15:05:50 | <mmorrow> | fib :: Int -> Int -> [Int]; fib m n = m : fib n (m + n); main = (print . length) (fib 0 1) |
| 15:06:03 | <mmorrow> | and it's first building the entire list |
| 15:06:11 | <mmorrow> | then it runs over and finds the length |
| 15:07:03 | <mmorrow> | (and also, even though it wouldn't matter here, there's no gc) |
| 15:08:11 | <mmorrow> | and without something like this getting fused in some/any way, there can't be a gc |
| 15:08:42 | <mmorrow> | (which was why i was looking at the C code in the first place :( |
| 15:08:49 | <copumpki> | :o |
| 15:11:31 | <DarthMeh> | given an expression (test1 == test 2), if I want test1 and test2 to evaluate in parallel with par, how do I express this? |
| 15:11:34 | <mmorrow> | well, actually i should make sure that it doesn't magically fuse if the list is finite |
| 15:12:24 | <koeien> | DarthMeh: test1 `par` (test2 `pseq` (test1==test2)) iirc |
| 15:12:38 | <DarthMeh> | thank you, koeien |
| 15:12:46 | <pumpkin> | sounds like someone should define liftP2 (==) test1 test2 :P |
| 15:13:01 | <koeien> | yeah, could be a nice function ;) |
| 15:13:41 | <pumpkin> | (==) <$$> test1 <||> test2 |
| 15:14:02 | <pumpkin> | maybe? |
| 15:14:05 | <pumpkin> | ACTION shrugs |
| 15:17:56 | <mmorrow> | it doesnt |
| 15:18:03 | <pumpkin> | ? |
| 15:18:06 | <pumpkin> | oh, JHC |
| 15:18:10 | <mmorrow> | oh, about magically fusing |
| 15:18:14 | <mmorrow> | yeah |
| 15:18:30 | <mmorrow> | main = (print . length . maxBound) (fib 0 1) -- ;) |
| 15:18:35 | <mmorrow> | err |
| 15:18:38 | <mmorrow> | main = (print . length . take maxBound) (fib 0 1) -- ;) |
| 15:20:33 | <quicksilver> | mmorrow: You lost me. jhc is or isn't fusing? |
| 15:22:28 | <b_jonas> | hi |
| 15:22:57 | <pumpkin> | hi indeed! |
| 15:23:11 | <b_jonas> | can you use the data types a la carte thing to create a product type, say for a state so that you can read or alter any field without having to know what other fields there are? |
| 15:23:15 | <zsol> | yo |
| 15:23:35 | <quicksilver> | b_jonas: well it's not a magic solution. |
| 15:23:44 | <quicksilver> | you can obviously write something which is polymorphic |
| 15:23:50 | <quicksilver> | you could do that with a normal tuple too: |
| 15:23:58 | <quicksilver> | (Int,a) -> .... |
| 15:24:05 | <quicksilver> | ^^ ignores the "other field" a. |
| 15:24:19 | <quicksilver> | but it remains sensitive to the orders of the fields and stuff. |
| 15:25:08 | <roconnor> | quicksilver: does that solution use the class system for overloading? |
| 15:25:22 | <quicksilver> | the solution I just sketched doesn't do any overloading. |
| 15:25:28 | <roconnor> | ok |
| 15:25:34 | <quicksilver> | to make something cleverer you would indeed have to use the class system. |
| 15:26:07 | <quicksilver> | haskell's data types don't support this kind of extension, and data types a la carte isn't changing haskell in any way |
| 15:26:13 | <quicksilver> | it's just a particular approach. |
| 15:26:31 | <quicksilver> | (...don't support *polymorphism* over this kind of extension...) I should say |
| 15:26:44 | <quicksilver> | you can hack up various kinds of extensible record stuffs with type classes |
| 15:26:51 | <paper_cc> | is there a GMap library done with type families? (gmap-0.1 on Hackage uses FunDeps) |
| 15:26:51 | <quicksilver> | some of it more clumsy looking than others. |
| 15:27:24 | <b_jonas> | quicksilver: thanks |
| 15:27:35 | <b_jonas> | I'll have to read the data types a la carte article again |
| 15:27:45 | <b_jonas> | because I still don't really understand how it works |
| 15:27:55 | <b_jonas> | I have another question though |
| 15:28:17 | <b_jonas> | how do the extensible exceptions in ghc work? is it magic or is there an article about it somewhere? |
| 15:29:14 | <quicksilver> | b_jonas: it's not magic. |
| 15:29:34 | <quicksilver> | b_jonas: http://www.haskell.org/~simonmar/papers/ext-exceptions.pdf |
| 15:30:00 | <b_jonas> | quicksilver: thanks for the link |
| 15:31:37 | <mmorrow> | awesome. this however jhc runs in constant memory: |
| 15:31:48 | <mmorrow> | fib m n x k = k x (m::Int) (\x -> fib (n::Int) (m + n) x k); gogo acc _ k = let acc' = acc + (1::Int) in acc' `seq` k acc'; main = print (fib 0 1 (0::Int) gogo :: Int) |
| 15:32:24 | <mmorrow> | so it looks like if you just pretend you're programming in a strict language, you're golden ;) |
| 15:34:58 | <mmorrow> | quicksilver: i'm gonna make a short writeup paste, because that'll be infinitely more clear than me trying to describe the particulars |
| 15:38:46 | <quicksilver> | mmorrow: OK :) |
| 15:44:42 | <mmorrow> | hmm, i take back that it's running in constant space, since there doesn't seem to be any use of free() in the entire file, and it's malling malloc 4 times per loop... |
| 15:45:05 | <pumpkin> | :o |
| 15:45:16 | <mmorrow> | (i guess the massive eat-all-4GB-of-memory-in-under-a-second with the other program was from stack) |
| 15:46:18 | <FalconNL_> | lilac: I've posted my solution as well |
| 15:50:04 | <Elly> | mmorrow: rule 1 of malloc is the same as rule 1 of air travel |
| 15:50:16 | <Elly> | "Attempt at all costs to keep your number of landings equal to your number of takeoffs." |
| 15:52:28 | <weilawei> | Elly: that's an awesome way to put it. never heard that analogy before |
| 15:52:51 | <Elly> | heh, thanks :) |
| 15:53:23 | <pumpkin> | @remember Elly Rule 1 of malloc is the same as rule 1 of air travel: "Attempt at all costs to keep your number of landings equal to your number of takeoffs." |
| 15:53:23 | <lambdabot> | I will remember. |
| 15:53:27 | <FalconNL_> | Alternatively you could say: If you don't land after taking off, you'll crash. |
| 15:55:02 | <dmwit> | If you land twice after takeoff... you've definitely crashed. |
| 15:58:53 | <roconnor> | I prefer to only let my planes take off throughout the day. Then I go home at the end of the day. In the morning, all my planes have been put back on the ground for me. |
| 15:59:59 | <maltem_> | As long as you're not going for intercontinental flights. |
| 16:00:27 | <roconnor> | I'm just running local flights on my laptop |
| 16:03:14 | <Gilly> | Any particular reason to why Data.Time.Calendar.Day isn't a instance of Read? |
| 16:03:19 | <skorpan> | i'm trying to write some sort of typechecker using the reader monad... i've only done this using the state monad before, which apparently is supposed to be more "heavyweight" than necessary. this is my question: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5128#a5128 |
| 16:04:35 | <roconnor> | skorpan: do you want to stop when you hit an error, or carry on? |
| 16:04:41 | <skorpan> | roconnor: carry on |
| 16:05:02 | <roconnor> | skorpan: maybe use Writer and tell the errors to it? |
| 16:05:40 | <skorpan> | roconnor: that was the idea i had in the beginning, but friends of mine told me that Reader is more suiting for some reasons i don't remember and they're not accessible to ask right now |
| 16:05:59 | <skorpan> | i think they said that "local" is awesome for some reason |
| 16:06:25 | <roconnor> | do you need to know what the errors were as you go along? |
| 16:06:36 | <skorpan> | nope, i don't care about existing errors |
| 16:07:05 | <quicksilver> | Gilly: it is. |
| 16:07:08 | <roconnor> | skorpan: definitely not Reader. "local" will only add information locally. After you leave the local block it will be gone. |
| 16:07:26 | <skorpan> | yeah, i know! i'm glad i'm not the only one not understanding what Reader would be good for :) |
| 16:07:29 | <quicksilver> | Gilly: but the instance is perhaps not in one of the modules you important. |
| 16:07:35 | <quicksilver> | Gilly: try Data.Time.Format |
| 16:07:35 | <roconnor> | skorpan: local would be great if you needed the error report to carry on, but didn't want to report it in the end. |
| 16:07:51 | <roconnor> | skorpan: but this is exactly the opposite of what you want. |
| 16:08:14 | <skorpan> | i'll just go for writer and see if that makes my life difficult |
| 16:08:17 | <skorpan> | thanks for your input |
| 16:08:40 | <roconnor> | skorpan: use DList in Writer. |
| 16:08:50 | <skorpan> | for speed? |
| 16:08:52 | <roconnor> | or CPS your monad stack. |
| 16:08:54 | <roconnor> | ya |
| 16:09:00 | <skorpan> | CPS? |
| 16:09:08 | <roconnor> | continuation passing style |
| 16:09:16 | <skorpan> | ah, well, i'm not too familiar with that just yet |
| 16:09:33 | <roconnor> | well |
| 16:09:37 | <Gilly> | quicksilver: Ahh, thanks :) |
| 16:09:42 | <roconnor> | you probably don't have too many errors |
| 16:09:51 | <roconnor> | so maybe using DList isn't that important |
| 16:09:55 | <roconnor> | you can probably start with list. |
| 16:10:04 | <roconnor> | hmm |
| 16:11:05 | <Jedai> | skorpan: local is nice to manage the environment in a functional language interpreter for instance |
| 16:11:37 | <skorpan> | ah, right, one environment per block? |
| 16:11:38 | <mmorrow> | quicksilver: http://moonpatio.com/fastcgi/hpaste.fcgi/view?id=2462 |
| 16:11:59 | <mmorrow> | Elly: hehe |
| 16:12:14 | <Jedai> | skorpan: Right, of course for your problematic I also fail to see why Reader would help, it seems a typical Writer task |
| 16:26:29 | <pumpkin> | shh, I'm hunting bunny wabbits |
| 16:27:44 | <skorpan> | if i'm using "DList a" as the "state" of my Writer, do i have to (tell . singleton) everytime? |
| 16:30:56 | <quicksilver> | skorpan: yes. |
| 16:31:08 | <quicksilver> | skorpan: you can give that a name if you like :) |
| 16:31:14 | <quicksilver> | sktell = tell . singleton |
| 16:31:22 | <skorpan> | say = tell . D.singleton |
| 16:31:23 | <skorpan> | :) |
| 16:36:21 | <pumpkin> | mmorrow: that little chunk of code you wrote for closures in c is like the adjustor in the ghc rts? |
| 16:37:12 | <skorpan> | @type tell . Data.DList.singleton |
| 16:37:14 | <lambdabot> | Couldn't find qualified module. |
| 16:37:38 | <skorpan> | GHC only says it's a -> m () but that's not enough |
| 16:37:43 | <skorpan> | i can't figure the type out! |
| 16:38:12 | <lilac> | @index Data.DList |
| 16:38:12 | <lambdabot> | bzzt |
| 16:38:25 | <quicksilver> | why isn't that enough? |
| 16:38:41 | <quicksilver> | singleton has type (a -> DList a) |
| 16:38:56 | <skorpan> | tell is only for MonadWriter so i have to put that constraint somehow |
| 16:39:12 | <quicksilver> | ghc should tell you it. |
| 16:39:22 | <skorpan> | but it doesn't |
| 16:39:29 | <quicksilver> | MonadWriter (DList a) m => a -> m () |
| 16:39:33 | <quicksilver> | skorpan: I don't believe you. |
| 16:39:45 | <skorpan> | Warning: Definition but no type signature for `say' |
| 16:39:45 | <skorpan> | Inferred type: say :: a -> m () |
| 16:39:52 | <quicksilver> | ask ghci. |
| 16:40:16 | <skorpan> | that's inconvenient as hell because i haven't installed the package i'm writing this for |
| 16:40:20 | <dmwit> | And file a RFE against GHC. |
| 16:40:36 | <quicksilver> | you don't need that package |
| 16:40:38 | <quicksilver> | you just need DList |
| 16:40:43 | <skorpan> | hm, true |
| 16:40:43 | <dmwit> | Wait, ghci does exactly the same things ghc does. |
| 16:40:59 | <dmwit> | So if ghc can see whatever it is you're using, so can ghci. |
| 16:41:08 | <quicksilver> | dmwit: it's only the warning message at fault. |
| 16:41:17 | <quicksilver> | right. |
| 16:41:24 | <skorpan> | ghci gives me the right result |
| 16:41:26 | <quicksilver> | you can always run ghci up in the same context ghc is compilng from |
| 16:41:47 | <skorpan> | listen i'm not *lying* so obviously something strange is going on |
| 16:42:38 | <skorpan> | could it be that i'm already using the definition of "say" in other functions when i cabal build, which i don't when i check the type in ghci? |
| 16:42:59 | <roconnor> | @hoogle say |
| 16:43:00 | <lambdabot> | No results found |
| 16:43:26 | <roconnor> | @where hayoo |
| 16:43:26 | <lambdabot> | http://holumbus.fh-wedel.de/hayoo/hayoo.html |
| 16:43:36 | <skorpan> | say = tell . Data.DList.singleton |
| 16:44:26 | <mauke> | I get No instance for (MonadWriter (DList a) m) for that line |
| 16:44:55 | <roconnor> | DList a needs to be a Monoid. |
| 16:45:07 | <roconnor> | if it isn't the package ought to be changed? |
| 16:45:15 | <mauke> | it is |
| 16:46:40 | <mauke> | compiles with FlexibleContexts and (MonadWriter (DList a) m) => a -> m () |
| 16:47:50 | <roconnor> | huh? |
| 16:48:09 | <roconnor> | what about (MonadWriter [a] m) => a -> m () |
| 16:48:16 | <roconnor> | does that requires FlexibleContexts? |
| 16:48:19 | <mauke> | oy |
| 16:48:23 | <mauke> | owned by MR |
| 16:48:29 | <roconnor> | that's what I figured |
| 16:48:33 | <roconnor> | I should have said |
| 16:49:05 | <roconnor> | monomoprhismRestriction-- |
| 16:50:12 | <mauke> | yes, [a] also requires flexible contexts |
| 16:50:18 | <koeien> | too long :( |
| 16:50:38 | <mauke> | MR-- |
| 16:51:23 | <koeien> | monomorphismrestriction.com is still free! |
| 16:51:48 | <mauke> | preflex: karma MR |
| 16:51:49 | <preflex> | MR: -1 |
| 16:52:11 | <bremner> | mutability++ |
| 16:52:34 | <pumpkin> | :o |
| 16:52:43 | <roconnor> | mauke: I assume that NoMonomoprhismRestriction works instead of FlexibleContexts? |
| 16:52:52 | <roconnor> | mauke: or eta expending the definition of say. |
| 16:53:10 | <roconnor> | @burnDown bremner |
| 16:53:11 | <lambdabot> | Unknown command, try @list |
| 16:53:15 | <mauke> | eta expanding works if I remove the type signature |
| 16:53:55 | <Twey> | koeien: Quick! Let's grab it! |
| 16:53:58 | <Twey> | *cough* |
| 16:55:21 | <koeien> | Twey: let's register it to prevent it from being used ;) |
| 16:56:02 | <b_jonas> | monomorphismRestriction++ I for one like iy |
| 16:56:37 | <mauke> | BURN HIM |
| 16:59:39 | <maltem_> | @remember koeien Let's register it [monomorphismrestriction.com] to prevent it from being used ;) |
| 16:59:39 | <lambdabot> | It is forever etched in my memory. |
| 17:17:21 | <mmorrow_> | pumpkin: sort of.. it is like that in the sense that it could be used to give something a view as a function pointer, but that code doesn't take any args, but if it did it'd be exactly like that (in purpose, i'm not sure how exactly the adjustor works) |
| 17:18:23 | <mmorrow_> | pumpkin: but on the other hand, it's pretty much exactly what a suspended function application is represented as by the rts |
| 17:21:32 | <skorpan> | @pl not . null |
| 17:21:33 | <lambdabot> | not . null |
| 17:21:44 | <skorpan> | just checking if there is notNull :P |
| 17:22:34 | <pumpkin> | @hoogle [a] -> Bool |
| 17:22:34 | <lambdabot> | Prelude null :: [a] -> Bool |
| 17:22:34 | <lambdabot> | Data.List null :: [a] -> Bool |
| 17:22:34 | <lambdabot> | Prelude all :: (a -> Bool) -> [a] -> Bool |
| 17:22:42 | <skorpan> | @pl not . elem |
| 17:22:43 | <lambdabot> | not . elem |
| 17:22:49 | <skorpan> | hm... i thought it was wiser |
| 17:22:58 | <Twey> | That's already pointfree |
| 17:23:30 | <Twey> | @pl \x y -> not (elem x y) |
| 17:23:30 | <lambdabot> | (not .) . elem |
| 17:23:42 | <Twey> | Hmn |
| 17:24:01 | <Twey> | @unpl not . null |
| 17:24:01 | <lambdabot> | (\ c -> not (null c)) |
| 17:24:06 | <Twey> | @unpl not . elem |
| 17:24:07 | <lambdabot> | (\ c -> not (elem c)) |
| 17:24:18 | <Twey> | That's not very useful |
| 17:24:22 | <skorpan> | nope! |
| 17:25:40 | <roconnor> | @pl just gets rid of lambdas |
| 17:26:11 | <roconnor> | hmm, maybe that isn't true |
| 17:26:13 | <Elly> | @pl \x -> x + x |
| 17:26:13 | <lambdabot> | join (+) |
| 17:26:18 | <Elly> | :t join |
| 17:26:19 | <lambdabot> | forall (m :: * -> *) a. (Monad m) => m (m a) -> m a |
| 17:26:21 | <roconnor> | > flip (flip id) |
| 17:26:22 | <lambdabot> | Overlapping instances for Show ((a -> c) -> a -> c) |
| 17:26:22 | <lambdabot> | arising from... |
| 17:26:25 | <Twey> | It does just get rid of lambdas |
| 17:26:27 | <roconnor> | @pl flip (flip id) |
| 17:26:28 | <lambdabot> | id |
| 17:26:28 | <Elly> | what the heck is that |
| 17:26:30 | <Twey> | But it has some neat tricks for doing so |
| 17:26:38 | <Twey> | Elly: join on the function monad |
| 17:26:47 | <Elly> | I didn't know there was a function monad |
| 17:26:53 | <Twey> | Essentially, it just repeats the argument |
| 17:27:09 | <Twey> | m here is (-> r) |
| 17:27:16 | <roconnor> | @djinn (e -> e -> a) -> (e -> a) |
| 17:27:17 | <lambdabot> | f a b = a b b |
| 17:27:23 | <Twey> | So, it's :: (a -> a -> r) -> (a -> r) |
| 17:27:33 | <Elly> | nice |
| 17:27:43 | <JuanDaugherty> | googling "haskell calculator" returns a half million hits but they all seem to be half assed. Anybody know of a decent one with an X interface? |
| 17:27:46 | <Twey> | Yeah, there's a function monad |
| 17:28:06 | <mauke> | JuanDaugherty: why an X interface? |
| 17:28:14 | <mauke> | I think ghci is an acceptable calculator |
| 17:28:24 | <roconnor> | JuanDaugherty: half-assed calculators is typically a test case for new GUI development libs. |
| 17:28:29 | <Twey> | > (length >>= (*2)) [1, 2, 3] |
| 17:28:30 | <lambdabot> | Couldn't match expected type `[a] -> b' against inferred type `Int' |
| 17:28:32 | <Twey> | Boo |
| 17:28:48 | <Twey> | I can never remember how to use it, but it exists :-P |
| 17:28:58 | <roconnor> | @let loeb x = fmap ($ loeb x) x |
| 17:28:58 | <lambdabot> | Defined. |
| 17:29:05 | <Twey> | > do { l <- length; return l } [1, 2, 3] |
| 17:29:06 | <mauke> | needs more fix |
| 17:29:06 | <lambdabot> | <no location info>: parse error on input `[' |
| 17:29:11 | <roconnor> | ACTION thinks ghci is also a fine spreadsheet |
| 17:29:17 | <roconnor> | > loeb [ (!!5), const 3, liftM2 (+) (!!0) (!!1), (*2) . (!!2), length, const 17] |
| 17:29:18 | <lambdabot> | [17,3,20,40,6,17] |
| 17:29:35 | <JuanDaugherty> | mauke: dunno of you are asking why a gui, or why that gui platform, but um, just because. |
| 17:29:39 | <JuanDaugherty> | *if |
| 17:29:41 | <mauke> | why a gui |
| 17:29:55 | <mauke> | I haven't seen a calculator that was improved by having a GUI |
| 17:30:02 | <JuanDaugherty> | for the reason gui exists |
| 17:30:09 | <Twey> | I've never been able to find good information on the (-> r) monad |
| 17:30:16 | <Twey> | Where does it lurk? |
| 17:30:28 | <roconnor> | Control.Monad.Instances I beleive |
| 17:30:37 | <maltem_> | Twey, btw it's ((->) r) |
| 17:30:56 | <Twey> | roconnor: I meant the information, not the monad |
| 17:31:15 | <Twey> | Oh, yeah |
| 17:31:16 | <roconnor> | it's isomorphic to (Reader r) |
| 17:31:18 | <Twey> | It's (r ->) |
| 17:31:44 | <Cale> | Good information lurks in the minds of the people of this channel :) |
| 17:32:02 | <Cale> | Also, it's a good exercise to try writing the instance yourself :) |
| 17:32:23 | <Cale> | If you get the types to match, it is almost impossible to have the wrong thing. |
| 17:32:24 | <roconnor> | Cale: actually everyone is a mindless zombie automoton, other than Twey. |
| 17:32:44 | <mauke> | JuanDaugherty: gui exists to display graphical information and let me click on things |
| 17:32:52 | <JuanDaugherty> | roconnor: Acknowledged, a decent one (as opposed to a half assed one) developed for that purpose would be fine. |
| 17:33:02 | <Cale> | > (do x <- id; y <- reverse; z <- map toUpper; return (x,y,z)) "Hello" |
| 17:33:03 | <lambdabot> | ("Hello","olleH","HELLO") |
| 17:33:04 | <Twey> | roconnor: I always knew it :-P |
| 17:33:13 | <JuanDaugherty> | mauke: yeah that's the ticket, just lemme at it |
| 17:33:16 | <zachk> | couldnt you use a map |
| 17:33:16 | <Twey> | Ah, that's the way to do it, is it? |
| 17:33:31 | <zachk> | omfg that was like 3 days ago forget i ever said that |
| 17:33:31 | <Cale> | > sequence [id, (+2), (*2), (^2), (2^)] 5 |
| 17:33:33 | <lambdabot> | [5,7,10,25,32] |
| 17:33:39 | <CalJohn> | is it possible to parse an integer token with attoparsec? |
| 17:33:39 | <Twey> | Hahaha, zachk |
| 17:33:40 | <b_jonas> | I may be mindless but I'm not a zombie |
| 17:33:59 | <CalJohn> | i'm learning parsec, but I'd like to use bytestrings |
| 17:34:24 | <Twey> | CalJohn: Get Parsec 3, use Text.Parsec.ByteString |
| 17:34:33 | <roconnor> | b_jonas: philosophical zombie, not hollywood zombie. |
| 17:36:13 | <Cale> | http://hackage.haskell.org/cgi-bin/hackage-scripts/package/ChristmasTree -- what a great contrived name for a package :) |
| 17:36:30 | <Twey> | Hahahaha |
| 17:36:31 | <Twey> | Nice! |
| 17:37:20 | <koeien> | lol lol lol |
| 17:37:22 | <koeien> | very nice |
| 17:37:27 | <b_jonas> | yeah |
| 17:37:38 | <koeien> | was that the utrecht thingy? |
| 17:37:41 | <koeien> | sounds like it |
| 17:37:51 | <CalJohn> | Twey: ah, wow, did not know |
| 17:38:05 | <koeien> | home page is cs.uu.nl, yeah apparently :) |
| 17:38:18 | <gio123> | please heeeeeeeeeeelp I need file amsmath.cls |
| 17:38:27 | <CalJohn> | koeien: i am attending that utrecht thingy |
| 17:38:41 | <mmorrow_> | pumpkin: this code though lets you make a closure of a C function and a value, then gives you back a function pointer which when called with another arg will call that original function with both saved arg and the new arg http://moonpatio.com/repos/cclos.c |
| 17:38:55 | <Cale> | gio123: That doesn't seem like a Haskell question... isn't that a LaTeX class file? |
| 17:39:01 | <b_jonas> | it is |
| 17:39:04 | <koeien> | CalJohn: i mean, i saw a presentation on that in Eindhoven by doaitse swierstra from Utrecht university |
| 17:39:08 | <mmorrow_> | pumpkin: (only have x86_64 asm though for that thus far) |
| 17:39:25 | <koeien> | gio123: try google. |
| 17:39:26 | <CalJohn> | koeien: oh, i was mistaken |
| 17:39:35 | <mmorrow_> | pumpkin: the thing that sucks about x86 is that you can't do pc-relative references easily |
| 17:39:35 | <gio123> | Cale: yes |
| 17:39:36 | <gio123> | :) |
| 17:39:52 | <Cale> | gio123: I'm pretty sure that it comes with most of the popular distributions of LaTeX... |
| 17:39:54 | <koeien> | CalJohn: the name (at least, the unabbreviated version) reminded me of that |
| 17:39:58 | <b_jonas> | mmorrow_: that's not the only think that sucks about x86 |
| 17:40:06 | <mmorrow_> | b_jonas: heh |
| 17:40:16 | <b_jonas> | I think I sound like a bot |
| 17:40:27 | <gio123> | Cale: noo, I am compiling the documents and it gives me error that that file can not fond |
| 17:40:28 | <b_jonas> | I make shallow comments like "yeah" and "um I don't think so" about whatever you say |
| 17:40:43 | <Cale> | How did you install LaTeX? |
| 17:40:52 | <b_jonas> | gio123: could you ask on another channel? like #haskell-blah ? |
| 17:41:14 | <gio123> | Cale: pm |
| 17:41:18 | <b_jonas> | (I must sound like that again) |
| 17:41:38 | <b_jonas> | anyway, I decided the magic isn't in extendible exceptions but in Typeable that is uses |
| 17:41:46 | <b_jonas> | so now I'll have to read how Typeable works |
| 17:42:06 | <skorpan> | @pl \x y -> not (null x) && not (null y) |
| 17:42:08 | <lambdabot> | (. (not . null)) . (&&) . not . null |
| 17:42:12 | <skorpan> | ...right... |
| 17:42:37 | <mauke> | :t (&&) `on` (not . null) |
| 17:42:37 | <kyagrd> | Is there a (development) version of TH that can parse GADT declerations? |
| 17:42:38 | <lambdabot> | forall a. [a] -> [a] -> Bool |
| 17:43:07 | <mauke> | :t join (liftM2 (&&)) (not . null) |
| 17:43:08 | <lambdabot> | forall a. [a] -> Bool |
| 17:43:12 | <mauke> | d'oh |
| 17:43:42 | <mmorrow_> | pumpkin: (and the best part of that cclos.c code is that it creates a new copy of the code for each closure, so you can have arbitrarily many of them and they won't interfere, no global vars used) |
| 17:52:14 | <mmorrow_> | kyagrd: unfortunately no |
| 17:56:12 | <mxc> | (a -> b -> c) |
| 17:56:14 | <mxc> | oops |
| 17:56:18 | <mxc> | oops |
| 17:59:29 | <lilac> | mmorrow_: am i right in thinking that they added rip-relative addressing in x86_64? |
| 18:00:16 | <mmorrow_> | lilac: yes, it's awesome |
| 18:00:31 | <lilac> | i imagine it's especially awesome for the authors of virii and trojans |
| 18:00:38 | <mmorrow_> | label(%rip) |
| 18:01:06 | <lilac> | what does the r stand for, anyway? ridiculously-big? |
| 18:01:13 | <mmorrow_> | lilac: exactly! |
| 18:01:20 | <mmorrow_> | i dunno |
| 18:01:22 | <mauke> | VIRII IS NOT A WORD |
| 18:01:49 | <mmorrow_> | is cactuses a word? |
| 18:02:02 | <mauke> | it is acceptable |
| 18:02:15 | <lilac> | > words "virii and trojans" |
| 18:02:16 | <lambdabot> | ["virii","and","trojans"] |
| 18:02:22 | <lilac> | lambdabot disagrees |
| 18:02:22 | <deech> | Hi all, |
| 18:02:30 | <mauke> | lilac: no, I disagree! |
| 18:02:43 | <lilac> | mauke: lambdabot says it's a word, though! |
| 18:02:47 | <lilac> | > words "virii" |
| 18:02:48 | <lambdabot> | ["virii"] |
| 18:03:02 | <mauke> | > words "`__)" |
| 18:03:03 | <lambdabot> | ["`__)"] |
| 18:03:10 | <mmorrow_> | mauke: heh, i actually just had to lookup the plural of cactus and it's "cacti" with only one i (or cactuses) |
| 18:03:11 | <inimino> | @faq can Haskell solve disputes over English? |
| 18:03:11 | <lambdabot> | The answer is: Yes! Haskell can do that. |
| 18:03:38 | <inimino> | good to know |
| 18:03:39 | <mauke> | there is no "-ii" suffix in latin |
| 18:03:41 | <mmorrow_> | mauke: so i guess it's different a thing altogether |
| 18:03:44 | <mmorrow_> | oh |
| 18:03:51 | <mmorrow_> | is there an -i ? |
| 18:03:57 | <mauke> | yes, several |
| 18:04:03 | <lilac> | @faq Can Haskell pose a polynomial-time-checkable problem so hard even Haskell cannot solve it in polynomial time? |
| 18:04:03 | <lambdabot> | The answer is: Yes! Haskell can do that. |
| 18:04:08 | <deech> | I'm trying to use Network.HTTP to browse a web address, but I want to handle the case when the network isn't available. I am currently getting a GHCI error in my handler function. The code is at http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5129#a5129. Appreciate your input. |
| 18:04:14 | <lilac> | wow. P != NP. you heard it here first. |
| 18:04:24 | <mauke> | "sons" is filii because a single son is filius |
| 18:05:11 | <lilac> | mauke: i envy how much you care about this :) |
| 18:05:34 | <mauke> | guess what I have in my highlight list |
| 18:05:46 | <inimino> | *ii ? |
| 18:05:51 | <mmorrow_> | hahah |
| 18:05:54 | <mauke> | no, "virii" |
| 18:06:12 | <mauke> | but don't worry, it's not an automatic /kick |
| 18:06:16 | <mauke> | (yet) |
| 18:06:29 | <lilac> | mauke: was this whole response merely a script in your IRC client playing itself out? |
| 18:06:31 | <da-x> | @faq Can Haskell explain what happens on Lost? |
| 18:06:31 | <lambdabot> | The answer is: Yes! Haskell can do that. |
| 18:06:38 | <inimino> | ACTION is glad mauke is on the case |
| 18:06:40 | <inimino> | some one has to defend Latin or it might die out! |
| 18:06:41 | <mauke> | lilac: haha, no :-) |
| 18:07:22 | <lilac> | inimino: latin is clearly actively maintained. iso latin15 didn't come out that long ago |
| 18:08:16 | <mmorrow_> | i wonder if there's a -fii-suffixii-allowed flag |
| 18:08:33 | <lilac> | sufficii, surely? ;-) |
| 18:08:47 | <mauke> | suffixen |
| 18:09:40 | <lilac> | mauke: hold on, you have virii in your highlight list? you're actively /hunting/ people who are wrong on the internets? |
| 18:09:43 | <inimino> | ACTION waits for someone to write Haskelligata, translating the entire type system into suffixes |
| 18:09:59 | <mauke> | lilac: precisely |
| 18:10:17 | <lilac> | ACTION waits for someone to paste an xkcd link |
| 18:10:37 | <mauke> | god help you if I catch you using "could care less" |
| 18:11:29 | <mmorrow_> | ha |
| 18:11:35 | <Twey> | ‘Virii’ is a word — it means ‘males’ |
| 18:11:45 | <Twey> | mauke: *shudder* |
| 18:11:45 | <lilac> | mauke: i'm with you on that |
| 18:11:51 | <mauke> | Twey: no, that's viri |
| 18:11:58 | <lilac> | virii is the plural of virium |
| 18:11:59 | <mauke> | men |
| 18:12:00 | <Twey> | mauke: It's virī |
| 18:12:09 | <lilac> | which is a word i made up |
| 18:12:13 | <mauke> | lilac: no, that would be viria |
| 18:12:20 | <JuanDaugherty> | ACTION concludes that the character set socalled "ISO-LATIN-15" has little or nothing to do with the Latin language (besides the obvious) |
| 18:12:24 | <lilac> | mauke: virius? |
| 18:12:27 | <mauke> | visum -> visa, etc |
| 18:12:32 | <mauke> | lilac: yes |
| 18:12:35 | <Twey> | The double-for-long-vowel convention was created by Accius |
| 18:13:09 | <Twey> | JuanDaugherty: Yeah — it has weird characters in it like ‘U’ and ‘W’ |
| 18:13:57 | <mauke> | I thought it was called ISO 8859-15 |
| 18:14:21 | <lilac> | mauke: Virius was a Roman first name. When referring to multiple people with this first name, is 'Virii' not appropriate? |
| 18:14:21 | <JuanDaugherty> | a "real" Latin set might have for example have had Claudius' characters |
| 18:14:51 | <mauke> | lilac: do you have a reference for that? |
| 18:15:30 | <b_jonas> | JuanDaugherty: is that the half H and turned C thing? |
| 18:15:36 | <JuanDaugherty> | y |
| 18:15:51 | <centrinia> | Which declension does "Virius" fall under? |
| 18:16:00 | <mauke> | tee hee |
| 18:16:06 | <b_jonas> | what I don't get is, why did they bother with introducing G as a variant form of C when they already had K and Q as extra letter Cs |
| 18:16:07 | <mauke> | Viriūs |
| 18:16:49 | <b_jonas> | I mean, the greeks used to have two letters for K as well, why do they have to repeat their mistake? |
| 18:17:09 | <mauke> | what was the other greek K? |
| 18:17:27 | <b_jonas> | English spelling still sucks because of the Latin's bad decisions |
| 18:17:53 | <centrinia> | varkappa? |
| 18:18:24 | <b_jonas> | no, Qoppa |
| 18:18:43 | <b_jonas> | it fell into disuse because the greeks did realize it was redundant |
| 18:18:49 | <centrinia> | That looks more like a 'Q' |
| 18:19:08 | <mauke> | ACTION faintly remembers things called "koppa" and "stigma" |
| 18:19:58 | <Twey> | centrinia: ‘Virius’ looks 2m |
| 18:20:31 | <centrinia> | Second declension masculine? |
| 18:20:42 | <abbe> | http://xkcd.com/587/ |
| 18:21:56 | <Twey> | centrinia: Yeah |
| 18:22:50 | <centrinia> | That would look like Viriī |
| 18:23:14 | <b_jonas> | if we could get people to use a writing that corresponds to phonetic once, then maybe phonetics wouldn't change much after that and then we wouldn't have all the historical accidents that caused us to use writings deviating from phonetics in first place |
| 18:23:37 | <b_jonas> | especially now in the age of global communication |
| 18:23:43 | <mauke> | and next week: "i.e." vs. "e.g." |
| 18:24:11 | <b_jonas> | because I can sort of understand how the geographic distances could have caused the problems in the ancient times |
| 18:27:32 | <Twey> | centrinia: The Romans rarely marked long vowels |
| 18:29:15 | <centrinia> | If zombies can only eat two brains and turn their victims into zombies; and you start out with one zombie, then shouldn't the n-th generation of brain eating victims have F_n members? |
| 18:30:44 | <b_jonas> | no |
| 18:31:04 | <centrinia> | Why not? |
| 18:31:13 | <lilac> | mauke: virius lupus was some famous roman dude |
| 18:31:14 | <b_jonas> | there'd be 2^n |
| 18:31:21 | <Cale> | 2^n - 1 |
| 18:31:59 | <b_jonas> | Cale: no, "the n-th generation of brain eating victims" doesn't include earlier ones I think |
| 18:32:15 | <Cale> | oh, if that's the case, then yes |
| 18:32:28 | <tromp> | :t showOct |
| 18:32:30 | <lambdabot> | forall a. (Integral a) => a -> String -> String |
| 18:32:31 | <b_jonas> | well, either can be a plausible interpretation |
| 18:32:37 | <b_jonas> | after all, zombies don't die |
| 18:32:44 | <b_jonas> | they can only be destroyed |
| 18:32:48 | <tromp> | > showOct 797 "" |
| 18:32:49 | <lambdabot> | "1435" |
| 18:32:50 | <Cale> | I thought it was talking about the total zombie population :) |
| 18:33:02 | <centrinia> | What if the zombie can only eat one brain during one generation? |
| 18:33:26 | <b_jonas> | centrinia: that would be a strange definition of "generation" |
| 18:33:28 | <Cale> | centrinia: But they keep earing? |
| 18:33:32 | <Cale> | eat* |
| 18:33:46 | <centrinia> | They go back to the grave after eating two brains. |
| 18:34:08 | <lizrail> | Hi, can anybody explain, what's the reason for this error in ghci: Loading package HTTP-4000.0.6 ... linking ... <interactive>: /usr/lib64/http-4000.0.6/ghc-6.10.1/HSHTTP-4000.0.6.o: unknown symbol `networkzm2zi2zi1_NetworkziURI_uri_closure' |
| 18:34:08 | <lizrail> | ghc: unable to load package `HTTP-4000.0.6' |
| 18:34:51 | <voker57_> | version mismatch in network |
| 18:35:06 | <mauke> | preflex: zdec networkzm2zi2zi1_NetworkziURI_uri_closure |
| 18:35:06 | <preflex> | network-2.2.1_Network.URI_uri_closure |
| 18:35:07 | <Cale> | lizrail: hmm, my first thought was "you're not compiling with --make" but since you got it in GHCi, it's probably a version mismatch like voker57_ says :) |
| 18:35:33 | <lizrail> | mismatch between what and what? I tried rebuild network and http |
| 18:35:54 | <Cale> | lizrail: hmm... |
| 18:36:00 | <centrinia> | So zombie a_{0,0} eats the brain of zombie a_{1,0}; zombie a_{0,0} and zombie a_{1,0} eat the brains of zombies a_{2,0} and a_{2,1}; zombie a_{1,0}, zombie a_{2,0}, and zombie a_{2,1} eat the brains of zombie a_{3,0}, a_{3,1}, and a_{3,2} |
| 18:37:18 | <tromp> | > showOct 410 "" |
| 18:37:20 | <lambdabot> | "632" |
| 18:37:48 | <lizrail> | Cale, I get this when trying to compile: http://pastebin.com/m165ce497 |
| 18:37:53 | <tromp> | > showOct 1024 "" |
| 18:37:55 | <lambdabot> | "2000" |
| 18:38:51 | <Cale> | lizrail: Yeah, it's got to be something strange with either network or http... I'm not sure what though, I've never run into that before. |
| 18:40:11 | <skorpan> | @type when |
| 18:40:12 | <lambdabot> | forall (m :: * -> *). (Monad m) => Bool -> m () -> m () |
| 18:40:19 | <Cale> | centrinia: The number of freshly bitten zombies on a given turn is the sum of the freshly bitten zombies from the previous two generations. So that's Fibonacci-ish... |
| 18:41:10 | <lizrail> | Cale, what can be done to investigate the problem? |
| 18:41:31 | <tetha> | mm, I remember that there is a simulator of a zombie outbreak in a building somewhere |
| 18:43:29 | <Cale> | lizrail: Let's start with ghc-pkg list network |
| 18:44:02 | <Cale> | Is there more than one entry? You should have a network package that came with your ghc installation most likely. |
| 18:44:13 | <Cale> | But there might also be a locally installed one? |
| 18:44:34 | <Cale> | I'm running 6.10.3 at the moment, so I'm not sure what version is appropriate. |
| 18:44:40 | <lizrail> | Cale, http://pastebin.com/d460e5836 |
| 18:45:07 | <Cale> | Oh, interesting, so there are two of them which are the same... |
| 18:46:01 | <Cale> | Maybe try uninstalling that local version of network: ghc-pkg --user unregister network |
| 18:46:02 | <lizrail> | Cale, I believe it because at first I built one as user with cabal and later second was installed from haskell overlay |
| 18:46:40 | <lizrail> | Cale, all fixed! |
| 18:46:47 | <lizrail> | thanks |
| 18:46:57 | <Cale> | Ah, that solved it? Nice :) |
| 18:47:45 | <lizrail> | still I'm puzzled how to prevent similar issues in the future.. |
| 18:48:11 | <lizrail> | by not using --user with cabal? |
| 18:48:13 | <Cale> | Well, a good rule of thumb is not to use cabal to install packages that normally come with ghc. It's not ideal... |
| 18:48:35 | <lizrail> | Cale, but what if I want to update them? |
| 18:48:57 | <lizrail> | or some other package wants more recent version? |
| 18:49:19 | <Cale> | Usually if a package wants a newer version of one of those, it's meant to be built with a newer GHC anyway. |
| 18:49:46 | <lizrail> | ) |
| 18:49:50 | <Cale> | I think that normally it's okay to have multiple versions of a package installed though |
| 18:50:06 | <Cale> | I think it may have got confused by the fact that they were the same version. |
| 18:50:52 | <Cale> | dcoutts would know more about this than I do |
| 18:51:23 | <Cale> | (He wrote much of Cabal and is really familiar with the package system) |
| 18:51:35 | <eivuokko> | It should be possible to link package with same name, but different versions, currently. |
| 18:51:37 | <lizrail> | yes, though I remember I runned into troubles in the past then two versions of package was used by two different packages in one program |
| 18:52:03 | <eivuokko> | The types between the different versions are never compatible in type system. |
| 18:52:34 | <Cale> | We have some annoying dependency problems like that in some of the packages on Hackage. Particularly stupid issues like packages demanding different versions of QuickCheck. |
| 18:52:42 | <Cale> | (Even though they only use it for testing) |
| 18:53:10 | <eivuokko> | I think that is also related to how cabal tries to optimise the dependencies? |
| 18:53:29 | <Cale> | Possibly. I seem to recall that installing them one by one separately helps. |
| 18:54:11 | <eivuokko> | Yeah, basically Cabal optimises dependencies on general case (not ghc), which means you can't have Qc<2.0 and Qc>=2.0 at one time in dependency solver. |
| 18:54:22 | <eivuokko> | Or did last I read it, but that's a while ago. |
| 18:55:18 | <lizrail> | deleting .cabal and rebuilding nessesary packages helped me one time |
| 18:55:25 | <Cale> | I've also had the dependency solver make bad decisions about which version to install, like it would inexplicably pick a really old version of something to download and install. |
| 18:55:46 | <Cale> | But if you asked for the deps separately, it was fine. |
| 18:55:53 | <eivuokko> | Hmmm. |
| 18:56:02 | <eivuokko> | That sounds like a bug in SAT solver thingy... |
| 18:56:17 | <mreh> | conal: is your webiste good material to read for game programing? |
| 18:56:28 | <mreh> | "Simply efficiennt functional reactivity" |
| 18:56:58 | <Cale> | mreh: Yes, but only if you're prepared to write your game in a new and somewhat experimental way :) |
| 18:57:27 | <Cale> | mreh: After playing around with Reactive, I'm pretty convinced that something like it is the future of game programming. |
| 18:57:37 | <conal> | mreh: i agree with Cale, though he may be being too generous. |
| 18:57:37 | <lambdabot> | conal: You have 4 new messages. '/msg lambdabot @messages' to read them. |
| 18:57:52 | <conal> | Cale: :) ! |
| 18:57:57 | <Cale> | mreh: But the current implementation still has some bugs which make it a bit hard to use practically... |
| 18:58:15 | <conal> | mreh: games definitely fit into what i want Reactive to support well. |
| 18:58:31 | <lizrail> | Cale, please, give a link ) |
| 18:58:32 | <conal> | yeah. what Cale said. |
| 18:58:50 | <conal> | the programming model needs work, also. |
| 19:00:02 | <conal> | so far, FRP doesn't handle input in a composable way. |
| 19:00:03 | <skorpan> | if i have a bunch of type constructors for the same type, e.g. data A t = A t t | B t t t | C t, and i want to be able to access the first parameter for each constructor, is there any generic way of doing this? i don't mind e.g. TH? |
| 19:00:06 | <Cale> | lizrail: To conal's blog? http://conal.net/blog/ |
| 19:00:26 | <FunctorSalad> | skorpan: gmapQi |
| 19:00:28 | <Cale> | lizrail: The best introductory articles are quite a ways back now though... |
| 19:00:44 | <skorpan> | FunctorSalad: i for index? |
| 19:00:46 | <conal> | the paper "genuinely functional user interfaces" set a direction for FRP with composable input, but it wasn't very much pursued. yampa got to a bit of it. |
| 19:00:49 | <mreh> | conal: i'll skip it |
| 19:00:59 | <mreh> | but I am looking for interesting projects |
| 19:01:19 | <conal> | mreh: sure. it'll give you more enlightenment than anything at this point. |
| 19:01:25 | <FunctorSalad> | skorpan: yes. head/gmapQ is fine too of course |
| 19:01:57 | <lizrail> | Cale, thanks, lots of promising titles. I must go now. Bye. |
| 19:01:59 | <skorpan> | i'd like some head... |
| 19:02:16 | <FunctorSalad> | skorpan: although I'm not sure syb can deal with this kind of polymorphism :( |
| 19:02:18 | <mreh> | im doing machine learning next year, FP should make it a whole lot easier |
| 19:02:40 | <FunctorSalad> | you can make "A" an instance of Typeable1 |
| 19:02:50 | <skorpan> | this should be fairly easy with TH though, right? |
| 19:03:04 | <FunctorSalad> | (but I'm not sure what the predicate to gmapQ should look like) |
| 19:03:09 | <skorpan> | sad thing is i know virtually no TH myself |
| 19:03:09 | <FunctorSalad> | skorpan: yeah, use "reify" |
| 19:03:23 | <FunctorSalad> | then extract the constructors from the info |
| 19:03:54 | <skorpan> | could you help me with that? :/ sorry if i'm asking too much of you! |
| 19:05:07 | <mreh> | I like conals |
| 19:05:13 | <mreh> | 's website |
| 19:05:25 | <mreh> | "The goal of my new research agenda is to give artists access to the expressive power of computers" |
| 19:06:07 | <conal> | mreh: thx. i better re-read that sentence regularly so i don't forget! |
| 19:06:33 | <conal> | i still get regularly wrapped up in type theory, denotational semantics, etc. |
| 19:08:16 | <Cale> | Did you ever figure out what was going on with the integral equation problem? |
| 19:08:28 | <solrize_> | i really like .lhs. i write much more heavily commented code just from not having to type "--" on every line. |
| 19:09:44 | <mreh> | what dat? |
| 19:10:26 | <Cale> | mreh: hm? |
| 19:10:31 | <Cale> | lhs? |
| 19:10:35 | <b_jonas> | {- haskell has long comments -} |
| 19:10:38 | <b_jonas> | Cale: literal hs |
| 19:10:44 | <wli> | What kind of integral equation? |
| 19:10:48 | <b_jonas> | it has that file extension |
| 19:10:55 | <Cale> | b_jonas: I know, I was asking what mreh was asking about. |
| 19:10:57 | <glguy> | its a great way to fill your source code with pages of outdated comments :) |
| 19:11:06 | <mreh> | Cale: I was asking what .lhs is |
| 19:11:17 | <b_jonas> | oh |
| 19:11:24 | <solrize_> | b_jonas yeah, but that still feels like writing comments that are interspersed into code. .lhs feels like writing text with code interspersed into it. |
| 19:11:39 | <solrize_> | and i want to start using lhs2latex |
| 19:11:39 | <glguy> | mreh: instead of surrounding your comments with delimiters, you surround your code with them |
| 19:11:45 | <glguy> | > code goes here |
| 19:11:46 | <lambdabot> | Not in scope: `code'Not in scope: `goes'Not in scope: `here' |
| 19:11:49 | <b_jonas> | well I don't really like literate programming |
| 19:12:01 | <b_jonas> | I don't write many comments either |
| 19:12:07 | <FunctorSalad> | skorpan: rough outline... http://hpaste.org/fastcgi/hpaste.fcgi/view?id=5130#a5130 |
| 19:12:09 | <b_jonas> | sometimes I do write some comments of course |
| 19:12:21 | <burp_> | wtf, > in front of each code line? |
| 19:12:25 | <burp_> | how mad is that |
| 19:12:26 | <Cale> | mreh: It's Literate Haskell. It makes commenting the default, and there are two styles of it. The first requires that each code line begins with "> " and the second puts code in between \begin{code} and \end{code} |
| 19:12:31 | <Vq^> | burp_: thats the idea |
| 19:12:45 | <Cale> | It's quite nice actually. |
| 19:13:05 | <FunctorSalad> | skorpan: reference for the abstract syntax tree..... http://hackage.haskell.org/packages/archive/template-haskell/2.3.0.1/doc/html/Language-Haskell-TH-Syntax.html |
| 19:13:12 | <Cale> | wli: Anything recursive. |
| 19:13:25 | <FunctorSalad> | (there are lower-case versions of all these constructors, which operate on values in the Q monad) |
| 19:13:25 | <Vq^> | burp_: consider the situation when you have a lot more documentation than actual sourcecode |
| 19:13:46 | <burp_> | ok |
| 19:13:49 | <MyCatVerbs> | The bird notation ("> " at the start of every line) is used on a lot of Haskell blogs so that one can just copy and paste their posts into a file, stick .lhs on the end of it and compile. |
| 19:13:56 | <Cale> | wli: But in my case, it was the integral equations for describing the position of an object in terms of its acceleration and velocity, when it had a drag acceleration dependent on the velocity. |
| 19:14:10 | <b_jonas> | Vq^: then I have some readme files or pod files or whatever files that only have docs, no code | <
