October 2015 - Haskell-Cafe

Building 32-bit binaries on a 64-bit system?
by Mike Meyer 19 Oct '15

19 Oct '15

I've got someone who wants to run one of my haskell apps on a 32-bit Windows 10 box. I've pretty much moved to 64-bit everywhere, so don't have such a system to build on. Is it possible to get stack/cabal/ghc to build a 32-bit Windows binary on a 64-bit Windows box? Google found a bunch of things about building 64 bit applications, and the ghc manual doesn't seem to have such an option. So do I need to get a 32-bit system for this?

5 6

STM and garbage collector
by Tomas Carnecky 18 Oct '15

18 Oct '15

If a thread is blocking indefinitely in an STM transaction (reading from a queue to which nobody else has a reference and thus can not write into), is the runtime smart enough to GC the thread? Or do I have to kill the thread manually?

3 3

make an PrimMonad-wrapped FFI call
by Alexey Egorov 18 Oct '15

18 Oct '15

Hello haskellers, I'm writing FFI bindings to C library which uses some allocations across functions call, for example: int *alloc_buf(void); void free_buf(int *); So I need to use newForeignPtr to make sure buffer will be free'd when all pointers to it get collected. But what if I want to do this in PrimMonad (say, ST) and not IO (returned by newForeignPtr)? Thanks.

1 0

How to specialize a type
by martin 17 Oct '15

17 Oct '15

Hello all, Suppose I have a type "Process" which stands either for a Train or a moving Belt. In both cases there is a place of departure and a place of arrival. However the other parameters which describe a Process differ. A Train has departure and arrival times, but a Belt has a speed. I tried the following: type PlaceDep = Int type PlaceArr = Int data Process = Train PlaceDep PlaceArr TP | MovingBelt PlaceDep PlaceArr BP deriving (Eq, Show) data TP = TP Int deriving (Eq, Show) data BP = BP Int deriving (Eq, Show) prc1 = Train 10 11 (TP 1) prc2 = MovingBelt 12 13 (BP 2) What I don't like about this is that the fact that all Processes have PlaceDep and PlaceArr appears somewhat "coincidental". This in contrast, captures the common parts more clearly: type PlaceDep = Int type PlaceArr = Int data ProcessParams = DepartureTime Int | Speed Int deriving (Eq, Show) data Process = Process PlaceDep PlaceArr ProcessParams deriving (Eq, Show) prc1 = Process 10 11 (DepartureTime 1) prc2 = Process 12 13 (Speed 2) Is this the classic way of specializing a type or are there better options See also: http://stackoverflow.com/questions/33156656/subclassing-a-type-in-haskell/3…

2 3

Trying to write an Embedded DSL that threads a monad?
by Oleg 17 Oct '15

17 Oct '15

> Ian Bloom wrote: > I've been hoping to design a domain specific embedded language in > Haskell that would let me pipe a commutative monad throughout an expression > written in the language. Special terms within the language will eventually > have access to this monad. Is there a reason tagless-final approach does not work for you? You can easily thread any monad, even if you don't make any provision for it initially. For example, http://okmij.org/ftp/tagless-final/index.html#call-by-any http://okmij.org/ftp/tagless-final/course/CBAny.hs uses monad to print out traces of the expression and then later to implement lazy evaluation for EDLS (which needs mutable cells).

2 2

Three open positions at Beautiful Destinations
by Tom Nielsen 16 Oct '15

16 Oct '15

The Data Science team at Beautiful Destinations is looking for three full-time developers and researchers with strong quantitative and programming skills. Beautiful Destinations ( http://www.beautifuldestinations.com/) is the world's largest travel influencer with a highly engaged audience of over 7 million followers across Instagram, Facebook and Snapchat. Beautiful Destinations distributes beautiful travel photography to its followers and helps clients to grow their social media channels. Our in-house photographers use drones and still photography to create content for our own accounts and our clients. The data science team has three aims: to predict and enhance the engagement of photographs on specific accounts; to measure influence on social media; and to determine the factors that contribute to the growth of social media accounts. Our internal API and core infrastructure are written in Haskell (using PostgreSQL), our Bayesian statistical time-series models use Stan ( http://mc-stan.org/) as well as some statistical models in Haskell. For image processing we use convolutional neural networks (in Python) and SIFT (C++/CUDA/Haskell). We are open to people bringing their preferred tools and maintain a polyglot environment. We are looking for developers who can fill the gap between data science and software engineering. We are at this stage only looking for full-time employees who can work in our London office. We welcome recent graduates and applicants with non-traditional backgrounds who show strong programming skills and capacity for independent learning. Please email contact(a)beautifuldestinations.com with any questions or with a CV if you are interested in this opportunity. -- <https://instagram.com/beautifuldestinations/>

1 0

Talk at tonight's Haskell Meetup at the Hacker Dojo.
by David Banas 16 Oct '15

16 Oct '15

Hi all, I plan to give a short talk, at tonight’s Haskell Meetup at the Hacker Dojo in Mountain View, CA, on using Conal’s Circat and Lambda-CCC machinery to implement the Fast Fourier Transform (FFT). Hope to see you there, -db

2 4

Specializing argument to forM_, mapM_ & foldM_ in base-4.8
by Amit Aryeh Levy 15 Oct '15

15 Oct '15

I've been running into a relatively small but frequent annoyance with base >= 4.8 (GHC 7.10). `Control.Monad.foldM_`, `Control.Monad.mapM_` and `Control.Monad.forM_` are generalized traverse over any `Foldable a` rather than just arrays (`[a]`). This is great, except I'm finding that, for a lot of my code that works well in previous versions, I need to specialize the argument to `[a]` now. If other people are encoutering a similar patter, I wonder what are your best practices for doing this: ScopedTypeVariables? Deconstruct the reconstruct the array? ... The most common example is when I deserialize a JSON array with aeson and want to traverse over that array (say, to store the objects to a DB): ``` let objArray = eitherDecode myjson case objArray of Left err -> ... Right (objs :: [MyObjType]) -> forM_ objs $ \obj -> saveToDb obj ``` The above fix requires `ScopedTypeVariables` (which is probably OK). Another option is to deconstruct and reconstruct the list: ``` Right (o:objs) -> forM_ (o:objs) $ \obj -> saveToDb obj ``` Does this get optimized away? Penny for your thoughts? Cheers! Amit

3 5

Trying to write an Embedded DSL that threads a monad?
by Ian Bloom 15 Oct '15

15 Oct '15

I've been wrestling with this for a while and I decided eventually to look for help. I've been hoping to design a domain specific embedded language in Haskell that would let me pipe a commutative monad throughout an expression written in the language. Special terms within the language will eventually have access to this monad. I've created a simplified version here to represent the main issues. Here is what I'd like from the language: * To use haskell syntax for substitution and pattern matching rather than implementing this myself. * To be able to express lambdas in my language. * To be able to embed any haskell terms including functions into the language. * I'd like the Haskell type checker to tell me about bad terms. * I'd like to thread a monad through the entire expression. So here is the first implementation that I tried of this (full source here: http://lpaste.net/142959) data Exp m x where Val :: m x -> Exp m x Lam1 :: m (a -> Exp m b) -> Exp m (a -> b) Lam2 :: m (a -> Exp m (b -> c)) -> Exp m (a -> b -> c) a function liftE allows me to lift a haskell term into an expression: liftE x = Val $ return x Application is a function <@> so: (<@>) :: forall m a b. Monad m => Exp m (a -> b) -> Exp m a -> Exp m b (<@>) (Val f) (Val x) = Val $ f `ap` x (<@>) (Lam2 f) (Val x) = Lam1 $ f >>= \f' -> x >>= \x' -> unLam1 $ f' x' (<@>) (Lam1 f) (Val x) = Val $ f >>= \f' -> x >>= \x' -> unVal $ f' x' Seems like it might work! In fact it does typecheck. So the first test expressions I'd like to try are these: mapE :: Monad m => Exp m ((a -> b) -> [a] -> [b]) mapE = Lam2 $ return $ \ f -> Lam1 $ return $ \ xxs -> case xxs of [] -> liftE [] (x:xs) -> liftE (:) <@> liftE (f x) <@> (mapE <@> liftE f <@> liftE xs) testExpression :: Monad m => Exp m [Int] testExpression = mapE <@> liftE (+10) <@> liftE [1,2,3,4] and just to justify doing any of this I'll create a Monad called BindCounter that counts the number of times bind is called: newtype BindCounter a = BC (Int -> (Int, a)) runBC (BC f) = f 0 instance Functor (BindCounter) where fmap f (BC x) = BC $ \t -> let (t', x') = x t in (t', f x') instance Applicative (BindCounter) where pure x = BC (\t -> (t,x)) f <*> x = f >>= \f' -> x >>= \x' -> return $ f' x' instance Monad (BindCounter) where return = pure BC f >>= g = BC $ \old -> let (new, val) = f (old + 1) BC f' = g val in f' new now I can try a test test = let (count, result) = runBC (unVal testExpression) in putStrLn $ "Count: " ++ show count ++ " Result: " ++ show result The result in ghci is: Count: 36 Result: [11,12,13,14] Ok so that's a lot. I was surprised I got this working. You can see from the code that my main gripe with this is I haven't found a way to remove the need to specify the number of embedded lambdas using Lam1 and Lam2 (we could easily add more) and I haven't found a way to apply a Lam to another Lam. I'm also curious if I am reinventing the wheel, I hadn't found a library yet that let's me do something similar.

3 3

Re: [Haskell-cafe] Talk at tonight's Haskell Meetup at the Hacker Dojo.
by David Banas 15 Oct '15

15 Oct '15

Yes, you may attend without being a member of the Hacker Dojo. -db On Thu, Oct 15, 2015 at 10:27 AM, Andrew Gibiansky < andrew.gibiansky(a)gmail.com> wrote: > Can I come if I'm not a member of Hacker Dojo? The website is unclear. > > -- Andrew > > On Thu, Oct 15, 2015 at 7:33 AM, David Banas <capn.freako(a)gmail.com> > wrote: > >> Hi all, >> >> I plan to give a short talk, at tonight’s Haskell Meetup at the Hacker >> Dojo in Mountain View, CA, on using Conal’s Circat and Lambda-CCC machinery >> to implement the Fast Fourier Transform (FFT). >> >> Hope to see you there, >> -db >> >> _______________________________________________ >> Haskell-Cafe mailing list >> Haskell-Cafe(a)haskell.org >> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe >> > >

1 0