- Haskell-Cafe - Haskell.org

Infrastructure for testing the impact of a Functor/Applicative/Monad hierarchy
by Niklas Hambüchen 22 May '13

22 May '13

Reading the other thread (Adding Applicative/Functor instances to all Monads in GHC) I was wondering if there was infrastructure for testing what effect making the often-discussed Functor/Monad change would have: How many packages on hackage would break etc. I have read a few times that people have "compiled all of hackage" to see the impact of whatever. How do you do that? Do you just run a loop around cabal install or have you built some more advanced tools to visualize the results better or compile the packages "from ground up", in order of their dependencies? I'm interested in anything in this direction.

3 3

Teaching FP with Haskell
by Rustom Mody 22 May '13

22 May '13

We are offering a MOOC on haskell : https://moocfellowship.org/submissions/the-dance-of-functional-programming-… Full Announcement on beginners list : http://www.haskell.org/pipermail/beginners/2013-May/012013.html One question that I have been grappling with in this regard: How to run ghc in lightweight/beginner mode? 2 examples of what I mean: 1. gofer used to come with an alternative standard prelude -- 'simple.pre' Using this, gofer would show many of the type-class based errors as simple (non-type-class based) errors. This was very useful for us teachers to help noobs start off without intimidating them. 2. Racket comes with a couple of levels. The easier numbers were not completely consistent with scheme semantics, but was gentle to beginners Any thoughts/inputs on this will be welcomed Rusi

4 3

parsing a CSV file
by Roger Mason 22 May '13

22 May '13

Hello, I'm attempting to write a parser for files that look like this: Bruker Nano GmbH Berlin, Germany Esprit 1.9 Date: 02/05/2013 10:06:49 AM Real time: 15000 Energy Counts -0.474 0 ..... The line before the ellipsis is repeated many times (such lines represents a spectrum). I need to be able to extract numbers from lines containing <string: > and I want to extract the number pairs following "Energy Counts\n". The extracted data will then be written to a file in a different format. For now I'll be satisfied with reading the "header" info, i.e. down to "Energy Counts\n". Thus far, I have: -- derived from RWH -- file: ch16/csv2.hs import Text.ParserCombinators.Parsec headerLines = endBy csvFile endHeader csvFile = endBy line eol line = sepBy cell (char ',') cell = many (noneOf ",\n") eol = char '\n' parseCSV :: String -> Either ParseError [[String]] parseCSV input = parse csvFile "(unknown)" input parseHDR :: String -> Either ParseError [[String]] parseHDR input = parse headerLines "(unknown)" input endHeader = string "Energy Counts" This loads into GHCi (7.6.2) OK. However, when I test it: parseHDR "Bruker Nano GmbH Berlin, Germany\nEsprit 1.9\n\nDate: 02/05/2013 10:06:49 AM\nReal time: 15000\nEnergy Counts" Not in scope: `parseHDR' which makes sense because ghci> :t endHeader <interactive>:1:1: Not in scope: `endHeader' Clearly, my naiive implementation of endHeader is no good. I appreciate any pointers. Thanks, Roger This electronic communication is governed by the terms and conditions at http://www.mun.ca/cc/policies/electronic_communications_disclaimer_2012.php

2 4

Announce: vsim the VHDL-subset simulator (unmaintained)
by Sergey Mironov 22 May '13

22 May '13

Hi. I'd like to announce the vsim - the simulator for (small subset of) VHDL language. Currently the project contains pretty large Java part and is not maintained. That's why I don't want to publish it on Hackage. Still, as far as I now have a permission to share the code, I would like to do it. https://github.com/grrwlf/vsim VSim is a VHDL simulator project aimed develop a methods of compiling VHDL code into a high-level language (Haskell). Currently it is able to compile simple VHDL programs, containing plain integer types, 1-dimentional arrays, records. It supports processes, procedures, functions (partly). Breakpoints and wait statements should work. Simulator compile VHDL into Haskell in several steps. Firstly, VHDL is translated into VIR-file by a translator written in Java (see tr/ folder and runtr function in ./simenv shell script). VIR file is lisp-like file, describing vhdl entities in a less complex manner. For example, it contains all the port maps expanded. Secondly, VSim tool is used to translate VIR into Haskell (see src/ folder and runsim functino in ./simenv). VSim is a small program which translates VIR line-by-line into Haskell program. haskell-src-exts is used to build the AST and print it to stdout. Finaly, Haskell program should be compiled into binary with ghc having runtime library included (refer to src\_r/ folder and runsim function in ./simenv). The runtime is the heart of the simulator and it's largest part. Java part is a main headache, because it is big, unsupported and has bugs lurking here and there. Haskell part is smaller and cleaner, but VHDL-standard coverage is still poor. For example, signal assignments are working as if they are declared with transport delay mechanism. Another problem is missing enum support except of some pre-defined. Sergey

1 0

Re: [Haskell-cafe] rip in the class-abstraction continuum
by oleg＠okmij.org 21 May '13

21 May '13

Type classes are the approach to constrain type variables, to bound polymorphism and limit the set of types the variables can be instantiated with. If we have two type variables to constrain, multi-parameter type classes are the natural answer then. Let's take this solution and see where it leads to. Here is the original type class > class XyConv a where > toXy :: a b -> [Xy b] and the problematic instance > data CircAppr a b = CircAppr a b b -- number of points, rotation angle, radius > deriving (Show) > > instance Integral a => XyConv (CircAppr a) where > toXy (CircAppr divns ang rad) = > let dAng = 2 * pi / (fromIntegral divns) in > let angles = map ((+ ang) . (* dAng) . fromIntegral) [0..divns] in > map (\a -> am2xy a rad) angles To be more explicit, the type class declaration has the form > class XyConv a where > toXy :: forall b. a b -> [Xy b] with the type variable 'b' universally quantified without any constraints. That means the user of (toXy x) is free to choose any type for 'b' whatsoever. Obviously that can't be true for (toXy (CircAppr x y)) since we can't instantiate pi to any type. It has to be a Floating type. Hence we have to constrain b. As I said, the obvious solution is to make it a parameter of the type class. We get the first solution: > class XYConv1 a b where > toXy1 :: a b -> [Xy b] > > instance (Floating b, Integral a) => XYConv1 (CircAppr a) b where > toXy1 (CircAppr divns ang rad) = > let dAng = 2 * pi / (fromIntegral divns) in > let angles = map ((+ ang) . (* dAng) . fromIntegral) [0..divns] in > map (\a -> am2xy a rad) angles The type class declaration proclaims that only certain combinations of 'a' and 'b' are admitted to the class XYConv1. In particular, 'a' is (CircAppr a) and 'b' is Floating. This reminds us of collections (with Int keys, for simplicity) > class Coll c where > empty :: c b > insert :: Int -> b -> c b -> c b > > instance Coll M.IntMap where > empty = M.empty > insert = M.insert The Coll declaration assumes that a collection is suitable for elements of any type. Later on one notices that if elements are Bools, they can be stuffed quite efficiently into an Integer. If we wish to add ad hoc, efficient collections to the framework, we have to restrict the element type as well: > class Coll1 c b where > empty1 :: c > insert1 :: Int -> b -> c -> c Coll1 is deficient since there is no way to specify the type of elements for the empty collection. When the type checker sees 'empty1', how can it figure out which instance for Coll1 (with the same c but different element types) to choose? We can help the type-checker by declaring (by adding the functional dependency c -> b) that for each collection type c, there can be only one instance of Coll1. In other words, the collection type determines the element type. Exactly the same principle works for XYConv. > class XYConv2 a b | a -> b where > toXy2 :: a -> [Xy b] > > instance (Floating b, Integral a) => XYConv2 (CircAppr a b) b where > toXy2 (CircAppr divns ang rad) = > let dAng = 2 * pi / (fromIntegral divns) in > let angles = map ((+ ang) . (* dAng) . fromIntegral) [0..divns] in > map (\a -> am2xy a rad) angles The third step is to move to associated types. At this stage you can consider them just as a different syntax of writing functional dependencies: > class XYConv3 a where > type XYT a :: * > toXy3 :: a -> [Xy (XYT a)] > > instance (Floating b, Integral a) => XYConv3 (CircAppr a b) where > type XYT (CircAppr a b) = b > toXy3 (CircAppr divns ang rad) = > let dAng = 2 * pi / (fromIntegral divns) in > let angles = map ((+ ang) . (* dAng) . fromIntegral) [0..divns] in > map (\a -> am2xy a rad) angles The step from XYConv2 to XYConv3 is mechanical. The class XYConv3 assumes that for each convertible 'a' there is one and only Xy type 'b' to which it can be converted. This was the case for (CircAppr a b). It may not be the case in general. But we can say that for each convertible 'a' there is a _class_ of Xy types 'b' to which they may be converted. This final step brings Tillmann Rendel's solution.

1 0

ghc-mod v2.0.1
by Kazu Yamamoto 21 May '13

21 May '13

Hi cafe! I have released ghc-mod v2.0.1. From this version, ghc-mod provides the ghc-mod library in addition to the ghc-mod command: http://hackage.haskell.org/package/ghc-mod http://mew.org/~kazu/proj/ghc-mod/en/ Enjoy! --Kazu

3 4

type-level integers for GHC
by Takayuki Muranushi 21 May '13

21 May '13

What is your recommendation for type-level integers? I'd like to use it to improve the unittyped, https://bitbucket.org/xnyhps/haskell-unittyped/ the library for physical dimension. Therefore, I need negative numbers, additions, but multiplications are not necessary. I've been looking forward for the type-nats extension of GHC, but I haven't been able to compile the type-nats branch. Just learned that it still takes a few month to merge the branch into the main. http://hackage.haskell.org/trac/ghc/wiki/Status/May13 Thijs, the original author of unittyped, has commited a branch that uses type-nats, but I can't try that out for the same reason. Best, -- Takayuki MURANUSHI The Hakubi Center for Advanced Research, Kyoto University http://www.hakubi.kyoto-u.ac.jp/02_mem/h22/muranushi.html

3 5

rip in the class-abstraction continuum
by Christopher Howard 21 May '13

21 May '13

Hi. I won't pretend to be an advanced Haskell programmer. However, I have a strong interest in abstraction, and I have been playing around with programming as abstractly as possible. Naturally, I find classes to be quite attractive and useful. However, something is bothering me. Lately I keep running into this situation where I have to cut across abstraction layers in order to make the code work. More specifically, I keep finding that I have to add constraints to a class definition, because eventually I find some instance of that class which needs those constraints to compile. For example, I wanted to create a class which represents all things that can be converted into X,Y coordinates. Naturally, I would like to do something like this: code: -------- class XyConv a where toXy :: a b -> [Xy b] -------- This leaves me free in the future to use any number type conceivable in the Xy coordinates - Floating or Integral types, or whatever. (Doesn't even have to be numbers, actually!) However the first instance I create, requires me to use operators in the function definition which require at least a Floating type. (The error will say Fractional, but there are other components that also require Floating.) code: -------- data CircAppr a b = CircAppr a b b -- number of points, rotation angle, radius deriving (Show) instance Integral a => XyConv (CircAppr a) where toXy (CircAppr divns ang rad) = let dAng = 2 * pi / (fromIntegral divns) in let angles = map ((+ ang) . (* dAng) . fromIntegral) [0..divns] in map (\a -> am2xy a rad) angles -------- This gives me the error code: -------- Could not deduce (Fractional b) arising from a use of `/' from the context (Integral a) bound by the instance declaration at /scratch/cmhoward/pulse-spin/pulse-spin.hs:51:10-42 Possible fix: add (Fractional b) to the context of the type signature for toXy :: CircAppr a b -> [Xy b] or the instance declaration In the expression: 2 * pi / (fromIntegral divns) In an equation for `dAng': dAng = 2 * pi / (fromIntegral divns) In the expression: let dAng = 2 * pi / (fromIntegral divns) in let angles = map ((+ ang) . (* dAng) . fromIntegral) [0 .. divns] in map (\ a -> am2xy a rad) angles -------- I can get a quick fix by adding Floating to the context of the /class/ definition: code: -------- class XyConv a where toXy :: Floating b => a b -> [Xy b] -------- But what I really want is to put Floating in the context of the /instance/ declaration. But I don't know how to do that syntactically. -- frigidcode.com

2 2

Failure deriving MonadRWS when using a type-family for the State part
by Nicolas Trangez 21 May '13

21 May '13

All, The following code results in a compilation error (I tried GHC 7.4.1 & a 7.7.20130430 build): {-# LANGUAGE TypeFamilies, GeneralizedNewtypeDeriving, StandaloneDeriving #-} module Main where import Control.Applicative import Control.Monad.RWS data C = C data E = E data S1 = S1 Int type family I a :: * type instance I S1 = Int newtype T a s m r = T { unT :: RWST C [E] (I s) m r } deriving ( Functor , Applicative , Monad , MonadReader C , MonadWriter [E] , MonadState (I s) , MonadRWS C [E] (I s) , MonadTrans ) Error: No instance for (MonadState (I s) (T a s m)) arising from the 'deriving' clause of a data type declaration Possible fix: use a standalone 'deriving instance' declaration, so you can specify the instance context yourself When deriving the instance for (MonadRWS C [E] (I s) (T a s m)) Commenting out the MonadRWS line from the derivings list (i.e. the line pointed at by the error) works as expected. I was somehow unable to get a suitable standalone-deriving clause working, so didn't test that. Is this expected? Regards, Nicolas

1 1

Stuck on design problem
by Nicolas Trangez 21 May '13

21 May '13

All, Since I'm stuck on a coding problem and can't figure out how to proceed, I decided to call for help. I'm unable to get some code to typecheck, so maybe I'm missing something obvious on the implementation side, or more likely the design is completely wrong. Here's the idea: I have some code which uses some log of "Entry a" values (for some polymorphic type 'a'), and provides actions which output "Command a" values which should be interpreted by some wrapper code, e.g. adding some new entries to the log. These actions are implemented in a custom WriterT transformer (in this demo code, the original is more complex) to output Commands, while the inner monad should give access to the log entries (this could be in-memory like in the example, or stored on disk so using IO). You can find a trimmed-down version at https://gist.github.com/NicolasT/4230251f4f87f110d197 This doesn't type-check, and I'm not sure how to proceed. Am I taking a wrong approach? Do I need a different design? Thanks, Nicolas

3 4