- Haskell-Cafe - Haskell.org

Re: [Haskell-cafe] partially applied data constructor and corresponding type
by Stephen Tetley 18 May '13

18 May '13

What you probably want are type level integers (naturals) Yury Sulsky used them in the message above - basically you can't use literal numbers 1,2,3,... etc as they are values of type Int (or Integer, etc...) instead you have to use type level numbers: data One data Two Work is ongoing for type level numbers in GHC and there are user libraries on Hackage so there is a lot of work to crib from.

3 4

question about GADT and deriving automatically a Show instance
by TP 18 May '13

18 May '13

Hi everybody, I have a question about deriving automatically a Show instance when using GADT. It works in this situation: ---------------------------- {-# LANGUAGE GADTs #-} data Male data Female data Person gender where Dead :: Person gender Alive :: { name :: String , weight :: Float , father :: Person gender } -> Person gender deriving Show main = do let a = Alive "Joe" 60 Dead :: Person Male let b = Alive "Jim" 70 a :: Person Male print a print b ---------------------------- Indeed: $ runghc test_show.hs Alive {name = "Joe", weight = 60.0, father = Dead} Alive {name = "Jim", weight = 70.0, father = Alive {name = "Joe", weight = 60.0, father = Dead}} But when I replace "father :: Person gender" by "father :: Person gender2" in the code (this is one of the advantages of GADT: with a classical algebraic data type declaration, gender2 would have to be a type variable for the data type), I obtain: Can't make a derived instance of `Show (Person gender)': Constructor `Alive' must have a Haskell-98 type Possible fix: use a standalone deriving declaration instead In the data declaration for `Person' So I modify my code by removing "deriving Show", and adding the line: ---------------------------- instance Show (Person gender) ---------------------------- But now, I obtain: $ runghc test_show.hs GHC stack-space overflow: current limit is 536870912 bytes. Use the `-K<size>' option to increase it. Why (I imagine this is because there is an infinite loop in the construction of the show function)? Is there any workaround? Thanks, TP

4 6

Compiler error in Parsec using ByteString
by mukesh tiwari 18 May '13

18 May '13

Hello all, I am trying to write a small calculator using Parsec. Every thing went fine and I wrote this [1] and now I am trying to use ByteString to make it more faster but getting compiler error which I am not able to figure out. Could some one please tell me what is wrong with this code. In case of indentation error, code on hpaste[2] {-# LANGUAGE OverloadedStrings #-} import qualified Text.Parsec.Token as Token import Text.Parsec.Prim import Text.Parsec.Char import Text.Parsec.Expr import Text.Parsec.Combinator import Text.Parsec.Language import qualified Text.Parsec.ByteString.Lazy as T import qualified Data.ByteString.Lazy.Char8 as BS import Control.Applicative hiding ( ( <|> ) , many ) import Data.Maybe ( fromJust ) languageDef = emptyDef { Token.commentStart = "/*" , Token.commentEnd = "*/" , Token.commentLine = "//" , Token.nestedComments = False , Token.identStart = letter <|> char '_' , Token.identLetter = alphaNum <|> oneOf "_'" , Token.opStart = Token.opLetter emptyDef , Token.opLetter = oneOf ":!#$%&*+./<=>?@\\^|-~" , Token.reservedOpNames= ["*" , "/" , "^" , "+" , "-" ] , Token.reservedNames = [] , Token.caseSensitive = True } lexer = Token.makeTokenParser languageDef identifier = Token.identifier lexer reserved = Token.reserved lexer operator = Token.operator lexer reservedOp = Token.reservedOp lexer natural = Token.natural lexer integer = Token.integer lexer float = Token.float lexer lexeme = Token.lexeme lexer parens = Token.parens lexer whiteSpace = Token.whiteSpace lexer semi = Token.semi lexer comma = Token.comma lexer colon = Token.colon lexer dot = Token.dot lexer semiSep = Token.semiSep lexer commaSep = Token.commaSep lexer data Expr = Num Integer | Add Expr Expr | Sub Expr Expr | Mul Expr Expr | Div Expr Expr | Exp Expr Expr deriving Show exprCal :: T.Parser Expr exprCal = buildExpressionParser table factor table = [ [ op "^" Exp AssocRight ] , [ op "*" Mul AssocLeft, op "/" Div AssocLeft ] , [ op "+" Add AssocLeft, op "-" Sub AssocLeft ] ] where op s f assoc = Infix ( reservedOp s >> return f ) assoc factor :: T.Parser Expr factor = parens exprCal <|> Num <$> integer calExpression :: Expr -> Integer calExpression ( Num n ) = n calExpression ( Add e1 e2 ) = calExpression e1 + calExpression e2 calExpression ( Sub e1 e2 ) = calExpression e1 - calExpression e2 calExpression ( Mul e1 e2 ) = calExpression e1 * calExpression e2 calExpression ( Div e1 e2 ) = div ( calExpression e1 ) ( calExpression e2 ) calExpression ( Exp e1 e2 ) = calExpression e1 ^ ( calExpression e2 ) calculator :: String -> Integer calculator expr = case parse ( whiteSpace >> exprCal ) "" ( BS.pack expr ) of Left msg -> error "failed to parse" Right ( val ) -> calExpression val *Main> :load "/Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs" [1 of 1] Compiling Main ( /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs, interpreted ) /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:56:36: Couldn't match type `[Char]' with `BS.ByteString' Expected type: OperatorTable BS.ByteString () Data.Functor.Identity.Identity Expr Actual type: [[Operator String () Data.Functor.Identity.Identity Expr]] In the first argument of `buildExpressionParser', namely `table' In the expression: buildExpressionParser table factor In an equation for `exprCal': exprCal = buildExpressionParser table factor /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:67:11: Couldn't match type `[Char]' with `BS.ByteString' Expected type: ParsecT BS.ByteString () Data.Functor.Identity.Identity Expr Actual type: ParsecT String () Data.Functor.Identity.Identity Expr In the return type of a call of `parens' In the first argument of `(<|>)', namely `parens exprCal' In the expression: parens exprCal <|> Num <$> integer /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:67:19: Couldn't match type `BS.ByteString' with `[Char]' Expected type: ParsecT String () Data.Functor.Identity.Identity Expr Actual type: T.Parser Expr In the first argument of `parens', namely `exprCal' In the first argument of `(<|>)', namely `parens exprCal' In the expression: parens exprCal <|> Num <$> integer /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:68:21: Couldn't match type `[Char]' with `BS.ByteString' Expected type: ParsecT BS.ByteString () Data.Functor.Identity.Identity Integer Actual type: ParsecT String () Data.Functor.Identity.Identity Integer In the second argument of `(<$>)', namely `integer' In the second argument of `(<|>)', namely `Num <$> integer' In the expression: parens exprCal <|> Num <$> integer /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:82:46: Couldn't match type `BS.ByteString' with `[Char]' Expected type: ParsecT String () Data.Functor.Identity.Identity Expr Actual type: T.Parser Expr In the second argument of `(>>)', namely `exprCal' In the first argument of `parse', namely `(whiteSpace >> exprCal)' In the expression: parse (whiteSpace >> exprCal) "" (BS.pack expr) /Users/mukeshtiwari/Programming/Code/Compilers/ParsecExample.hs:82:62: Couldn't match type `BS.ByteString' with `[Char]' Expected type: String Actual type: BS.ByteString In the return type of a call of `BS.pack' In the third argument of `parse', namely `(BS.pack expr)' In the expression: parse (whiteSpace >> exprCal) "" (BS.pack expr) Failed, modules loaded: none. Prelude> Mukesh Tiwari [1] http://hpaste.org/88155 [2] http://hpaste.org/88156

2 2

Extensible Records using Implicit Parameters
by Thomas Jaeger 18 May '13

18 May '13

Hello, Records in Haskell are somewhat of a contentious issue and many proposals have been put forth to address the shortcomings of the current record system [1]. Below, I introduce a small library relying on several GHC extensions, crucially Implicit Parameters and Constraint Kinds, which implements an extensible record system. While by no means production-ready, it is remarkably close to how I think an extensible record system should function. Record access is very convenient (reminiscent of Pascal's with statement), while record update is somewhat cumbersome (but could potentially be improved using Template Haskell). Two caveats: * Type inference for records does not work, so type signatures have to be provided. * When more than one binding for an implicit parameter is in scope, it is not always clear which one takes precedence. However, I think it is safe to assume that if the innermost binding is a let/where binding, it will carry the day. In all other cases, it is probably safest to rely on explicit type signatures to resolve the ambiguity. > {-# LANGUAGE ImplicitParams #-} > {-# LANGUAGE ConstraintKinds #-} > {-# LANGUAGE GADTs #-} > {-# LANGUAGE RankNTypes #-} The record type is simply the dictionary type, which is more commonly used to reify type class dictionaries. Here, we use it to record a number of implicit parameter bindings, i.e. if implicit parameters of types `fields' are in scope, then the data constructor `Rec' will create a dictionary from them, provided an appropriate type signature is given. > data Record fields where > Rec :: fields => Record fields The `?' operator will bring the implicit parameter bindings previously captured by `Rec' into scope. > infixr 1 ? > (?) :: Record fields -> (fields => r) -> r > Rec ? e = e This is the complete library. Examples follow below. ======================================================================== Type synonyms to replace the awkward syntax for implicit parameter constraints. > type X a = ?x::a > type Y a = ?y::a > type Z a = ?z::a As a first example, construct a record with fields ?x and ?y, both of type Int. > xy :: Record (X Int, Y Int) > xy = let ?x = 1; ?y = 2 in Rec To access the fields of the record, we can use the `?' operator. Expressions involving any of the fields can appear to the right of the operator. * xy ? ?x ==> 1 * xy ? ?x + ?y ==> 3 The next example illustrates record update > xy' :: Record (X Int, Y Int) > xy' = xy ? let ?y = -2 in Rec * xy' ? ?y ==> -2 Type signatures are required, but can be placed directly after `Rec'. > xz = let ?x = -1; ?z = 3 in Rec :: Record (X Int, Z Int) Record Restriction > y :: Record (Y Int) > y = xy ? Rec We can combine records as follows, but the result may be implementation- dependent if both records share fields. > xyz, xyz' :: Record (X Int, Y Int, Z Int) > xyz = xz ? xy ? Rec * xyz ? ?x ==> -1 So it is probably better to disambiguate. > xyz' = xz ? y ? Rec * xyz' ? ?x ==> -1 ======================================================================== As an application, we use records to build a rudimentary object system. A class is a record which can access its own fields, which are supplied as an argument. > type Class fields = Record fields -> Record fields Originally I tried the type 'fields => Record fields', which leads to problems when tying the knot. Given a class, we can get a record for property and method access by using a fixed point combinator. > runClass :: Class fields -> Record fields > runClass inst = inst (runClass inst) Type signatures for our next example. > type N = ?n::Int > type AbstractFact = (?fact::Int, ?factHelper::Int -> Int -> Int) > type ConcreteFact = (AbstractFact, N) This is an abstract class that computes the factorial of its abstract property `?n'. Notice how `?n' is not included in the function's return type. > abstractFact :: Record ConcreteFact -> Record AbstractFact > abstractFact r = Rec where > ?fact = r ? ?factHelper 1 ?n > ?factHelper = r ? > \p k -> if k == 1 then p else ?factHelper (p*k) (k-1) It is important that `r' only be opened inside the definitions of the methods, otherwise `runClass' will cause an infinite loop. To go from an abstract class to a concrete class, all we need to do is provide the missing property. > concreteFact :: Int -> Class ConcreteFact > concreteFact n r = abstractFact r ? let ?n = n in Rec Quick sanity check > testFact :: Class ConcreteFact > testFact = concreteFact 10 * runClass testFact ? ?fact ==> 3628800 Example of overloading. > overrideFact :: Class ConcreteFact > overrideFact r = testFact r ? let ?fact = fact in Rec where > fact = r ? product [1.. ?n] It is tempting to use the following definition overrideFact r = testFact r ? let ?fact = product [1.. ?n] in Rec However, this will use `?n' from the current environment and not respond to `?n' being overridden in a subclass. The final example concerns multiple inheritance. We first define a class for computing Fibonacci numbers, similar to the one for the factorial above. > type Fib = (?fib::Int, N) > fib :: Int -> Class Fib > fib n r = Rec where > ?n = n > ?fib = let fibs = 0:zipWith (+) fibs (1:fibs) in r ? fibs !! ?n A multiple-inheritance combinator > diamond :: Class c1 -> Class c2 -> Class (c1,c2) > diamond c1 c2 r = c1 (r ? Rec) ? c2 (r ? Rec) ? Rec > testFactFib :: Class (ConcreteFact, Fib) > testFactFib = diamond testFact (fib 5) This is C++-style inheritance: We now have two copies of the field `?n'. * runClass testFactFib ? ?fact ==> 3628800 * runClass testFactFib ? ?fib ==> 5 However, if we were to update `?n' in a subclass, both the factorial and the Fibonacci class would both use the new value. Alternatively, we can force the class to use a single value for `?n' by making sure `?n' only occurs once in the list of fields, but it's anyone's guess which value GHC will pick for `?n'. > testFactFib' :: Class (AbstractFact, Fib) > testFactFib' r = testFactFib (r ? Rec) ? Rec * runClass testFactFib' ? ?fact ==> 120 * runClass testFactFib' ? ?fib ==> 5 Thomas [1] http://hackage.haskell.org/trac/ghc/wiki/Records

1 0

A use case for *real* existential types
by oleg＠okmij.org 18 May '13

18 May '13

But Haskell (and GHC) have existential types, and your prototype code works with GHC after a couple of trivial changes: > main = do > W nd0 <- init > wd0 <- addWatch nd0 "foo" > wd1 <- addWatch nd0 "bar" > W nd1 <- init > wd3 <- addWatch nd1 "baz" > printInotifyDesc nd0 > printInotifyDesc nd1 > rmWatch nd0 wd0 > rmWatch nd1 wd3 > -- These lines cause type errors: > -- rmWatch nd1 wd0 > -- rmWatch nd0 wd3 > printInotifyDesc nd0 > printInotifyDesc nd1 The only change is that you have to write W nd <- init and that's all. The type-checker won't let the user forget about the W. The commented out lines do lead to type errors as desired. Here is what W is > data W where > W :: Inotify a -> W > init :: IO W [trivial modification to init's code] I must say though that I'd rather prefer Adres solution because his init > init :: (forall a. Inotify a -> IO b) -> IO b ensures that Inotify does not leak, and so can be disposed of at the end. So his init enforces the region discipline and could, after a trivial modification to the code, automatically do a clean-up of notify descriptors -- something you'd probably want to do.

2 2

Propositions in Haskell
by Patrick Browne 17 May '13

17 May '13

6 8

Lambda expressions (core) to categorical form
by Conal Elliott 16 May '13

16 May '13

I want to convert lambda expressions into a vocabulary of monoidal categories, so that they can be given multiple interpretations, including circuit generation and timing analysis, and hopefully some other far-out alternatives (3D visualization, animated evaluation, etc). More specifically, I want a GHC plugin that makes this transformation on GHC's Core language. If you know of related work, have suggestions, and/or are interested in collaborating/consulting, I'd love to hear. Thanks, - Conal

1 0

List comprehensions with Word8
by Jose A. Lopes 16 May '13

16 May '13

Hello everyone, I was playing with Word8 and list comprehensions and the following examples came up. I have to admit the behavior looks quite strange because it does not seem to be consistent. Can someone shed some light on reason behind some of these outputs? By the way, I have abbreviated some outputs with ellipsis ... [1..10] :: [Word8] [1,2,3,4,5,6,7,8,9,10] [1..100] :: [Word8] [1,2,3,4,5,6,7,8,9,10,...,100] [1..1000] :: [Word8] [1,2,3,4,5,6,7,8,9,10,...,232] [1..10000] :: [Word8] [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16] [1..100000] :: [Word8] [1,2,3,4,5,6,7,8,9,10,...,160] [1..1000000] :: [Word8] [1,2,3,4,5,6,7,8,9,10,...,64] [1..10000000] :: [Word8] [1,2,3,4,5,6,7,8,9,10,...,128] [1..100000000] :: [Word8] [] [1..1000000000] :: [Word8] [] Thank you, Jose -- José António Branquinho de Oliveira Lopes Instituto Superior Técnico Technical University of Lisbon

5 4

Re: [Haskell-cafe] Infrastructure for testing the impact of a Functor/Applicative/Monad hierarchy
by Stephen Tetley 16 May '13

16 May '13

Has anyone surveyed the in-print textbooks, tutorials, or tried to assess how much Haskell (H98, H2010, Glasgow Haskell?) is used in teaching? Having the wrong hierarchy is a minor annoyance to us members of the cognoscenti, but a change outside a revision of the language standard could leave a lot of beginners and the teaching material they rely on stranded.

2 1

Parallel ghc --make
by Niklas Hambüchen 16 May '13

16 May '13

I know this has been talked about before and also a bit in the recent GSoC discussion. I would like to know what prevents ghc --make from working in parallel, who worked at that in the past, what their findings were and a general estimation of the difficulty of the problem. Afterwards, I would update http://hackage.haskell.org/trac/ghc/ticket/910 with a short summary of what the current situation is. Thanks to those who know more!

5 6