Haskell function signature and "Could not deduce" compiler error - haskell

I'm writing a Haskell library which uses Data.Vector's. I successfully wrote library function, but I don't know how to add signature to it. Below is a simple example illustrating the problem:
import qualified Data.Vector.Generic as V
-- zip two vectors and return first element as a tuple
test :: (V.Vector v a, Fractional a) => v a -> v a -> (a, a)
test a b = (V.zip a b) V.! 0
This code causes following compilation error:
Could not deduce (V.Vector v (a, a))
from the context (V.Vector v a, Fractional a)
arising from a use of `V.zip' at MyLib.hs:7:12-20
Possible fix:
add (V.Vector v (a, a)) to the context of
the type signature for `test'
or add an instance declaration for (V.Vector v (a, a))
In the first argument of `(V.!)', namely `(V.zip a b)'
In the expression: (V.zip a b) V.! 0
In the definition of `test': test a b = (V.zip a b) V.! 0
Code is complied if I comment out the signature of test function. What is a correct signature here?
I'm using GHC 6.12.3, vector library 0.7.0.1.
Thanks.

ghci says:
Prelude Data.Vector.Generic> :t \a b -> (Data.Vector.Generic.zip a b) Data.Vector.Generic.! 0
\a b -> (Data.Vector.Generic.zip a b) Data.Vector.Generic.! 0
:: (Vector v a, Vector v b, Vector v (a, b)) =>
v a -> v b -> (a, b)
Matching with your case, the signature should be
test :: (V.Vector v a, Fractional a, V.Vector v (a, a)) => v a -> v a -> (a, a)
(oh and you need FlexibleContexts)

Related

Problem with 'rigid type variable' in a Monad in haskell

I'm trying to learn monads better and am playing around with it in Haskell. I defined a monad in this way:
module TESTMonad where
import Control.Monad
newtype TEST i = TEST {getTEST :: ((i, Int), Int)} deriving (Show, Eq, Ord)
instance Functor TEST where
fmap f (TEST ((x,y), z)) = TEST ((f x, y), z)
instance Applicative TEST where
pure = return
tf <*> tx = tf >>= \f -> tx >>= \x -> return (f x)
instance Monad TEST where
return x = TEST ((x, 1), 1)
(TEST ((x, y), z)) >>= f = TEST ((plusOne a, b), c)
where
((a, b), c) = getTEST (f x)
plusOne :: Int -> Int
plusOne x = x+1
but I get the following error when I'm trying to compile it:
TESTMonad.hs:16:47: error:
• Couldn't match expected type ‘Int’ with actual type ‘b’
‘b’ is a rigid type variable bound by
the type signature for:
(>>=) :: forall a b. TEST a -> (a -> TEST b) -> TEST b
at TESTMonad.hs:16:24
• In the first argument of ‘plusOne’, namely ‘a’
In the expression: plusOne a
In the expression: (plusOne a, b)
• Relevant bindings include
a :: b (bound at TESTMonad.hs:18:19)
f :: a -> TEST b (bound at TESTMonad.hs:16:28)
(>>=) :: TEST a -> (a -> TEST b) -> TEST b
(bound at TESTMonad.hs:16:5)
Failed, modules loaded: none.
I clearly know that I might be doing lots of things in the wrong way but I have no idea what are they. Any comment would be appreciated. Thank you in advance!
The Monad instance cannot be constrained. The type of (>>=) must be
Monad m => m a -> (a -> m b) -> m b
but your definition, using plusOne :: Int -> Int, makes the type
Monad m => m Int -> (Int -> m Int) -> m Int
You could safely apply plusOne to either of the other values wrapped inside TEST, as they are already defined to be Ints.
The definition of (>>=) has no idea what the type of x might be, and the caller gets to choose f, so it has no idea what the type of f x might either. As a result, you can't really do anything with it except use it as-is.

Coercible and existential

data T t where
A :: Show (t a) => t a -> T t
B :: Coercible Int (t a) => t a -> T t
f :: T t -> String
f (A t) = show t
g :: T t -> Int
g (B t) = coerce t
Why does f compile but g generate an error like follows? I'm using GHC 8.4.
• Couldn't match representation of type ‘Int’ with that of ‘t a’
Inaccessible code in
a pattern with constructor:
B :: forall k (t :: k -> *) (a :: k).
Coercible Int (t a) =>
t a -> T t,
in an equation for ‘g’
• In the pattern: B t
In an equation for ‘g’: g (B t) = coerce t
Also, are Coercible constraints zero-cost even when they are embedded in GADTs?
UPD: Compiler bug: https://ghc.haskell.org/trac/ghc/ticket/15431
As a workaround, you may replace the constraint (which is not free in the first place) with a Data.Type.Coercion.Coercion (which adds an extra data wrapper around the dictionary).
data T t where
A :: Show (t a) => t a -> T t
B :: !(Coercion Int (t a)) -> t a -> T t
-- ! for correctness: you can’t have wishy-washy values like B _|_ (I "a")
-- Such values decay to _|_
f :: T t -> String
f (A x) = show x
f (B c x) = show (coerceWith (sym c) x)
newtype I a = I a
main = putStrLn $ f $ B Coercion $ I (5 :: Int)
GHC 8.6 will improve this situation in two ways:
Your original code will work, as the underlying bug was fixed.
The Coercion can be unpacked to a Coercible constraint, and this will happen automatically, due to -funbox-small-strict-fields. Thus, this T will get performance characteristics equivalent to your original for free.

When is forall not for all

In the program below test₁ will not compile but test₂ will. The reason seems to be because of the forall s. in withModulus₁. It seems that the s is a different type for each and every call to withModulus₁ because of the forall s.. Why is that the case?
{-# LANGUAGE
GADTs
, KindSignatures
, RankNTypes
, TupleSections
, ViewPatterns #-}
module Main where
import Data.Reflection
newtype Modulus :: * -> * -> * where
Modulus :: a -> Modulus s a
deriving (Eq, Show)
newtype M :: * -> * -> * where
M :: a -> M s a
deriving (Eq, Show)
add :: Integral a => Modulus s a -> M s a -> M s a -> M s a
add (Modulus m) (M a) (M b) = M (mod (a + b) m)
mul :: Integral a => Modulus s a -> M s a -> M s a -> M s a
mul (Modulus m) (M a) (M b) = M (mod (a * b) m)
unM :: M s a -> a
unM (M a) = a
withModulus₁ :: a -> (forall s. Modulus s a -> w) -> w
withModulus₁ m k = k (Modulus m)
withModulus₂ :: a -> (Modulus s a -> w) -> w
withModulus₂ m k = k (Modulus m)
test₁ = withModulus₁ 89 (\m ->
withModulus₁ 7 (\m' ->
let
a = M 131
b = M 127
in
unM $ add m' (mul m a a) (mul m b b)))
test₂ = withModulus₂ 89 (\m ->
withModulus₂ 7 (\m' ->
let
a = M 131
b = M 127
in
unM $ add m' (mul m a a) (mul m b b)))
Here is the error message:
Modulus.hs:41:29: error:
• Couldn't match type ‘s’ with ‘s1’
‘s’ is a rigid type variable bound by
a type expected by the context:
forall s. Modulus s Integer -> Integer
at app/Modulus.hs:(35,9)-(41,52)
‘s1’ is a rigid type variable bound by
a type expected by the context:
forall s1. Modulus s1 Integer -> Integer
at app/Modulus.hs:(36,11)-(41,51)
Expected type: M s1 Integer
Actual type: M s Integer
• In the second argument of ‘add’, namely ‘(mul m a a)’
In the second argument of ‘($)’, namely
‘add m' (mul m a a) (mul m b b)’
In the expression: unM $ add m' (mul m a a) (mul m b b)
• Relevant bindings include
m' :: Modulus s1 Integer (bound at app/Modulus.hs:36:28)
m :: Modulus s Integer (bound at app/Modulus.hs:35:27)
|
41 | unM $ add m' (mul m a a) (mul m b b)))
| ^^^^^^^^^
Briefly put, a function
foo :: forall s . T s -> U s
lets its caller to choose what the type s is. Indeed, it works on all types s. By comparison,
bar :: (forall s . T s) -> U
requires that its caller provides an argument x :: forall s. T s, i.e. a polymorphic value that will work on all types s. This means that bar will choose what the type s will be.
For instance,
foo :: forall a. a -> [a]
foo x = [x,x,x]
is obvious. Instead,
bar :: (forall a. a->a) -> Bool
bar x = x 12 > length (x "hello")
is more subtle. Here, bar first uses x choosing a ~ Int for x 12, and then uses x again choosing a ~ String for x "hello".
Another example:
bar2 :: Int -> (forall a. a->a) -> Bool
bar2 n x | n > 10 = x 12 > 5
| otherwise = length (x "hello") > 7
Here a is chosen to be Int or String depending on n > 10.
Your own type
withModulus₁ :: a -> (forall s. Modulus s a -> w) -> w
states that withModulus₁ must be allowed to choose s to any type it wishes. When calling this as
withModulus₁ arg (\m -> ...)
m will have type Modulus s0 a where a was chosen by the caller, while s was chosen by withModulus₁ itself. It is required that ... must be compatible with any choice withModulus₁ may take.
What if we nest calls?
withModulus₁ arg (\m1 -> ...
withModulus₁ arg (\m2 -> ...)
...
)
Now, m1 :: Modulus s0 a as before. Further m2 :: Modulus s1 a where s1 is chosen by the innermost call to withModulus₁.
The crucial point, here, is that there is no guarantee that s0 is chosen to be the same as s1. Each call might make a different choice: see e.g. bar2 above which indeed does so.
Hence, the compiler can not assume that s0 and s1 are equal. Hence, if we call a function that requires their equality, like add, we get a type error, since this would constrain the freedom of choice of s by the two withModulus₁ calls.

Pattern matching on rank-2 type

I'm trying to understand why one version of this code compiles, and one version does not.
{-# LANGUAGE RankNTypes, FlexibleContexts #-}
module Foo where
import Data.Vector.Generic.Mutable as M
import Data.Vector.Generic as V
import Control.Monad.ST
import Control.Monad.Primitive
data DimFun v m r =
DimFun {dim::Int, func :: v (PrimState m) r -> m ()}
runFun1 :: (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun1 (DimFun dim t) x | V.length x == dim = runST $ do
y <- thaw x
t y
unsafeFreeze y
runFun2 :: (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun2 t x = runST $ do
y <- thaw x
evalFun t y
unsafeFreeze y
evalFun :: (PrimMonad m, MVector v r) => DimFun v m r -> v (PrimState m) r -> m ()
evalFun (DimFun dim f) y | dim == M.length y = f y
runFun2 compiles fine (GHC-7.8.2), but runFun1 results in errors:
Could not deduce (PrimMonad m0) arising from a pattern
from the context (Vector v r, MVector (Mutable v) r)
bound by the type signature for
tfb :: (Vector v r, MVector (Mutable v) r) =>
(forall (m :: * -> *). PrimMonad m => TensorFunc m r) -> v r -> v r
at Testing/Foo.hs:(26,8)-(28,15)
The type variable ‘m0’ is ambiguous
Note: there are several potential instances:
instance PrimMonad IO -- Defined in ‘Control.Monad.Primitive’
instance PrimMonad (ST s) -- Defined in ‘Control.Monad.Primitive’
In the pattern: TensorFunc _ f
In an equation for ‘tfb’:
tfb (TensorFunc _ f) x
= runST
$ do { y <- thaw x;
f y;
unsafeFreeze y }
Couldn't match type ‘m0’ with ‘ST s’
because type variable ‘s’ would escape its scope
This (rigid, skolem) type variable is bound by
a type expected by the context: ST s (v r)
at Testing/Foo.hs:(29,26)-(32,18)
Expected type: ST s ()
Actual type: m0 ()
Relevant bindings include
y :: Mutable v s r (bound at Testing/Foo.hs:30:3)
f :: forall (v :: * -> * -> *).
MVector v r =>
v (PrimState m0) r -> m0 ()
(bound at Testing/Foo.hs:29:19)
In a stmt of a 'do' block: f y
In the second argument of ‘($)’, namely
‘do { y <- thaw x;
f y;
unsafeFreeze y }’
Could not deduce (s ~ PrimState m0)
from the context (Vector v r, MVector (Mutable v) r)
bound by the type signature for
tfb :: (Vector v r, MVector (Mutable v) r) =>
(forall (m :: * -> *). PrimMonad m => TensorFunc m r) -> v r -> v r
at Testing/Foo.hs:(26,8)-(28,15)
‘s’ is a rigid type variable bound by
a type expected by the context: ST s (v r) at Testing/Foo.hs:29:26
Expected type: Mutable v (PrimState m0) r
Actual type: Mutable v s r
Relevant bindings include
y :: Mutable v s r (bound at Testing/Foo.hs:30:3)
f :: forall (v :: * -> * -> *).
MVector v r =>
v (PrimState m0) r -> m0 ()
(bound at Testing/Foo.hs:29:19)
In the first argument of ‘f’, namely ‘y’
In a stmt of a 'do' block: f y
I'm pretty sure the rank-2 type is to blame, possibly caused by a monomorphism restriction. However, as suggested in a previous question of mine, I enabled -XNoMonomorphismRestriction, but got the same error.
What is the difference between these seemingly identical code snippets?
I think that having a rough mental model of the type-level plumbing involved here is essential, so I'm going go talk about "implicit things" in a bit more detail, and scrutinize your problem only after that. Readers only interested in the direct solution to the question may skip to the "Pattern matching on polymorhpic values" subsection and the end.
1. Implicit function arguments
Type arguments
GHC compiles Haskell to a small intermediate language called Core, which is essentially a rank-n polymorphic typed lambda calculus called System F (plus some extensions). Below I am going use Haskell alongside a notation somewhat resembling Core; I hope it's not overly confusing.
In Core, polymorphic functions are functions which take types as additional arguments, and arguments further down the line can refer to those types or have those types:
-- in Haskell
const :: forall (a :: *) (b :: *). a -> b -> a
const x y = x
-- in pseudo-Core
const' :: (a :: *) -> (b :: *) -> a -> b -> a
const' a b x y = x
This means that we must also supply type arguments to these functions whenever we want to use them. In Haskell type inference usually figures out the type arguments and supplies them automatically, but if we look at the Core output (for example, see this introduction for how to do that), type arguments and applications are visible everywhere. Building a mental model of this makes figuring out higher-rank code a whole lot easier:
-- Haskell
poly :: (forall a. a -> a) -> b -> (Int, b)
poly f x = (f 0, f x)
-- pseudo-Core
poly' :: (b :: *) -> ((a :: *) -> a -> a) -> b -> (Int, b)
poly' b f x = (f Int 0, f b x)
And it makes clear why some things don't typecheck:
wrong :: (a -> a) -> (Int, Bool)
wrong f = (f 0, f True)
wrong' :: (a :: *) -> (a -> a) -> (Int, Bool)
wrong' a f = (f ?, f ?) -- f takes an "a", not Int or Bool.
Class constraint arguments
-- Haskell
show :: forall a. Show a => a -> String
show x = show x
-- pseudo-Core
show' :: (a :: *) -> Show a -> a -> String
show' a (ShowDict showa) x = showa x
What is ShowDict and Show a here? ShowDict is just a Haskell record containing a show instance, and GHC generates such records for each instance of a class. Show a is just the type of this instance record:
-- We translate classes to a record type:
class Show a where show :: a -> string
data Show a = ShowDict (show :: a -> String)
-- And translate instances to concrete records of the class type:
instance Show () where show () = "()"
showUnit :: Show ()
showUnit = ShowDict (\() -> "()")
For example, whenever we want to apply show, the compiler has to search the scope in order to find a suitable type argument and an instance dictionary for that type. Note that while instances are always top level, quite often in polymorphic functions the instances are passed in as arguments:
data Foo = Foo
-- instance Show Foo where show _ = "Foo"
showFoo :: Show Foo
showFoo = ShowDict (\_ -> "Foo")
-- The compiler fills in an instance from top level
fooStr :: String
fooStr = show' Foo showFoo Foo
polyShow :: (Show a, Show b) => a -> b -> String
polyShow a b = show a ++ show b
-- Here we get the instances as arguments (also, note how (++) also takes an extra
-- type argument, since (++) :: forall a. [a] -> [a] -> [a])
polyShow' :: (a :: *) -> (b :: *) -> Show a -> Show b -> a -> b -> String
polyShow' a b (ShowDict showa) (ShowDict showb) a b -> (++) Char (showa a) (showb b)
Pattern matching on polymorphic values
In Haskell, pattern matching on functions doesn't make sense. Polymorphic values can be also viewed as functions, but we can pattern match on them, just like in OP's erroneous runfun1 example. However, all the implicit arguments must be inferable in the scope, or else the mere act of pattern matching is a type error:
import Data.Monoid
-- it's a type error even if we don't use "a" or "n".
-- foo :: (forall a. Monoid a => (a, Int)) -> Int
-- foo (a, n) = 0
foo :: ((a :: *) -> Monoid a -> (a, Int)) -> Int
foo f = ? -- What are we going to apply f to?
In other words, by pattern matching on a polymorphic value, we assert that all implicit arguments have been already applied. In the case of foo here, although there isn't a syntax for type application in Haskell, we can sprinkle around type annotations:
{-# LANGUAGE ScopedTypeVariables, RankNTypes #-}
foo :: (forall a. Monoid a => (a, Int)) -> Int
foo x = case (x :: (String, Int)) of (_, n) -> n
-- or alternatively
foo ((_ :: String), n) = n
Again, pseudo-Core makes the situation clearer:
foo :: ((a :: *) -> Monoid a -> (a, Int)) -> Int
foo f = case f String monoidString of (_ , n) -> n
Here monoidString is some available Monoid instance of String.
2. Implicit data fields
Implicit data fields usually correspond to the notion of "existential types" in Haskell. In a sense, they are dual to implicit function arguments with respect to term obligations:
When we construct functions, the implicit arguments are available in the function body.
When we apply functions, we have extra obligations to fulfill.
When we construct data with implicit fields, we must supply those extra fields.
When we pattern match on data, the implicit fields also come into scope.
Standard example:
{-# LANGUAGE GADTs #-}
data Showy where
Showy :: forall a. Show a => a -> Showy
-- pseudo-Core
data Showy where
Showy :: (a :: *) -> Show a -> a -> Showy
-- when constructing "Showy", "Show a" must be also available:
someShowy :: Showy
someShowy = Showy (300 :: Int)
-- in pseudo-Core
someShowy' = Showy Int showInt 300
-- When pattern matching on "Showy", we get an instance in scope too
showShowy :: Showy -> String
showShowy (Showy x) = show x
showShowy' :: Showy -> String
showShowy' (Showy a showa x) = showa x
3. Taking a look at OP's example
We have the function
runFun1 :: (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun1 dfun#(DimFun dim t) x | V.length x == dim = runST $ do
y <- thaw x
t y
unsafeFreeze y
Remember that pattern matching on polymorphic values asserts that all implicit arguments are available in the scope. Except that here, at the point of pattern matching there is no m at all in scope, let alone a PrimMonad instance for it.
With GHC 7.8.x it's is good practice to use type holes liberally:
runFun1 :: (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun1 (DimFun dim t) x | V.length x == dim = _
Now GHC will duly display the type of the hole, and also the types of the variables in the context. We can see that t has type Mutable v (PrimState m0) r -> m0 (), and we also see that m0 is not listed as bound anywhere. Indeed, it is a notorious "ambiguous" type variable conjured up by GHC as a placeholder.
So, why don't we try manually supplying the arguments, just as in the prior example with the Monoid instance? We know that we will use t inside an ST action, so we can try fixing m as ST s and GHC automatically applies the PrimMonad instance for us:
runFun1 :: forall v r. (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun1 (DimFun dim (t :: Mutable v s r -> ST s ())) x
| V.length x == dim = runST $ do
y <- thaw x
t y
unsafeFreeze y
... except it doesn't work and we get the error "Couldn't match type ‘s’ with ‘s1’ because type variable ‘s1’ would escape its scope".
It turns out - comes as no surprise - that we've forgotten about yet another implicit argument. Recall the type of runST:
runST :: (forall s. ST s a) -> a
We can imagine that runST takes a function of type ((s :: PrimState ST) -> ST s a), and then our code looks like this:
runST $ \s -> do
y <- thaw x -- y :: Mutable v s r
t y -- error: "t" takes a "Mutable v s r" with a different "s".
unsafeFreeze y
The s in t's argument type is silently introduced at the outermost scope:
runFun1 :: forall v s r. ...
And thus the two s-es are distinct.
A possible solution is to pattern match on the DimFun argument inside the ST action. There, the correct s is in scope, and GHC can supply ST s as m:
runFun1 :: forall v r. (Vector v r, MVector (Mutable v) r) =>
(forall m . PrimMonad m => DimFun (Mutable v) m r) -> v r -> v r
runFun1 dimfun x = runST $ do
y <- thaw x
case dimfun of
DimFun dim t | dim == M.length y -> t y
unsafeFreeze y
With some parameters made explicit:
runST $ \s -> do
y <- thaw x
case dimfun (ST s) primMonadST of
DimFun dim t | dim == M.length y -> t y
unsafeFreeze y
As an exercise, let's convert all of the function to pseudo-Core (but let's not desugar the do syntax, because that would be way too ugly):
-- the full types of the functions involved, for reference
thaw :: forall m v a. (PrimMonad m, V.Vector v a) => v a -> m (V.Mutable v (PrimState m) a)
runST :: forall a. (forall s. ST s a) -> a
unsafeFreeze :: forall m v a. (PrimMonad m, Vector v a) => Mutable v (PrimState m) a -> v a
M.length :: forall v s a. MVector v s a -> Int
(==) :: forall a. Eq a => a -> a -> Bool
runFun1 ::
(v :: * -> *) -> (r :: *)
-> Vector v r -> MVector (Mutable v) r
-> ((m :: (* -> *)) -> PrimMonad m -> DimFun (Mutable v) m r)
-> v r -> v r
runFun1 v r vecInstance mvecInstance dimfun x = runST r $ \s -> do
y <- thaw (ST s) v r primMonadST vecInstance x
case dimFun (ST s) primMonadST of
DimFun dim t | (==) Int eqInt dim (M.length v s r y) -> t y
unsafeFreeze (ST s) v r primMonadST vecInstance y
That was a mouthful.
Now we are well-equipped to explain why runFun2 worked:
runFun2 :: (Vector v r, MVector (Mutable v) r) =>
(forall m . (PrimMonad m) => DimFun (Mutable v) m r) -> v r -> v r
runFun2 t x = runST $ do
y <- thaw x
evalFun t y
unsafeFreeze y
evalFun :: (PrimMonad m, MVector v r) => DimFun v m r -> v (PrimState m) r -> m ()
evalFun (DimFun dim f) y | dim == M.length y = f y
evalFun is just a polymorphic function that gets called in the right place (we ultimately pattern match on t in the right place), where the correct ST s is available as the m argument.
As a type system gets more sophisticated, pattern matching becomes a progressively more serious affair, with far-reaching consequences and non-trivial requirements. At the end of the spectrum you find full-dependent languages and proof assistants such as Agda, Idris or Coq, where pattern matching on a piece of data can mean accepting an arbitrary logical proposition as true in a certain branch of your program.
Though #AndrasKovacs gave a great answer, I think it is worth pointing out how to avoid this nastiness altogether. This answer to a related question by me shows how the "correct" definition for DimFun makes all of the rank-2 stuff go away.
By defining DimFun as
data DimFun v r =
DimFun {dim::Int, func :: forall s . (PrimMonad s) => v (PrimState s) r -> s ()}
runFun1 becomes:
runFun1 :: (Vector v r)
=> DimFun (Mutable v) r -> v r -> v r
runFun1 (DimFun dim t) x | dim == V.length x = runST $ do
y <- thaw x
t y
unsafeFreeze y
and compiles without issue.
Pattern-match on a constrained value is not allowed, I think. In particular, you could use a pattern-match, but only for a GADT constructor that fixed the type(s) in the constraint and choose a specific instance. Otherwise, I get the ambiguous type variable error.
That is, I don't think that GHC can unify the type of a value matching the pattern (DimFun dim t) with the type (forall m . (PrimMonad m) => DimFun (Mutable v) m r).
Note that the pattern match in evalFun looks similar, but it is allowed to put constraints on m since the quantification is scoped over the whole evalFun; in constrast, runFun1 as a smaller scope for the quantification of m.
HTH

Currying Product Types

Using type families, we can define the function fold over a type and the underlying algebra for that type represented as an n-tuple of functions and constant values. This permits the definition of a generalized foldr function, defined in the Foldable type class:
import Data.Set (Set)
import Data.Map (Map)
import qualified Data.Set as S
import qualified Data.Map as M
class Foldable m where
type Algebra m b :: *
fold :: Algebra m b -> m -> b
instance (Ord a) => Foldable (Set a) where
type Algebra (Set a) b = (b, a -> b -> b)
fold = uncurry $ flip S.fold
instance (Ord k) => Foldable (Map k a) where
type Algebra (Map k a) b = (b, k -> a -> b -> b)
fold = uncurry $ flip M.foldWithKey
Similarly, constraint kinds permit the definition of a generalized map function. The map function differs from fmap by considering each value field of an algebraic data type:
class Mappable m where
type Contains m :: *
type Mapped m r b :: Constraint
map :: (Mapped m r b) => (Contains m -> b) -> m -> r
instance (Ord a) => Mappable (Set a) where
type Contains (Set a) = a
type Mapped (Set a) r b = (Ord b, r ~ Set b)
map = S.map
instance (Ord k) => Mappable (Map k a) where
type Contains (Map k a) = (k, a)
type Mapped (Map k a) r b = (Ord k, r ~ Map k b)
map = M.mapWithKey . curry
From the user's perspective, neither function is particularly friendly. In particular, neither technique permits the definition of curried functions. This means that the user cannot easily apply either fold or the mapped function partially. What I would like is a type-level function that curries tuples of functions and values, in order to generate curried versions of the above. Thus, I would like to write something approximating the following type-function:
Curry :: Product -> Type -> Type
Curry () m = m
Curry (a × as) m = a -> (Curry as m b)
If so, we could generate a curried fold function from the underlying algebra. For instance:
fold :: Curry (Algebra [a] b) ([a] -> b)
≡ fold :: Curry (b, a -> b -> b) ([a] -> b)
≡ fold :: b -> (Curry (a -> b -> b)) ([a] -> b)
≡ fold :: b -> (a -> b -> b -> (Curry () ([a] -> b))
≡ fold :: b -> ((a -> b -> b) -> ([a] -> b))
map :: (Mapped (Map k a) r b) => (Curry (Contains (Map k a)) b) -> Map k a -> r
≡ map :: (Mapped (Map k a) r b) => (Curry (k, a) b) -> Map k a -> r
≡ map :: (Mapped (Map k a) r b) => (k -> (Curry (a) b) -> Map k a -> r
≡ map :: (Mapped (Map k a) r b) => (k -> (a -> Curry () b)) -> Map k a -> r
≡ map :: (Mapped (Map k a) r b) => (k -> (a -> b)) -> Map k a -> r
I know that Haskell doesn't have type functions, and the proper representation of the n-tuple would probably be something like a type-level length-indexed list of types. Is this possible?
EDIT: For completeness, my current attempt at a solution is attached below. I am using empty data types to represent products of types, and type families to represent the function Curry, above. This solution appears to work for the map function, but not the fold function. I believe, but am not certain, that Curry is not being reduced properly when type checking.
data Unit
data Times a b
type family Curry a m :: *
type instance Curry Unit m = m
type instance Curry (Times a l) m = a -> Curry l m
class Foldable m where
type Algebra m b :: *
fold :: Curry (Algebra m b) (m -> b)
instance (Ord a) => Foldable (Set a) where
type Algebra (Set a) b = Times (a -> b -> b) (Times b Unit)
fold = S.fold
instance (Ord k) => Foldable (Map k a) where
type Algebra (Map k a) b = Times (k -> a -> b -> b) (Times b Unit)
fold = M.foldWithKey
class Mappable m where
type Contains m :: *
type Mapped m r b :: Constraint
map :: (Mapped m r b) => Curry (Contains m) b -> m -> r
instance (Ord a) => Mappable (Set a) where
type Contains (Set a) = Times a Unit
type Mapped (Set a) r b = (Ord b, r ~ Set b)
map = S.map
instance (Ord k) => Mappable (Map k a) where
type Contains (Map k a) = Times k (Times a Unit)
type Mapped (Map k a) r b = (Ord k, r ~ Map k b)
map = M.mapWithKey
Ok, if I understand you correctly, you can create inconvenient folds, but want to have convenient curried folds.
Below is an explanation how to achieve this as a separate step. Yes, it can also be done all at once, I've done something similar before. However, I think the separate phase makes it clearer what's going on.
We need the following language extensions:
{-# LANGUAGE TypeFamilies, TypeOperators, FlexibleInstances #-}
I'm using the following product and unit types:
data U = U
data a :*: b = a :*: b
infixr 8 :*:
As an example, let's assume we have an inconvenient version of a fold on lists:
type ListAlgType a r = (U -> r)
:*: (a :*: r :*: U -> r)
:*: U
inconvenientFold :: ListAlgType a r -> [a] -> r
inconvenientFold (nil :*: cons :*: U) [] = nil U
inconvenientFold a#(nil :*: cons :*: U) (x : xs) = cons (x :*: inconvenientFold a xs :*: U)
We have a nested product type, and we want to curry both levels. I'm defining two type classes for this, one for each layer. (It might be doable with one more general function, I haven't tried in this case.)
class CurryInner a where
type CurryI a k :: *
curryI :: (a -> b) -> CurryI a b
uncurryI :: CurryI a b -> a -> b
class CurryOuter a where
type CurryO a k :: *
curryO :: (a -> b) -> CurryO a b
uncurryO :: CurryO a b -> (a -> b) -- not really required here
Each type class implements the isomorphism between the curried and uncurried types. The type classes look identical, but CurryOuter will call CurryInner for each component of the outer nested tuple.
The instances are relatively straightforward:
instance CurryInner U where
type CurryI U k = k
curryI f = f U
uncurryI x = \ U -> x
instance CurryInner ts => CurryInner (t :*: ts) where
type CurryI (t :*: ts) k = t -> CurryI ts k
curryI f = \ t -> curryI (\ ts -> f (t :*: ts))
uncurryI f = \ (t :*: ts) -> uncurryI (f t) ts
instance CurryOuter U where
type CurryO U k = k
curryO f = f U
uncurryO x = \ U -> x
instance (CurryInner a, CurryOuter ts) => CurryOuter ((a -> b) :*: ts) where
type CurryO ((a -> b) :*: ts) k = CurryI a b -> CurryO ts k
curryO f = \ t -> curryO (\ ts -> f (uncurryI t :*: ts))
uncurryO f = \ (t :*: ts) -> uncurryO (f (curryI t)) ts
That's it. Note that
*Main> :kind! CurryO (ListAlgType A R) ([A] -> R)
CurryO (ListAlgType A R) ([A] -> R) :: *
= R -> (A -> R -> R) -> [A] -> R
(for suitably defined placeholder types A and R). We can use it as follows:
*Main> curryO inconvenientFold 0 (+) [1..10]
55
Edit: I now see you're actually only asking about currying the outer layer. You then only need one class, but can use the same idea. I used this example because I had written something for a sum-of-product based generic programming library which needed two levels of currying before, and thought at first you are in the same setting.
Ok, I think my other answer isn't actually really an answer to your question. Sorry for that.
In your final code, compare the types of fold and map:
fold :: Curry (Algebra m b) (m -> b)
map :: (Mapped m r b) => Curry (Contains m) b -> m -> r
There's a substantial difference here. The type of fold is just a type family application, whereas the type of map contains the final m -> r, mentioning the class parameter m. So in the case of map, it's easy for GHC to learn at which type you want to instance the class from the context.
Not so in the case of fold, unfortunately, because type families need not be injective, and therefore aren't easy to invert. So by seeing a particular type you use fold at, it's impossible for GHC to infer what m is.
The standard solution to this problem is to use a proxy argument that fixes the type of m, by defining
data Proxy m = P
and then giving fold this type instead:
fold :: Proxy m -> Curry (Algebra m b) (m -> b)
You have to adapt the instances to take and discard the proxy argument. Then you can use:
fold (P :: Proxy (Set Int)) (+) 0 (S.fromList [1..10])
or similar to call the fold function on sets.
To see more clearly why this situation is difficult for GHC to solve, consider this toy example instead:
class C a where
type F a :: *
f :: F a
instance C Bool where
type F Bool = Char -> Char
f = id
instance C () where
type F () = Char -> Char
f = toUpper
Now, if you call f 'x', there's no meaningful way for GHC to detect which instance you meant. The proxy would help here as well.
A type-level list is exactly what you need! You got very close, but you need the full power of both DataKinds and ScopedTypeVariables for this to work properly:
{-# LANGUAGE ConstraintKinds, DataKinds, FlexibleContexts, FlexibleInstances, TypeFamilies, TypeOperators, ScopedTypeVariables #-}
import GHC.Exts (Constraint)
import Data.Set (Set)
import Data.Map (Map)
import qualified Data.Set as S
import qualified Data.Map as M
-- | A "multifunction" from a list of inhabitable types to an inhabitable type (curried from the start).
type family (->>) (l :: [*]) (y :: *) :: *
type instance '[] ->> y = y
type instance (x ': xs) ->> y = x -> (xs ->> y)
class Foldable (m :: *) where
type Algebra m (b :: *) :: [*]
fold :: forall (b :: *). Algebra m b ->> (m -> b)
instance (Ord a) => Foldable (Set a) where
type Algebra (Set a) b = '[(a -> b -> b), b]
fold = S.fold :: forall (b :: *). (a -> b -> b) -> b -> Set a -> b
instance (Ord k) => Foldable (Map k a) where
type Algebra (Map k a) b = '[(k -> a -> b -> b), b]
fold = M.foldWithKey :: forall (b :: *). (k -> a -> b -> b) -> b -> Map k a -> b
class Mappable m where
type Contains m :: [*]
type Mapped m (b :: *) (r :: *) :: Constraint
map :: forall (b :: *) (r :: *). Mapped m b r => (Contains m ->> b) -> m -> r
instance (Ord a) => Mappable (Set a) where
type Contains (Set a) = '[a]
type Mapped (Set a) b r = (Ord b, r ~ Set b)
map = S.map :: forall (b :: *). (Ord b) => (a -> b) -> Set a -> Set b
instance (Ord k) => Mappable (Map k a) where
type Contains (Map k a) = '[k, a]
type Mapped (Map k a) b r = r ~ Map k b
map = M.mapWithKey :: forall (b :: *). (k -> a -> b) -> Map k a -> Map k b

Resources