Excersize from an online course.
Suppose, that for a standard list Applicative functor the <*> operator is defined in a standard way, while pure is changed to
pure x = [x,x]
What laws of Applicative typeclass will be violated?
Homomorphism: pure g <*> pure x ≡ pure (g x)
Identity: pure id <*> xs ≡ xs
Interchange: fs <*> pure x ≡ pure ($ x) <*> fs
Applicative-Functor: g <$> xs ≡ pure g <*> xs
Composition: (.) <$> us <*> vs <*> xs ≡ us <*> (vs <*> xs)
I created the following file:
newtype MyList a = MyList {getMyList :: [a]}
deriving Show
instance Functor MyList where
fmap f (MyList xs) = MyList (map f xs)
instance Applicative MyList where
pure x = MyList [x,x]
MyList gs <*> MyList xs = MyList ([g x | g <- gs, x <- xs])
fs = MyList [\x -> 2*x, \x -> 3*x]
xs = MyList [1,2]
x = 1
g = (\x -> 2*x)
us = MyList [(\x -> 2*x)]
vs = MyList [(\x -> 3*x)]
Then I tried:
Homomorphism: pure g <*> pure x ≡ pure (g x)
*Main> pure g <*> pure x :: MyList Integer
MyList {getMyList = [2,2,2,2]}
*Main> pure (g x) :: MyList Integer
MyList {getMyList = [2,2]}
Identity: pure id <*> xs ≡ xs
*Main> pure id <*> xs :: MyList Integer
MyList {getMyList = [1,2,1,2]}
*Main> xs :: MyList Integer
MyList {getMyList = [1,2]}
Interchange: fs <*> pure x ≡ pure ($ x) <*> fs
*Main> fs <*> pure x
[2,3]
*Main> pure ($ x) <*> fs
[2,3]
Applicative-Functor: g <$> xs ≡ pure g <*> xs
*Main> g <$> xs
MyList {getMyList = [2,4]}
*Main> pure g <*> xs
MyList {getMyList = [2,4,2,4]}
Composition: (.) <$> us <*> vs <*> xs ≡ us <*> (vs <*> xs)
*Main> (.) <$> us <*> vs <*> xs
MyList {getMyList = [6,12]}
*Main> us <*> (vs <*> xs)
MyList {getMyList = [6,12]}
Composition shouldn't be violated, because pure is not used here.
It seems that Homomorphism, Identity and Applicative-Functor don't work. But when I select them in the course, it says that the answer is wrong. So, who is the fool: me or the author of the course?
As per #DavidFletcher's comment, using your code, I see different output for the interchange test:
> fs <*> pure x
MyList {getMyList = [2,2,3,3]}
> pure ($ x) <*> fs
MyList {getMyList = [2,3,2,3]}
so you might want to double-check that one.
Related
I have two composed applicative functors Maybe [Integer] and want to combine them with <$>/<*> but I am stuck with applying the applicative operation. The following does not typecheck:
(<*>) (<*>) ((<$>) ((+) <$>) $ Just [1,2,3]) $ Just [4,5,6]
Expected result:
Just [5,6,7,6,7,8,7,8,9]
The functor part works, i.e. the intermediate value passed to <*> as the first argument is Just [Integer -> Integer]. I am used to S-expressions so I have a hard time with the Haskell syntax. I know of Compose but I am interested in the mere composition wihtout abstraction.
As Li-yao Xia said, using liftA2 makes it a lot less confusing.
But if you still what to see what it becomes in terms of the underlaying operations, we can expand the definition of liftA2:
liftA2 :: (a -> b -> c) -> f a -> f b -> f c
liftA2 f x y = f <$> x <*> y
so the solution becomes
(liftA2 . liftA2) (+) (Just [1,2,3]) (Just [4,5,6])
= liftA2 (liftA2 (+)) (Just [1,2,3]) (Just [4,5,6])
= (\f x y -> f <$> x <*> y) ((\f x y -> f <$> x <*> y) (+)) (Just [1,2,3]) (Just [4,5,6])
= ((\f x y -> f <$> x <*> y) (+)) <$> Just [1,2,3] <*> Just [4,5,6]
= (\x y -> (+) <$> x <*> y) <$> Just [1,2,3] <*> Just [4,5,6]
Now, this is not in point free style like your example above, and I really don't think it's helpful to convert it into point free, but here's the output from http://pointfree.io:
((<*>) . ((+) <$>)) <$> Just [1, 2, 3] <*> Just [4, 5, 6]
we can see that this is the same by eta-expanding:
(<*>) . ((+) <$>)
= \x y -> ((<*>) . ((+) <$>)) x y
= \x y -> ((<*>) $ ((+) <$>) x) y
= \x y -> ((<*>) ((+) <$> x)) y
= \x y -> (<*>) ((+) <$> x) y
= \x y -> ((+) <$> x) <*> y
= \x y -> (+) <$> x <*> y
liftA2 might be less confusing for this than (<*>).
(+) :: Int -> Int -> Int
liftA2 (+) :: [Int] -> [Int] -> [Int]
liftA2 (liftA2 (+)) :: Maybe [Int] -> Maybe [Int] -> Maybe [Int]
liftA2 (liftA2 (+)) (Just [1,2,3]) (Just [4,5,6])
The composition of two Applicatives is always an Applicative (unlike the case for Monad).
We can use this to our advantage here with the Compose newtype from Data.Functor.Compose:
newtype Compose f g a = Compose { getCompose :: f (g a) }
It requires a bit of wrapping, but this kind of solution could be useful under the right circumstances:
example :: Maybe [Int]
example =
getCompose ((+) <$> Compose (Just [1,2,3]) <*> Compose (Just [4,5,6]))
One other way could be to use the ListT transformer. While it works just fine in this case, for some reason it's a depreciated transformer, marked in red with "Deprecated: This transformer is invalid on most monads".
import Control.Monad.Trans.List
doit :: (Int-> Int -> Int) -> Maybe [Int] -> Maybe [Int] -> Maybe [Int]
doit f mt1 mt2 = runListT $ f <$> (ListT mt1) <*> (ListT mt2)
λ> doit (+) (Just [1,2,3]) (Just [4,5,6])
Just [5,6,7,6,7,8,7,8,9]
In Maybe applicative, <*> can be implemented based on fmap. Is it incidental, or can it be generalized to other applicative(s)?
(<*>) :: Maybe (a -> b) -> Maybe a -> Maybe b
Nothing <*> _ = Nothing
(Just g) <*> mx = fmap g mx
Thanks.
See also In applicative, how can `<*>` be represented in terms of `fmap_i, i=0,1,2,...`?
It cannot be generalized. A Functor instance is unique:
instance Functor [] where
fmap = map
but there can be multiple valid Applicative instances for the same type constructor.
-- "Canonical" instance: [f, g] <*> [x, y] == [f x, f y, g x, g y]
instance Applicative [] where
pure x = [x]
[] <*> _ = []
(f:fs) <*> xs = fmap f xs ++ (fs <*> xs)
-- Zip instance: [f, g] <*> [x, y] == [f x, g y]
instance Applicative [] where
pure x = repeat x
(f:fs) <*> (x:xs) = f x : (fs <*> xs)
_ <*> _ = []
In the latter, we neither want to apply any single function from the left argument to all elements of the right, nor apply all the functions on the left to any single element on the right, making fmap useless.
During my study of Typoclassopedia I encountered this proof, but I'm not sure if my proof is correct. The question is:
One might imagine a variant of the interchange law that says something about applying a pure function to an effectful argument. Using the above laws, prove that:
pure f <*> x = pure (flip ($)) <*> x <*> pure f
Where "above laws" points to Applicative Laws, briefly:
pure id <*> v = v -- identity law
pure f <*> pure x = pure (f x) -- homomorphism
u <*> pure y = pure ($ y) <*> u -- interchange
u <*> (v <*> w) = pure (.) <*> u <*> v <*> w -- composition
My proof is as follows:
pure f <*> x = pure (($) f) <*> x -- identical
pure f <*> x = pure ($) <*> pure f <*> x -- homomorphism
pure f <*> x = pure (flip ($)) <*> x <*> pure f -- flip arguments
The first two steps of your proof look fine, but the last step doesn't. While the definition of flip allows you to use a law like:
f a b = flip f b a
that doesn't mean:
pure f <*> a <*> b = pure (flip f) <*> b <*> a
In fact, this is false in general. Compare the output of these two lines:
pure (+) <*> [1,2,3] <*> [4,5]
pure (flip (+)) <*> [4,5] <*> [1,2,3]
If you want a hint, you are going to need to use the original interchange law at some point to prove this variant.
In fact, I found I had to use the homomorphism, interchange, and composition laws to prove this, and part of the proof was pretty tricky, especially getting the sections right --like ($ f), which is different from (($) f). It was helpful to have GHCi open to double-check that each step of my proof type checked and gave the right result. (Your proof above type checks fine; it's just that the last step wasn't justified.)
> let f = sqrt
> let x = [1,4,9]
> pure f <*> x
[1.0,2.0,3.0]
> pure (flip ($)) <*> x <*> pure f
[1.0,2.0,3.0]
>
I ended up proving it backwards:
pure (flip ($)) <*> x <*> pure f
= (pure (flip ($)) <*> x) <*> pure f -- <*> is left-associative
= pure ($ f) <*> (pure (flip ($)) <*> x) -- interchange
= pure (.) <*> pure ($ f) <*> pure (flip ($)) <*> x -- composition
= pure (($ f) . (flip ($))) <*> x -- homomorphism
= pure (flip ($) f . flip ($)) <*> x -- identical
= pure f <*> x
Explanation of the last transformation:
flip ($) has type a -> (a -> c) -> c, intuitively, it first takes an argument of type a, then a function that accepts that argument, and in the end it calls the function with the first argument. So flip ($) 5 takes as argument a function which gets called with 5 as it's argument. If we pass (+ 2) to flip ($) 5, we get flip ($) 5 (+2) which is equivalent to the expression (+2) $ 5, evaluating to 7.
flip ($) f is equivalent to \x -> x $ f, that means, it takes as input a function and calls it with the function f as argument.
The composition of these functions works like this: First flip ($) takes x as it's first argument, and returns a function flip ($) x, this function is awaiting a function as it's last argument, which will be called with x as it's argument. Now this function flip ($) x is passed to flip ($) f, or to write it's equivalent (\x -> x $ f) (flip ($) x), this results in the expression (flip ($) x) f, which is equivalent to f $ x.
You can check the type of flip ($) f . flip ($) is something like this (depending on your function f):
λ: let f = sqrt
λ: :t (flip ($) f) . (flip ($))
(flip ($) f) . (flip ($)) :: Floating c => c -> c
I'd remark that such theorems are, as a rule, a lot less involved when written in mathematical style of a monoidal functor, rather than the applicative version, i.e. with the equivalent class
class Functor f => Monoidal f where
pure :: a -> f a
(⑂) :: f a -> f b -> f (a,b)
Then the laws are
id <$> v = v
f <$> (g <$> v) = f . g <$> v
f <$> pure x = pure (f x)
x ⑂ pure y = fmap (,y) x
a⑂(b⑂c) = assoc <$> (a⑂b)⑂c
where assoc ((x,y),z) = (x,(y,z)).
The theorem then reads
pure u ⑂ x = swap <$> x ⑂ pure u
Proof:
swap <$> x ⑂ pure u
= swap <$> fmap (,u) x
= swap . (,u) <$> x
= (u,) <$> x
= pure u ⑂ x
□
For a List, why does right apply (*>) behave as repeating and appending the second argument n times, where n is the length of the first argument?
ghci> [1,2,3] *> [4,5]
[4,5,4,5,4,5]
The *> operator is defined, by default, as
xs *> ys = id <$ xs <*> ys
which in turn translates, by default, to
const id <$> xs <*> ys
That is, it replaces each element of xs with id to get xs' and then calculates xs' <*> ys. [] is a Monad instance, where (=<<) = concatMap. One of the laws of Applicative lays out the relationship between Applicative and Monad instances:
pure = return
fs <*> as = fs `ap` as = fs >>= \f -> as >>= \a -> f a
For lists, this is
fs <*> as = [f a | f <- fs, a <- as]
So *> for lists is ultimately determined by the Monad instance.
Note that there is another very sensible Applicative instance for lists, which is made available through a newtype in Control.Applicative:
newtype ZipList a = ZipList [a]
instance Applicative ZipList where
pure = repeat
(<*>) = zipWith ($)
When I don't grasp how an expression in Haskell works I often find it helps to decompose it into a more basic form.
Using the following definitions
sequenceA :: (Applicative f) => [f a] -> f [a]
sequenceA [] = pure []
sequenceA (x:xs) = (:) <$> x <*> sequenceA xs
instance Applicative ((->) r) where
pure x = (\_ -> x)
f <*> g = \x -> f x (g x)
I rewrote sequenceA [(+3),(+2)] 3 as
(\_ -> (:)) <*> (+3) <*> ((\_ -> (:)) <*> (+2) <*> (\_-> [])) $ 3
And then turned it into (please excuse the format; I'm not sure what the convention is for splitting lines)
(\d ->(\c->(\b -> (\a -> (\_ -> (:)) a (+3) a) b (\_ -> (:)) b) c (+2) c) d (\_ -> []) d) 3
This seems right when I work through it by hand, but I can't get GHCi to accept it. What have I done wrong here? My second question is how to convert from this form into functional composition. I've tried substituing dots in various combinations, but GHCi rejects all of them....
Being an idle goodfornothing, I thought I would make a computer do the expansion for me. So into GHCi, I typed
let pu x = "(\\_ -> " ++ x ++ ")"
let f >*< a = "(\\g -> " ++ f ++ " g (" ++ a ++ " g))"
So now I have funny versions of pure and <*> which map strings which look like expressions to string which look like more complicated expressions. I then defined, similarly, the analogue of sequenceA, replacing functions by strings.
let sqa [] = pu "[]" ; sqa (f : fs) = (pu "(:)" >*< f) >*< sqa fs
I was then able to generate the expanded form of the example as follows
putStrLn $ sqa ["(+3)","(+2)"] ++ " 3"
which duly printed
(\g -> (\g -> (\_ -> (:)) g ((+3) g)) g ((\g -> (\g -> (\_ -> (:)) g ((+2) g)) g ((\_ -> []) g)) g)) 3
This last, copied to the prompt, yielded
[6,5]
Comparing the output from my "metaprogram" with the attempt in the question shows a shorter initial prefix of lambdas, arising from a shallower nesting of <*> operations. Remember, it's
(pure (:) <*> (+3)) <*> ((pure (:) <*> (+2)) <*> pure [])
so the outer (:) should be only three lambdas deep. I suspect the proposed expansion may correspond to a differently bracketed version of the above, perhaps
pure (:) <*> (+3) <*> pure (:) <*> (+2) <*> pure []
Indeed, when I evaluate
putStrLn $ pu "(:)" >*< "(+3)" >*< pu "(:)" >*< "(+2)" >*< pu "[]" ++ " 3 "
I get
(\g -> (\g -> (\g -> (\g -> (\_ -> (:)) g ((+3) g)) g ((\_ -> (:)) g)) g ((+2) g)) g ((\_ -> []) g)) 3
which looks like it matches the (updated)
(\d -> (\c -> (\b -> (\a -> (\_ -> (:)) a ((+3) a)) b ((\_ -> (:)) b)) c ((+2) c)) d ((\_ -> []) d)) 3
I hope this machine-assisted investigation helps to clarify what's going on.
You rewrote (\_ -> (:)) <*> (+3) as \a -> (\_ -> (:)) a (+3) a, which is rewriting f <*> g as f x g x instead of f x (g x). I think you made that mistake for every <*>.
It might be easier to use combinators, e.g. _S and _K, symbolically, and not their definitions as lambda-expressions,
_S f g x = f x (g x)
_K x y = x
With functions, fmap is (.) and <*> is _S, as others already mentioned. So,
sequenceA [(+3),(+2)] 3 ==
( ((:) <$> (+3)) <*> sequenceA [(+2)] ) 3 ==
_S ((:).(+3)) ( ((:) <$> (+2)) <*> pure [] ) 3 ==
_S ((:).(+3)) ( _S ((:).(+2)) (_K []) ) 3 ==
((:).(+3)) 3 ( _S ((:).(+2)) (_K []) 3 ) ==
((:).(+3)) 3 ( ((:).(+2)) 3 (_K [] 3) ) ==
(6:) ( (5:) [] ) ==
[6,5]
So it might be easier to decompose expressions down to basic functions and combinators and stop there (i.e. not decomposing them to their lambda expressions), using their "re-write rules" in manipulating the expression to find its more comprehensible form.
If you wanted to, you could now write down for yourself a more abstract, informal re-write rule for sequenceA as
sequenceA [f,g,..., z] ==
_S ((:).f) . _S ((:).g) . _S ..... . _S ((:).z) . _K []
and so
sequenceA [f,g,..., z] a ==
((:).f) a $ ((:).g) a $ ..... $ ((:).z) a $ _K [] a ==
(f a:) $ (g a:) $ ..... $ (z a:) $ [] ==
[f a, g a, ..., z a]
and hence
sequenceA fs a == map ($ a) fs == flip (map . flip ($)) fs a
to wit,
Prelude Control.Applicative> flip (map . flip ($)) [(+3),(+2)] 3
[6,5]