This excellent answer in this question demonstrates how bind can be written in terms of join and fmap:
(>>=) :: m v -> (v -> m w) -> m w
says "if you have a strategy to produce a v, and for each v a
follow-on strategy to produce a w, then you have a strategy to produce
a w". How can we capture that in terms of join?
mv >>= v2mw = join (fmap v2mw mv)
But, I don't understand how v2mw, which has a type of a -> m b type checks to the first argument of fmap.
fmap :: Functor f => (a -> b) -> f a -> f b
Let's say v2mw :: c -> m d, just so things aren't ambiguous, and
fmap :: Functor f => (a -> b) -> f a -> f b
Then fmap v2mw works out so that f ~ m, a ~ c and b ~ m d, so
fmap v2mw :: m c -> m (m d)
and join :: m (m e) -> m e, so join (fmap v2mw mv) has type m d as expected.
Related
In Learn you a Haskell, it is given that
fmap (+3) (*3)
is equivalent to
\x -> ((x*3)+3))
However, I can't understand why. Isn't it supposed to be \x -> ((x+3)*3)?
I don't know the implementation of fmap for (*3) functor, but my intuition tells me that since the functor (*3) is equivalent to \x -> x * 3, the map (+3) would be first applied and then (*3) be applied, but it is the other way around. What am I missing in here?
my intuition tells me that since the functor (*3) is equivalent to \x -> x * 3
Functions form a functor instance
instance Functor ((->) r) where ...
Those are the functions mapping from r.
Given a function g :: r -> a you can form a new function h :: r -> b with the help of f :: a -> b via h = fmap f g. It should now be clear that f :: a -> b cannot be applied first but must be applied second. That is h' = (g :: r -> a) . (f :: a -> b) does not make any sense but h = (f :: a -> b) . (g :: r -> a) does.
fmap has to obey two laws:
fmap id == id
fmap (f . g) == fmap f . fmap g
Your proposed definition, fmap' f g == g . f, satisfies the first law but violates the second law:
fmap' id f == f . id == f == id f -- OK
fmap' (f . g) h == h . (f . g)
== (h . f) . g
== (fmap' f h) . g
== fmap' g (fmap' f h)
== (fmap' g . fmap' f) h -- violation, since (.) is not commutative
The correct definition, fmap f g = f . h, satisfies both:
fmap id f == id . f == f == id f
fmap (f . g) h == (f . g) . h
== f . (g . h)
== fmap f (g . h)
== fmap f (fmap g h)
== (fmap f . fmap g) h
A "functor", in Haskell, is a higher order type, F, -- "higher order" meaning, it accepts another type variable, a (denoting any type whatever), -- such that we can have
F a fa
(a -> b) ab
-----------------
F b fmap ab fa
which is known as "flip fmap" (flip just means that the order of arguments is flipped, flip fmap fa ab == fmap ab fa). There are also some "common sense" laws it must follow.
For F ~ Maybe, say, it means
flip fmap :: Maybe a ->
(a -> b) ->
Maybe b
and for F ~ [],
flip fmap :: [] a ->
(a -> b) ->
[] b
which is more conventionally written as [a] -> (a -> b) -> [b].
In our case here, F a ~ r -> a, or more formally, F a ~ ((->) r) a, which means F ~ ((->) r),
flip fmap :: ((->) r) a ->
(a -> b) ->
((->) r) b
which is more conventionally written as (r -> a) -> (a -> b) -> (r -> b),
r -> a
a -> b
-------------
r -> b
which is the same as (r -> a) -> (a -> b) -> r -> b since with types, the arrows associate on the right, corresponding to the fact that applications associate on the left: f a b c is actually ((f a) b) c and
f a b c = d -- f a b c
f a b = \c -> d -- (f a b) c
f a = \b c -> d -- ((f a) b) c
f = \a b c -> d -- (((f) a) b) c
are all different ways to write down the same definition, and different ways to write down the same function call.
This then means we need to implement
fmap :: (a -> b) -> (r -> a) -> (r -> b)
fmap ab ra r = b
where
b =
So what could the definition be? Is it up to us to decode what goes where? Well, we must produce a b type value. The only thing we have that can do it for us, is ab :: a -> b.
Can we produce a b without it? Out of the blue? Except for erroring out, no, we can't -- we know nothing about that b type. It can be anything. So we're left with
b = ab a
a =
and now we must get an a somewhere, to use it as an argument to ab. Fortunately, ra can give it to us:
a = ra r
and r, we already got! So the types did write this implementation for us:
fmap :: (a -> b) -> (r -> a) -> (r -> b)
fmap ab ra r = b
where
b = ab a
a = ra r
or, simplifying and renaming, we get
fmap f g r = f ( g r)
= (f . g) r
by definition of the function composition, ., as
(.) :: (a -> b) -> (r -> a) -> (r -> b)
(f . g) r = f (g r)
which is a valid syntax definition, otherwise written as
(.) :: (a -> b) -> (r -> a) -> (r -> b)
(.) f g r = f (g r)
or
(.) :: (a -> b) -> (r -> a) -> (r -> b)
(.) f g = \ r -> f (g r)
or
(.) :: (a -> b) -> (r -> a) -> (r -> b)
(.) = \f g r -> f (g r)
All these are equivalent. And its type diagram
a -> b
r -> a
--------------
r -> b
As for the intuition, a functorial type value of type F A is, loosely, an F-type "something" than can somehow produce an A-type something, in some F-type sense.
The functor laws mean that F does so in some purely "structural", mechanical way, without regard to what that A value that it produces actually is. In other words, the A values do not influence how they are produced, only the F type itself determines that.
For example, Maybe Int could Maybe produce an Int. Or [Int] could produce several Ints. (*3) can also produce an Int, if we supply it with an Int argument.
What then this fmap is? What does it do? It transforms that would-be produced value. Every functor type must define its fmap, that is what makes it be a functorial type, that it defines the
instance Functor Maybe where
fmap ab (Just a) = (Just (ab a))
etc. So then, with functions r -> a, which produce that a type value promised by their type, after being applied to an argument, we transform that value by applying the transformation function to it:
fmap transf mult3 arg = tansf (mult3 arg)
which is just the definition of the functional composition itself, with arguments renamed.
So that's why, in this case,
fmap (+3) (*3) r = (+3) ((*3) r)
= (+3) (r*3)
= (r*3) + 3
we (+3) transform the value produced by (*3) in the ((->) r) sense, which is application to some user-supplied argument, r. So (*3) must be applied first, to get (it to produce) that value.
I'm learning Haskell by taking fp-course exercise. There is a question block my way. I don't know how Haskell infer lift2 (<$>) (,)'s type, and turn out Functor k => (a1 -> k a2) -> a1 -> k (a1, a2).
I have tried out lift2 (<$>)'s type, and verified by GHCI's command :t lift2 (<$>). step as follow.
I know lift2 :: Applicative k => (a -> b -> c) -> k a -> k b -> k c
I also know (<$>) :: Functor f => (m -> n) -> (f m) -> (f n)
Then by lambda calculus's Beta conversion, I can figure out lift2 (<$>)'s type is
(Applicative k, Functor f) => k (m -> n) -> k (f m) -> k (f n) by replacing a with (m -> n), b with (f m), c with (f n)
When I going to figure out lift2 (<$>) (,)'s type, It block me.
I know (,) :: a -> b -> (a,b)
And lift2 (<$>) :: (Applicative k, Functor f) => k (m -> n) -> k (f m) -> k (f n).
How does Haskell apply lift2 (<$>) to (,)?
The first variable of lift2 (<$>) is Applicative k => k (m -> n).
The to be applied value is (,) :: a -> b -> (a, b)
How the k, m, n replace by a, b?
GHCI's answer is lift2 (<$>) (,) :: Functor k => (a1 -> k a2) -> a1 -> k (a1, a2) by typing :t lift2 (<$>) (,). I cannot infer out this answer by myself.
So I have 2 questions.
1.Could someone show me the inference step by step?
2.In this case the conversion seems not be Beta conversion in lambda calculus (May be I am wrong). What the conversion is?
Type derivation is a mechanical affair.(*) The key is that the function arrow -> is actually a binary operator here, associating on the right (while the application / juxtaposition associates on the left).
Thus A -> B -> C is actually A -> (B -> C) is actually (->) A ((->) B C) is actually ((->) A) (((->) B) C). In this form it is clear that it consists of two parts so can match up with e.g. f t, noting the equivalences f ~ ((->) A) and t ~ (((->) B) C) (or in pseudocode f ~ (A ->), and also t ~ (B -> C) in normal notation).
When "applying" two type terms a structural unification is performed. The structures of two terms are matched up, their sub-parts are matched up, and the resulting equivalences are noted as "substitutions" (... ~ ...) available to be performed and ensured in further simplifications of the resulting type terms (and if some incompatibility were to be thus discovered, the type would be then rejected).
This follows a general structure / type derivation rule rooted in the logical rule of Modus Ponens:
A -> B C
--------------
B , where A ~ C
And thus,
liftA2 :: A f => ( a -> b -> c ) -> f a -> f b -> f c
(<$>) :: F h => (d -> e) -> h d -> h e
(,) :: s -> (t -> (s, t))
---------------------------------------------------------------------------------
liftA2 (<$>) (,) :: f b -> f c
---------------------------------------------------------------------------------
b ~ h d f ~ (s->)
a ~ d->e c ~ h e a ~ t->(s,t)
\_ _ _ _ _ _ _ _ _ _ _ _ _ _ a ~ d->e
----------------------------------------------------
d ~ t e ~ (s,t)
liftA2 (<$>) (,) :: f b -> f c
~ (s -> b ) -> (s -> c )
~ F h => (s -> h d) -> (s -> h e )
~ F h => (s -> h t) -> (s -> h (s,t))
(writing A for Applicative and F for Functor, as an abbreviation). The substitutions stop when there are no more type variables to substitute.
There's some freedom as to which type variables are chosen to be substituted on each step, but the resulting terms will be equivalent up to consistent renaming of the type variables, anyway. For example we could choose
~ F h => (s -> h d) -> (s -> h e )
~ F h => (s -> h d) -> (s -> h (s,t))
~ F h => (s -> h d) -> (s -> h (s,d))
The Applicative ((->) s) constraint was discovered in the process. It checks out since this instance exists for all s. We can see it by typing :i Applicative at the prompt in GHCi. Looking through the list of instances it prints, we find instance Applicative ((->) a) -- Defined in `Control.Applicative'.
If there were no such instance the type derivation would stop and report the error, it wouldn't just skip over it. But since the constraint holds, it just disappears as it does not constrain the derived type, Functor h => (s -> h t) -> (s -> h (s,t)). It's already "baked in".
The instance defines (f <*> g) x = f x $ g x but the definition itself is not needed in type derivations, only the fact that it exists. As for the liftA2, it is defined as
liftA2 h f g x = (h <$> f <*> g) x -- for any Applicative (sans the `x`)
= (h . f <*> g) x -- for functions
= (h . f) x (g x)
= f x `h` g x -- just another combinator
(yes, (<*>) = liftA2 ($) ), so
liftA2 (<$>) (,) g s = (,) s <$> g s
= do { r <- g s -- in pseudocode, with
; return (s, r) -- "Functorial" Do
}
Or in other words, liftA2 (<$>) (,) = \ g s -> (s ,) <$> g s.
With the type Functor m => (s -> m t) -> s -> m (s,t). Which is what we have derived.
(*) See also:
Haskell: how to infer the type of an expression manually
Some playing around with functors and monads in ghci led me to a value whose type and behaviour I would like to understand better.
The type of \x -> join . x is (Monad m) => (a -> m (m b)) -> (a -> m b) and the type of \y -> y . (flip fmap) is (Functor f) => ((a -> b) -> f b) -> (f a -> c).
Version 8.2.2 of ghci permits the definition h = join . (flip fmap).
Why does h have type ((A -> B) -> A) -> (A -> B) -> B?
In particular, why do the functor and monad constraints disappear? Is this really the correct and expected behaviour? As a follow up, I would also like to ask:
Why does evaluating h (\f -> f u) (\x -> x + v) for integers u and v give u + 2v in every case?
In short: due to type deduction, Haskell knows that m and f are in fact a partially instantiated arrow.
Deriving the type
Well let us do the math. The function join . (flip fmap) is basically your given lambda expression \x -> join . x with as argument (flip fmap), so:
h = (\x -> join . x) (flip fmap)
Now the lambda expression has type:
(\x -> join . x) :: Monad m => (a -> m (m b)) -> (a -> m b)
Now the argument flip fmap has type:
flip fmap :: Functor f => f c -> ((c -> d) -> f d)
(we here use c and d instead of a and b to avoid confusion between two possibly different types).
So that means that the type of flip fmap is the same as the type of the argument of the lambda expression, hence we know that:
Monad m => a -> m (m b)
~ Functor f => f c -> ((c -> d) -> f d)
---------------------------------------
a ~ f c, m (m b) ~ ((c -> d) -> f d)
So we now know that a has the same type as f c (this is the meaning of the tilde ~).
But we have to do some extra computations:
Monad m => m (m b)
~ Functor f => ((c -> d) -> f d)
--------------------------------
m ~ (->) (c -> d), m b ~ f d
Hence we know that m is the same as (->) (c -> d) (basically this is a function where we know that input type, here (c -> d), and the output type is a type parameter of m.
So that means that m b ~ (c -> d) -> b ~ f d, so this means that f ~ (->) (c -> d) and b ~ d. An extra consequence is that since a ~ f c, we know that a ~ (c -> d) -> c
So to list what we derived:
f ~ m
m ~ (->) (c -> d)
b ~ d
a ~ (c -> d) -> c
So we now can "specialize" the types of both our lambda expression, and our flip fmap function:
(\x -> join . x)
:: (((c -> d) -> c) -> (c -> d) -> (c -> d) -> d) -> ((c -> d) -> c) -> (c -> d) -> d
flip fmap
:: ((c -> d) -> c) -> (c -> d) -> (c -> d) -> d
and type of flip fmap now perfectly matches with the type of the argument of the lambda expression. So the type of (\x -> join . x) (flip fmap) is the result type of the lambda expression type, and that is:
(\x -> join . x) (flip fmap)
:: ((c -> d) -> c) -> (c -> d) -> d
But now we of course did not yet obtained the implementation of this function. We are however already a step further.
Deriving the implementation
Since we now know that m ~ (->) (c -> d), we know we should lookup the arrow instance of a monad:
instance Monad ((->) r) where
f >>= k = \ r -> k (f r) r
So for a given function f :: r -> a, as left operand, and a function k :: a -> (r -> b) ~ a -> r -> b as operand, we construct a new function that maps a variable x to k applied to f applied to x, and x. It is thus a way to perform some sort of preprocessing on an input variable x, and then do the processing both taking into account the preprocessing and the original view (well this is an interpretation a human reader can use).
Now join :: Monad m => m (m a) -> m a is implemented as:
join :: Monad m => m (m a) -> m a
join x = x >>= id
So for the (->) r monad, this means that we implement this as:
-- specialized for `m ~ (->) a
join f = \r -> id (f r) r
Since id :: a -> a (the identity function) returns its argument, we can further simplify it to:
-- specialized for `m ~ (->) a
join f = \r -> (f r) r
or cleaner:
-- specialized for `m ~ (->) a
join f x = f x x
So it basically is given a function f, and will then apply an argument twice to that function.
Furthermore we know that the Functor instance for the arrow type is defined as:
instance Functor ((->) r) where
fmap = (.)
So it is basically used as a "post processor" on the result of the function: we construct a new function that will do the post processing with the given function.
So now that we specialized the function enough for the given Functor/Monad, we can derive the implementation as:
-- alternative implementation
h = (.) (\f x -> f x x) (flip (.))
or by using more lambda expressions:
h = \a -> (\f x -> f x x) ((flip (.)) a)
which we can now further specialize as:
h = \a -> (\f x -> f x x) ((\y z -> z . y) a)
-- apply a in the lambda expression
h = \a -> (\f x -> f x x) (\z -> z . a)
-- apply (\z -> z . a) in the first lambda expression
h = \a -> (\x -> (\z -> z . a) x x)
-- cleaning syntax
h a = (\x -> (\z -> z . a) x x)
-- cleaning syntax
h a x = (\z -> z . a) x x
-- apply lambda expression
h a x = (x . a) x
-- remove the (.) part
h a x = x (a x)
So h basically takes two arguments: a and x, it then performs function application with a as function and x as parameter, and the output is passed to the x function again.
Sample usage
As sample usage you use:
h (\f -> f u) (\x -> x + v)
or nicer:
h (\f -> f u) (+v)
so we can analyze this like:
h (\f -> f u) (+v)
-> (+v) ((\f -> f u) (+v))
-> (+v) ((+v) u)
-> (+v) (u+v)
-> ((u+v)+v)
So we add u+v to v.
Types line up easier with >>>:
a -> b >>>
b -> c ::
a -> c
Here, we have
join . flip fmap == flip fmap >>> join
flip fmap :: Functor f => f a -> ((a -> b) -> f b )
join :: Monad m => (m (m b)) -> m b
----------------------------------------------------------
flip fmap >>> join ::
(Functor f, Monad m) => f a -> m b , ((a -> b) ->) ~ m, f ~ m
::
(Functor f, Monad f) => f a -> f b , f ~ ((a -> b) ->)
:: ((a -> b) -> a) -> ((a -> b) -> b)
Simple, mechanical, mundane.
To see what it does, combinatory style definitions are usually easiest to twiddle with,
(join . flip fmap) f g x =
join (flip fmap f) g x = -- join f x = f x x
(`fmap` f) g g x = -- f `fmap` g = f . g
(g . f) g x
g (f g) x
So we don't need x after all (or do we?). The join and fmap definitions for functions are given in the margins. We've arrived at
(join . flip fmap) f g = g (f g) -- f :: (a -> b) -> a, g :: a -> b
-- f g :: a , g (f g) :: b
Another way is starting from the types, going by the rule of modus ponens,
((a -> b) -> a) (a -> b) -- f g
---------------------------
(a -> b) a -- g (f g)
---------------------------------------
b
Applicative's has the (<*>) function:
(<*>) :: (Applicative f) => f (a -> b) -> f a -> f b
Learn You a Haskell shows the following function.
Given:
ap :: (Monad m) => m (a -> b) -> m a -> m b
ap f m = do
g <- f -- '<-' extracts f's (a -> b) from m (a -> b)
m2 <- m -- '<-' extracts a from m a
return (g m2) -- g m2 has type `b` and return makes it a Monad
How could ap be written with bind alone, i.e. >>=?
I'm not sure how to extract the (a -> b) from m (a -> b). Perhaps once I understand how <- works in do notation, I'll understand the answer to my above question.
How could ap be written with bind alone, i.e. >>= ?
This is one sample implementation I can come up with:
ap :: (Monad m) => m (a -> b) -> m a -> m b
ap xs a = xs >>= (\f -> liftM f a)
Of if you don't want to even use liftM then:
ap :: (Monad m) => m (a -> b) -> m a -> m b
ap mf ma = mf >>= (\f -> ma >>= (\a' -> return $ f a'))
Intially these are the types:
mf :: m (a -> b)
ma :: m a
Now, when you apply bind (>>=) operator to mf: mf >>= (\f-> ..., then f has the type of:
f :: (a -> b)
In the next step, ma is also applied with >>=: ma >>= (\a'-> ..., here a' has the type of:
a' :: a
So, now when you apply f a', you get the type b from that because:
f :: (a -> b)
a' :: a
f a' :: b
And you apply return over f a' which will wrap it with the monadic layer and hence the final type you get will be:
return (f a') :: m b
And hence everything typechecks.
In Haskell, what does the monad instance of functions give over just applicative? Looking at their implementations, they seem almost identical:
(<*>) f g x = f x (g x)
(>>=) f g x = g (f x) x
Is there anything you can do with >>= that you can't do with just <*>?
They are equivalent in power for the function instance: flip f <*> g == g >>= f. This is not true for most types that are instances of Monad though.
It's a little more clear if we compare <*> and =<< (which is flip (>>=)) specialized to the ((->) r) instance:
(<*>) :: Applicative f => f (a -> b) -> f a -> f b
-- Specialized to ((->) r):
(<*>) :: (r -> a -> b) -> (r -> a) -> r -> b
(=<<) :: Monad m => (a -> m b) -> m a -> m b
-- Specialized to ((->) r):
(=<<) :: (a -> r -> b) -> (r -> a) -> r -> b