I understand that it's impossible to pattern match functions in Haskell, and I fully understand why. However, I have two closely related questions. First, in cases where you'd like to partially apply functions for use later, is there a way of defining and capturing the return if it's a tuple? Or am I wrong, and this is still trying to pattern match functions under my nose?
For example, suppose I'm trying to get the quotient and remainder of a value with various multiples of ten. Then, how would I write something like this?
q, r :: Integral a => a -> a
(q, r) = (12345 `quotRem`)
I realize here, there are separate functions that exist, so I could do this instead:
q, r :: Integral a => a -> a
q = (12345 `quot`)
r = (12345 `rem`)
However, that's a very specific case, and there are unlimited other examples of functions that return tuples that would be nice to generalize. For example, a function that returns the number of evens and odds in a list.
evens, odds :: Integral a => [a] -> Int
(evens, odds) = (length . (filter even), length . (filter odd))
This leads me to my second question. The above works just fine in GHCi.
Prelude> let (evens, odds) = (length . (filter even), length . (filter odd))
Prelude> :t evens
evens :: Integral a => [a] -> Int
Prelude> evens [1..10]
5
What's even more confusing is it even works by "pattern-matching" in the same way that I was playing with (q, r) in the beginning:
Prelude> let evensOdds = (length . (filter even), length . (filter odd))
Prelude> :t evensOdds
evensOdds :: (Integral a1, Integral a) => ([a1] -> Int, [a] -> Int)
Prelude> let (ev,od) = evensOdds
Prelude> :t ev
ev :: Integral a1 => [a1] -> Int
Prelude> ev [1..10]
5
It also works just fine in an actual file loaded into GHCi, even though (evens, odds) doesn't. Why are these two different, and why does the second one work in GHCi at all if it doesn't work normally? Can what's different here be leveraged in some way?
You never pattern matched on a function. You always pattern matched on the pair-constructor (,). Your (even, odds) example
(evens, odds) = (length . (filter even), length . (filter odd))
just works like
(first, second) = (x, y)
It doesn't matter what type x and y have at that point.
Your (q, r) example doesn't work due to quotRem's type. Let's recall it and compare it with (q, r)'s type:
quotRem :: Integral n => n -> n -> (n , n)
quotRem 12345 :: Integral n => n -> (n , n)
(q, r) :: Integral n => (n -> n, n -> n)
As you can see, the pair (q, r)'type differs from quotRem's one. Still, it's possible to write your function:
pairify :: (a -> (b, c)) -> (a -> b, a -> c)
pairify f = (fst . f, snd . f)
(q,r) = pairify (quotRem 12345)
But as you can see we don't gain too much from pairify. By the way, partition from Data.List provides your (even, odds) functionality:
(even, odds) = pairify (partition even)
Look at the type of (12345 `quotRem`):
Integral a => a -> (a, a)
It’s a single function that returns a tuple. If you want to make this into a tuple of functions, you can compose it with fst and snd:
(q, r) = (fst . f, snd . f)
where f = (12345 `quotRem`)
If you want to do this in a point-free way, one way is to use the &&& combinator from Control.Arrow. Its fully general type is:
Arrow a => a b c -> a b d -> a b (c, d)
Specialised to the -> arrow, that’s:
(b -> c) -> (b -> d) -> b -> (c, d)
So it takes two functions, each taking a value of type b, and returns both their results (of types c and d) in a tuple. So here you can do something like this:
split = (fst .) &&& (snd .)
(q, r) = split (12345 `quotRem`)
Whereas if you look at the type of (length . filter even, length . filter odd), it’s a tuple already,
(Integral a, Integral b) => ([a] -> Int, [b] -> Int)
Which is why of course you can destructure this tuple to bind evens and odds.
Related
I want to fix this code
h :: (a -> b) -> [a] -> [b]
h f = foldr (\x y -> f x : y) []
if i put h (+100) [1,2,3,4,5] in GHCI
it returns to me [101,202,303,404,505]
when i put h (*10) [1,2,3,4,5] then
i want to get [10,200,3000,40000,500000] list
can anyone help me fixing this code?
You here implemented a map, but in order to repeat the same operation multiple times, you need to perform a mapping on the tail y:
h :: (a -> a) -> [a] -> [a]
h f = foldr (\x y -> f x : map f y) []
Solving the general problem, as Willem Van Onsem's answer does, requires O(n^2) time to calculate the first n elements, because the function has to be applied k times to calculate the kth element.
To solve this sort of problem efficiently, you will need to take advantage of some additional structure. Based on your examples, I think the most obvious approach is to think about semigroup actions. That is, instead of applying an arbitrary function repeatedly, look for an efficient way to represent the compositions of the function. For example, (*x) can be represented by x, allowing (*x) . (*y) to be represented by x*y.
To apply this idea, we first need to transform Willem's solution to make the compositions explicit.
h :: (a -> a) -> [a] -> [a]
h f0 as0 = go as0 f0
where
go [] _ = []
go (a:as) f = f a : go as (f0 . f)
If we like, we can write that as a fold:
h :: (a -> a) -> [a] -> [a]
h f0 as = foldr go stop as f0
where
stop _ = []
go a r f = f a : r (f0 . f)
Now we've structured the function using an accumulator (which is a function). As we compose onto the accumulator, it will get slower and slower to apply it. We want to replace that accumulator with one we can "apply" quickly.
{-# language BangPatterns #-}
import Data.Semigroup
repeatedly :: Semigroup s => (s -> a -> a) -> s -> [a] -> [a]
repeatedly act s0 as = foldr go stop as s0
where
stop _ = []
go a r !s = act s a : r (s0 <> s)
Now you can use, for example,
repeatedly (\(Product s) -> (s*)) (Product 10) [1..5]
==> [10,200,3000,40000,500000]
repeatedly (\(Sum s) -> (s+)) (Sum 100) [1..5]
==> [101,202,303,404,505]
In each of these, you accumulate a product/sum which is added to/multiplied by the current list element.
I need a function that does this:
>>> func (+1) [1,2,3]
[[2,2,3],[2,3,3],[2,3,4]]
My real case is more complex, but this example shows the gist of the problem. The main difference is that in reality using indexes would be infeasible. The List should be a Traversable or Foldable.
EDIT: This should be the signature of the function:
func :: Traversable t => (a -> a) -> t a -> [t a]
And closer to what I really want is the same signature to traverse but can't figure out the function I have to use, to get the desired result.
func :: (Traversable t, Applicative f) :: (a -> f a) -> t a -> f (t a)
It looks like #Benjamin Hodgson misread your question and thought you wanted f applied to a single element in each partial result. Because of this, you've ended up thinking his approach doesn't apply to your problem, but I think it does. Consider the following variation:
import Control.Monad.State
indexed :: (Traversable t) => t a -> (t (Int, a), Int)
indexed t = runState (traverse addIndex t) 0
where addIndex x = state (\k -> ((k, x), k+1))
scanMap :: (Traversable t) => (a -> a) -> t a -> [t a]
scanMap f t =
let (ti, n) = indexed (fmap (\x -> (x, f x)) t)
partial i = fmap (\(k, (x, y)) -> if k < i then y else x) ti
in map partial [1..n]
Here, indexed operates in the state monad to add an incrementing index to elements of a traversable object (and gets the length "for free", whatever that means):
> indexed ['a','b','c']
([(0,'a'),(1,'b'),(2,'c')],3)
and, again, as Ben pointed out, it could also be written using mapAccumL:
indexed = swap . mapAccumL (\k x -> (k+1, (k, x))) 0
Then, scanMap takes the traversable object, fmaps it to a similar structure of before/after pairs, uses indexed to index it, and applies a sequence of partial functions, where partial i selects "afters" for the first i elements and "befores" for the rest.
> scanMap (*2) [1,2,3]
[[2,2,3],[2,4,3],[2,4,6]]
As for generalizing this from lists to something else, I can't figure out exactly what you're trying to do with your second signature:
func :: (Traversable t, Applicative f) => (a -> f a) -> t a -> f (t a)
because if you specialize this to a list you get:
func' :: (Traversable t) => (a -> [a]) -> t a -> [t a]
and it's not at all clear what you'd want this to do here.
On lists, I'd use the following. Feel free to discard the first element, if not wanted.
> let mymap f [] = [[]] ; mymap f ys#(x:xs) = ys : map (f x:) (mymap f xs)
> mymap (+1) [1,2,3]
[[1,2,3],[2,2,3],[2,3,3],[2,3,4]]
This can also work on Foldable, of course, after one uses toList to convert the foldable to a list. One might still want a better implementation that would avoid that step, though, especially if we want to preserve the original foldable type, and not just obtain a list.
I just called it func, per your question, because I couldn't think of a better name.
import Control.Monad.State
func f t = [evalState (traverse update t) n | n <- [0..length t - 1]]
where update x = do
n <- get
let y = if n == 0 then f x else x
put (n-1)
return y
The idea is that update counts down from n, and when it reaches 0 we apply f. We keep n in the state monad so that traverse can plumb n through as you walk across the traversable.
ghci> func (+1) [1,1,1]
[[2,1,1],[1,2,1],[1,1,2]]
You could probably save a few keystrokes using mapAccumL, a HOF which captures the pattern of traversing in the state monad.
This sounds a little like a zipper without a focus; maybe something like this:
data Zippy a b = Zippy { accum :: [b] -> [b], rest :: [a] }
mapZippy :: (a -> b) -> [a] -> [Zippy a b]
mapZippy f = go id where
go a [] = []
go a (x:xs) = Zippy b xs : go b xs where
b = a . (f x :)
instance (Show a, Show b) => Show (Zippy a b) where
show (Zippy xs ys) = show (xs [], ys)
mapZippy succ [1,2,3]
-- [([2],[2,3]),([2,3],[3]),([2,3,4],[])]
(using difference lists here for efficiency's sake)
To convert to a fold looks a little like a paramorphism:
para :: (a -> [a] -> b -> b) -> b -> [a] -> b
para f b [] = b
para f b (x:xs) = f x xs (para f b xs)
mapZippy :: (a -> b) -> [a] -> [Zippy a b]
mapZippy f xs = para g (const []) xs id where
g e zs r d = Zippy nd zs : r nd where
nd = d . (f e:)
For arbitrary traversals, there's a cool time-travelling state transformer called Tardis that lets you pass state forwards and backwards:
mapZippy :: Traversable t => (a -> b) -> t a -> t (Zippy a b)
mapZippy f = flip evalTardis ([],id) . traverse g where
g x = do
modifyBackwards (x:)
modifyForwards (. (f x:))
Zippy <$> getPast <*> getFuture
> type Client = (Text, WS.Connection)
The state kept on the server is simply a list of connected clients. We've added
an alias and some utility functions, so it will be easier to extend this state
later on.
> type ServerState = [Client]
Check if a user already exists (based on username):
> clientExists :: Client -> ServerState -> Bool
> clientExists client = any ((== fst client) . fst)
Remove a client:
> removeClient :: Client -> ServerState -> ServerState
> removeClient client = filter ((/= fst client) . fst)
This is a literal haskell code taken from websockets. I don't understand how does clientExists function works,
clientExists client = any ((== fst client) . fst)
This function is invoked as,
clientExists client clients
So, how does the function refer the second argument clients? and what does . operator do?
And again at removeClient, what does the operator `/=' stand for?
The . operator is the function composition operator, it's defined as
f . g = \x -> f (g x)
The /= operator is "not equals", usually written as != in other languages.
The clientExists function has two arguments, but the second one is left off since it's redundant. It could have been written as
clientExists client clients = all ((== fst client) . fst) clients
But Haskell allows you to drop the last argument in situations like this. The any function has the type (a -> Bool) -> [a] -> Bool, and the function any ((== fst client) . fst) has the type [a] -> Bool. This is saying that the function clientExists client is the same function as any ((== fst client) . fst).
Another way to think of it is that Haskell does not have multi-argument functions, only single argument functions that return new functions. This is because -> is right associative, so a type signature like
a -> b -> c -> d
Can be written as
a -> (b -> (c -> d))
without changing its meaning. With the second type signature it's more clear that you have a function that when given an a, returns a function of type b -> (c -> d). If it's next given a b, it returns a function of type c -> d. Finally, if this is given a c it just returns a d. Since function application in Haskell is so cheap (just a space), this is transparent, but it comes in handy. For example, it means that you can write code like
incrementAll = map (+1)
or
onlyPassingStudents = filter ((>= 70) . grade)
In both of these cases I've also used operator sections, where you can supply either argument to an operator, and so long as its wrapped in parens it works. Internally it looks more like
(x +) = \y -> x + y
(+ x) = \y -> y + x
Where you can swap out the + for any operator you please. If you were to expand the definition of clientExists to have all argument specified it would look more like
clientExists client clients = any (\c -> fst c == fst client) clients
This definition is exactly equivalent to the one you have, just de-sugared to what the compiler really uses internally.
When in doubt, use the GHCi interpreter to find out the types of the functions.
First off, the /= operator is the not-equal:
ghci> :t (/=)
(/=) :: Eq a => a -> a -> Bool
ghci> 5 /= 5
False
ghci> 10 /= 5
True
. is the composition of two functions. It glues two functions together, just like in mathematics:
ghci> :t (.)
(.) :: (b -> c) -> (a -> b) -> a -> c
ghci> :t head.tail
head.tail :: [c] -> c
ghci> (head.tail) [1, 2, 3]
2
With the basics covered, let's see how it is used in your function definition:
ghci> :t (\x -> (== fst x))
(\x-> (== fst x)) :: Eq a => (a, b) -> a -> Bool
ghci> :t (\x-> (== fst x) . fst)
(\x-> (== fst x) . fst) :: Eq b => (b, b1) -> (b, b2) -> Bool
ghci> (\x -> (== fst x) . fst) (1, "a") (1, "b")
True
ghci> (\x -> (== fst x) . fst) (1, "a") (2, "b")
False
As we can see, the (== fst x) . fst is used to take two tuples, and compare the first element each for equality. Now, this expression (let's call it fstComp) has type fstComp :: Eq b => (b, b1) -> (b, b2) -> Bool, but we are already passing it a defined tuple (client :: (Text, WS.Connection)), we curry it to (Text, b2) -> Bool.
Since we have any :: (a -> Bool) -> [a] -> Bool, we can unify the first parameter with the previous type to have an expression of type (Text, b2) -> [(Text, b2)] -> Bool. Instantiating b2 = WS.Connection we get the type of clientExists :: (Text, WS.Connection) -> [(Text, WS.Connection)] -> Bool, or using the type synonyms, clientExists :: Client -> ServerState -> Bool.
Haskell newb here
I'm working on this problem in haskell:
(**) Eliminate consecutive duplicates of list elements.
If a list contains repeated elements they should be replaced with a single copy of the element. The order of the elements should not be changed.
Example:
* (compress '(a a a a b c c a a d e e e e))
(A B C A D E)
The solution (which I had to look up) uses foldr:
compress' :: (Eq a) => [a] -> [a]
compress' xs = foldr (\x acc -> if x == (head acc) then acc else x:acc) [last xs] xs
This foldr, according to the solution, takes two parameters, x and acc. It would seem like all foldr's take these parameters; is there any exception to this? Like a foldr that takes 3 or more? If not, is this convention redundant and can the formula be written with less code?
foldr takes a function of 2 arguments, but this doesn't prevent it from taking a function of 3 arguments provided that function has the right type signature.
If we had a function
g :: x -> y -> z -> w
With
foldr :: (a -> b -> b) -> b -> [a] -> b
Where we want to pass g to foldr, then (a -> b -> b) ~ (x -> y -> z -> w) (where ~ is type equality). Since -> is right associative, this means we can write g's signature as
x -> y -> (z -> w)
and its meaning is the same. Now we've produced a function of two parameters that returns a function of one parameter. In order to unify this with the type a -> b -> b, we just need to line up the arguments:
a -> | x ->
b -> | y ->
b | (z -> w)
This means that b ~ z -> w, so y ~ b ~ z -> w and a ~ x so g's type really has to be
g :: x -> (z -> w) -> (z -> w)
implying
foldr g :: (z -> w) -> [x] -> (z -> w)
This is certainly not impossible, although more unlikely. Our accumulator is a function instead, and to me this begs to be demonstrated with DiffLists:
type DiffList a = [a] -> [a]
append :: a -> DiffList a -> DiffList a
append x dl = \xs -> dl xs ++ [x]
reverse' :: [a] -> [a]
reverse' xs = foldr append (const []) xs $ []
Note that foldr append (const []) xs returns a function which we apply to [] to reverse a list. In this case we've given an alias to functions of the type [a] -> [a] called DiffList, but it's really no different than having written
append :: a -> ([a] -> [a]) -> [a] -> [a]
which is a function of 3 arguments.
As with all things in haskell have a look at the types of things to guide your way you can do this for any function in ghci.
Looking at this for foldr we see:
Prelude> :t foldr
foldr :: (a -> b -> b) -> b -> [a] -> b
This slightly abstract string can be written in english as:
foldr is a function that takes
1 ) a function with two parameters one of type a and one of type b and returns something of type b
2 ) A value of type b
3 ) A list of values of type a
And returns a value of type b
Where a and b are type variables (see here for a good tutorial on them) which can be filled in with any type you like.
It turns out that you can solve your compress problem using a foldr with a three-argument function.
compress :: Eq a => [a] -> [a]
compress [] = []
compress (z:zs) = z : foldr f (const []) zs z
where f x k w | x==w = k x
| otherwise = x : k x
Let's dissect that. First, we can improve readability by changing the last two lines to
where f x k = \w -> if x==w then k x else x : k x
This makes it evident that a ternary function is nothing but a binary function returning a unary function. The advantage of looking at it in this way is that foldr is best understood when passed a binary function. Indeed, we are passing a binary function, which just happens to return a function.
Let's focus on types now:
f :: a -> (a -> [a]) -> (a -> [a])
f x k
So, x::a is the element of the list we are folding on. Function k is the result of the fold on the list tail. The result of f x k is something having the same type as k.
\w -> if .... :: (a -> [a])
The overall idea behind this anonymous function is as follows. The parameter k plays the same role as acc in the OP code, except it is a function expecting the previous element w in the list before producing the accumulated compressed list.
Concretely, we use now k x when we used acc, passing on the current element to the next step, since by that time x will become the previous element w. At the top-level, we pass z to the function which is returned by foldr f (const []).
This compress variant is lazy, unlike the posted solution. In fact, the posted solution needs to scan the whole list before starting producing something: this is due to (\x acc -> ...) being strict in acc, and to the use of last xs. Instead, the above compress outputs list elements in a "streaming" fashion. Indeed, it works with infinite lists as well:
> take 10 $ compress [1..]
[1,2,3,4,5,6,7,8,9,10]
That being said, I think using a foldr here feels a bit weird: the code above is arguably less readable than the explicit recursion.
I'm working my way through the Project Euler problems in Haskell. I have got a solution for Problem 3 below, I have tested it on small numbers and it works, however due to the brute force implementation by deriving all the primes numbers first it is exponentially slow for larger numbers.
-- Project Euler 3
module Main
where
import System.IO
import Data.List
main = do
hSetBuffering stdin LineBuffering
putStrLn "This program returns the prime factors of a given integer"
putStrLn "Please enter a number"
nums <- getPrimes
putStrLn "The prime factors are: "
print (sort nums)
getPrimes = do
userNum <- getLine
let n = read userNum :: Int
let xs = [2..n]
return $ getFactors n (primeGen xs)
--primeGen :: (Integral a) => [a] -> [a]
primeGen [] = []
primeGen (x:xs) =
if x >= 2
then x:primeGen (filter (\n->n`mod` x/=0) xs)
else 1:[2]
--getFactors
getFactors :: (Integral a) => a -> [a] -> [a]
getFactors n xs = [ x | x <- xs, n `mod` x == 0]
I have looked at the solution here and can see how it is optimised by the first guard in factor. What I dont understand is this:
primes = 2 : filter ((==1) . length . primeFactors) [3,5..]
Specifically the first argument of filter.
((==1) . length . primeFactors)
As primeFactors is itself a function I don't understand how it is used in this context. Could somebody explain what is happening here please?
If you were to open ghci on the command line and type
Prelude> :t filter
You would get an output of
filter :: (a -> Bool) -> [a] -> [a]
What this means is that filter takes 2 arguments.
(a -> Bool) is a function that takes a single input, and returns a Bool.
[a] is a list of any type, as longs as it is the same type from the first argument.
filter will loop over every element in the list of its second argument, and apply it to the function that is its first argument. If the first argument returns True, it is added to the resulting list.
Again, in ghci, if you were to type
Prelude> :t (((==1) . length . primeFactors))
You should get
(((==1) . length . primeFactors)) :: a -> Bool
(==1) is a partially applied function.
Prelude> :t (==)
(==) :: Eq a => a -> a -> Bool
Prelude> :t (==1)
(==1) :: (Eq a, Num a) => a -> Bool
It only needs to take a single argument instead of two.
Meaning that together, it will take a single argument, and return a Boolean.
The way it works is as follows.
primeFactors will take a single argument, and calculate the results, which is a [Int].
length will take this list, and calculate the length of the list, and return an Int
(==1) will
look to see if the values returned by length is equal to 1.
If the length of the list is 1, that means it is a prime number.
(.) :: (b -> c) -> (a -> b) -> a -> c is the composition function, so
f . g = \x -> f (g x)
We can chain more than two functions together with this operator
f . g . h === \x -> f (g (h x))
This is what is happening in the expression ((==1) . length . primeFactors).
The expression
filter ((==1) . length . primeFactors) [3,5..]
is filtering the list [3, 5..] using the function (==1) . length . primeFactors. This notation is usually called point free, not because it doesn't have . points, but because it doesn't have any explicit arguments (called "points" in some mathematical contexts).
The . is actually a function, and in particular it performs function composition. If you have two functions f and g, then f . g = \x -> f (g x), that's all there is to it! The precedence of this operator lets you chain together many functions quite smoothly, so if you have f . g . h, this is the same as \x -> f (g (h x)). When you have many functions to chain together, the composition operator is very useful.
So in this case, you have the functions (==1), length, and primeFactors being compose together. (==1) is a function through what is called operator sections, meaning that you provide an argument to one side of an operator, and it results in a function that takes one argument and applies it to the other side. Other examples and their equivalent lambda forms are
(+1) => \x -> x + 1
(==1) => \x -> x == 1
(++"world") => \x -> x ++ "world"
("hello"++) => \x -> "hello" ++ x
If you wanted, you could re-write this expression using a lambda:
(==1) . length . primeFactors => (\x0 -> x0 == 1) . length . primeFactors
=> (\x1 -> (\x0 -> x0 == 1) (length (primeFactors x1)))
Or a bit cleaner using the $ operator:
(\x1 -> (\x0 -> x0 == 1) $ length $ primeFactors x1)
But this is still a lot more "wordy" than simply
(==1) . length . primeFactors
One thing to keep in mind is the type signature for .:
(.) :: (b -> c) -> (a -> b) -> a -> c
But I think it looks better with some extra parentheses:
(.) :: (b -> c) -> (a -> b) -> (a -> c)
This makes it more clear that this function takes two other functions and returns a third one. Pay close attention the the order of the type variables in this function. The first argument to . is a function (b -> c), and the second is a function (a -> b). You can think of it as going right to left, rather than the left to right behavior that we're used to in most OOP languages (something like myObj.someProperty.getSomeList().length()). We can get this functionality by defining a new operator that has the reverse order of arguments. If we use the F# convention, our operator is called |>:
(|>) :: (a -> b) -> (b -> c) -> (a -> c)
(|>) = flip (.)
Then we could have written this as
filter (primeFactors |> length |> (==1)) [3, 5..]
And you can think of |> as an arrow "feeding" the result of one function into the next.
This simply means, keep only the odd numbers that have only one prime factor.
In other pseodo-code: filter(x -> length(primeFactors(x)) == 1) for any x in [3,5,..]