Given
> foldr (+) 5 [1,2,3,4]
15
this second version
foldr (\x n -> x + n) 5 [1,2,3,4]
also returns 15. The first thing I don't understand about the second version is how foldr knows which variable is associated with the accumulator-seed 5 and which with the list variable's elements [1,2,3,4]. In the lambda calculus way, x would seem to be the dependent variable and n the independent variable. So if this
foldr :: (a -> b -> b) -> b -> [a] -> b
foldr _ z [] = z
foldr f z (x:xs) = f x (foldr f z xs)
is foldr and these
:type foldr
foldr :: Foldable t => (a -> b -> b) -> b -> t a -> b
:t +d foldr
foldr :: (a -> b -> b) -> b -> [a] -> b
its type declarations, can I glean, deduce the answer to "which is dependent and which is independent" from the type declaration itself? It would seem both examples of foldr above must be doing this
(+) 1 ((+) 2 ((+) 3 ((+) 4 ((+) 5 0))))
I simply guessed the second, lambda function version above, but I don't really understand how it works, whereas the first version with (+) breaks down as shown directly above.
Another example would be this
length' = foldr (const (1+)) 0
where, again, const seems to know to "throw out" the incoming list elements and simply increment, starting with the initial accumulator value. This is the same as
length' = foldr (\_ acc -> 1 + acc) 0
where, again, Haskell knows which of foldr's second and third arguments -- accumulator and list -- to treat as the dependent and independent variable, seemingly by magic. But no, I'm sure the answer lies in the type declaration (which I can't decipher, hence, this post), as well as the lore of lambda calculus, of which I'm a beginner.
Update
I've found this
reverse = foldl (flip (:)) []
and then applying to a list
foldl (flip (:)) [] [1,2,3]
foldl (flip (:)) (1:[]) [2,3]
foldl (flip (:)) (2:1:[]) [3]
foldl (flip (:)) (3:2:1:[]) []
. . .
Here it's obvious that the order is "accumulator" and then list, and flip is flipping the first and second variables, then subjecting them to (:). Again, this
reverse = foldl (\acc x -> x : acc) []
foldl (\acc x -> x : acc) [] [1,2,3]
foldl (\acc x -> x : acc) (1:[]) [1,2,3]
. . .
seems also to rely on order, but in the example from further above
length' = foldr (\_ acc -> 1 + acc) 0
foldr (\_ acc -> 1 + acc) 0 [1,2,3]
how does it know 0 is the accumulator and is bound to acc and not the first (ghost) variable? So as I understand (the first five pages of) lambda calculus, any variable that is "lambda'd," e.g., \x is a dependent variable, and all other non-lambda'd variables are independent. Above, the \_ is associated with [1,2,3] and the acc, ostensibly the independent variable, is 0; hence, order is not dictating assignment. It's as if acc was some keyword that when used always binds to the accumulator, while x is always talking about the incoming list members.
Also, what is the "algebra" in the type definition where t a is transformed to [a]? Is this something from category theory? I see
Data.Foldable.toList :: t a -> [a]
in the Foldable definition. Is that all it is?
By "dependent" you most probably mean bound variable.
By "independent" you most probably mean free (i.e. not bound) variable.
There are no free variables in (\x n -> x + n). Both x and n appear to the left of the arrow, ->, so they are named parameters of this lambda function, bound inside its body, to the right of the arrow. Being bound means that each reference to n, say, in the function's body is replaced with the reference to the corresponding argument when this lambda function is indeed applied to its argument(s).
Similarly both _ and acc are bound in (\_ acc -> 1 + acc)'s body. The fact that the wildcard is used here, is immaterial. We could just have written _we_dont_care_ all the same.
The parameters in lambda function definition get "assigned" (also called "bound") the values of the arguments in an application, purely positionally. The first argument will be bound / assigned to the first parameter, the second argument - to the second parameter. Then the lambda function's body will be entered and further reduced according to the rules.
This can be seen a bit differently stating that actually in lambda calculus all functions have only one parameter, and multi-parameter functions are actually nested uni-parameter lambda functions; and that the application is left-associative i.e. nested to the left.
What this actually means is quite simply
(\ x n -> x + n) 5 0
=
(\ x -> (\ n -> x + n)) 5 0
=
((\ x -> (\ n -> x + n)) 5) 0
=
(\ n -> 5 + n) 0
=
5 + 0
As to how Haskell knows which is which from the type signatures, again, the type variables in the functional types are also positional, with first type variable corresponding to the type of the first expected argument, the second type variable to the second expected argument's type, and so on.
It is all purely positional.
Thus, as a matter of purely mechanical and careful substitution, since by the definition of foldr it holds that foldr g 0 [1,2,3] = g 1 (foldr g 0 [2,3]) = ... = g 1 (g 2 (g 3 0)), we have
foldr (\x n -> x + n) 0 [1,2,3]
=
(\x n -> x + n) 1 ( (\x n -> x + n) 2 ( (\x n -> x + n) 3 0 ))
=
(\x -> (\n -> x + n)) 1 ( (\x n -> x + n) 2 ( (\x n -> x + n) 3 0 ))
=
(\n -> 1 + n) ( (\x n -> x + n) 2 ( (\x n -> x + n) 3 0 ))
=
1 + ( (\x n -> x + n) 2 ( (\x n -> x + n) 3 0 ))
=
1 + ( (\x (\n -> x + n)) 2 ( (\x n -> x + n) 3 0 ))
=
1 + (\n -> 2 + n) ( (\x n -> x + n) 3 0 )
=
1 + (2 + (\x n -> x + n) 3 0 )
=
1 + (2 + (\x -> (\n -> x + n)) 3 0 )
=
1 + (2 + (\n -> 3 + n) 0 )
=
1 + (2 + ( 3 + 0))
In other words, there is absolutely no difference between (\x n -> x + n) and (+).
As for that t in foldr :: Foldable t => (a -> b -> b) -> b -> t a -> b, what that means is that given a certain type T a, if instance Foldable T exists, then the type becomes foldr :: (a -> b -> b) -> b -> T a -> b, when it's used with a value of type T a.
One example is Maybe a and thus foldr (g :: a -> b -> b) (z :: b) :: Maybe a -> b.
Another example is [] a and thus foldr (g :: a -> b -> b) (z :: b) :: [a] -> b.
(edit:) So let's focus on lists. What does it mean for a function foo to have that type,
foo :: (a -> b -> b) -> b -> [a] -> b
? It means that it expects an argument of type a -> b -> b, i.e. a function, let's call it g, so that
foo :: (a -> b -> b) -> b -> [a] -> b
g :: a -> b -> b
-------------------------------------
foo g :: b -> [a] -> b
which is itself a function, expecting of some argument z of type b, so that
foo :: (a -> b -> b) -> b -> [a] -> b
g :: a -> b -> b
z :: b
-------------------------------------
foo g z :: [a] -> b
which is itself a function, expecting of some argument xs of type [a], so that
foo :: (a -> b -> b) -> b -> [a] -> b
g :: a -> b -> b
z :: b
xs :: [a]
-------------------------------------
foo g z xs :: b
And what could such function foo g z do, given a list, say, [x] (i.e. x :: a, [x] :: [a])?
foo g z [x] = b where
We need to produce a b value, but how? Well, g :: a -> b -> b produces a function b -> b given an value of type a. Wait, we have that!
f = g x -- f :: b -> b
and what does it help us? Well, we have z :: b, so
b = f z
And what if it's [] we're given? We don't have any as then at all, but we have a b type value, z -- so instead of the above we'd just define
b = z
And what if it's [x,y] we're given? We'll do the same f-building trick, twice:
f1 = g x -- f1 :: b -> b
f2 = g y -- f2 :: b -> b
and to produce b we have many options now: it's z! or maybe, it's f1 z!? or f2 z? But the most general thing we can do, making use of all the data we have access to, is
b = f1 (f2 z)
for a right-fold (...... or,
b = f2 (f1 z)
for a left).
And if we substitute and simplify, we get
foldr g z [] = z
foldr g z [x] = g x z -- = g x (foldr g z [])
foldr g z [x,y] = g x (g y z) -- = g x (foldr g z [y])
foldr g z [x,y,w] = g x (g y (g w z)) -- = g x (foldr g z [y,w])
A pattern emerges.
Etc., etc., etc.
A sidenote: b is a bad naming choice, as is usual in Haskell. r would be much much better -- a mnemonic for "recursive result".
Another mnemonic is the order of g's arguments: a -> r -> r suggests, nay dictates, that a list's element a comes as a first argument; r the recursive result comes second (the Result of Recursively processing the Rest of the input list -- recursively, thus in the same manner); and the overall result is then produced by this "step"-function, g.
And that's the essence of recursion: recursively process self-similar sub-part(s) of the input structure, and complete the processing by a simple single step:
a a
: `g`
[a] r
------------- -------------
[a] r
[a]
a [a]
--------
(x : xs) -> r
xs -> r
----------------------
( x , r ) -> r --- or, equivalently, x -> r -> r
Well, the foldr itself knows this by definition. It was defined in such way that its function argument accepts the accumulator as 2nd argument.
Just like when you write a div x y = ... function you are free to use y as dividend.
Maybe you got confused by the fact that foldr and foldl has swapped arguments in the accumulator funtions?
As Steven Leiva says here, a foldr (1) takes a list and replaces the cons operators (:) with the given function and (2) replaces the last empty list [] with the accumulator-seed, which is what the definition of foldr says it will do
foldr :: (a -> b -> b) -> b -> [a] -> b
foldr _ z [] = z
foldr f z (x:xs) = f x (foldr f z xs)
So de-sugared [1,2,3] is
(:) 1 ((:) 2 ((:) 3 []))
and the recursion is in effect replacing the (:) with f, and as we see in foldr f z (x:xs) = f x (foldr f z xs), the z seed value is going along for the ride until the base case where it is substituted for the [], fulfilling (1) and (2) above.
My first confusion was seeing this
foldr (\x n -> x + n) 0 [1,2,3]
and not understanding it would be expanded out, per definition above, to
(\x n -> x + n) 1 ((\x n -> x + n) 2 ((\x n -> x + n) 3 0 ))
Next, due to a weak understanding of how the actual beta reduction would progress, I didn't understand the second-to-third step below
(\x -> (\n -> x + n)) 1 ...
(\n -> 1 + n) ...
1 + ...
That second-to-third step is lambda calculus being bizarre all right, but is at the root of why (+) and (\x n -> x + n) are the same thing. I don't think it's pure lambda calculus addition, but it (verbosely) mimics addition in recursion. I probably need to jump back into lambda calculus to really grasp why (\n -> 1 + n) turns into 1 +
My worse mental block was thinking I was looking at some sort of eager evaluation inside the parentheses first
foldr ((\x n -> x + n) 0 [1,2,3,4])
where the three arguments to foldr would interact first, i.e., 0 would be bound to the x and the list member to the n
(\x n -> x + n) 0 [1,2,3,4]
0 + 1
. . . then I didn't know what to think. Totally wrong-headed, even though, as Will Ness points out above, beta reduction is positional in binding arguments to variables. But, of course, I left out the fact that Haskell currying means we follow the expansion of foldr first.
I still don't fully understand the type definition
foldr :: (a -> b -> b) -> b -> [a] -> b
other than to comment/guess that the first a and the [a] mean a is of the type of the members of the incoming list and that the (a -> b -> b) is a prelim-microcosm of what foldr will do, i.e., it will take an argument of the incoming list's type (in our case the elements of the list?) then another object of type b and produce an object b. So the seed argument is of type b and the whole process will finally produce something of type b, also the given function argument will take an a and ultimately give back an object b which actually might be of type a as well, and in fact is in the above example with integers... IOW, I don't really have a firm grasp of the type definition...
I've got a few exercises to prepare for the exam in Haskell/Prolog.
One Haskell task is to rewrite the function below:
original :: [Integer] -> Integer
original [] = 0
original (x:xs) | x < 20 = 5 * x - 3 + original xs
| otherwise = original xs
But the condition is that I am only allowed to remove the two "undifined" in the scheme below:
alternative :: [Integer] -> Integer
alternative = foldr undefined undefined
My problem is that I dont know how this could match the normal foldr structure with 3 parameters (function, "start value" or how is it called?,list)?
Maybe an equivalent example would be helpfull, not the full soultion please!
Futhermore I am not allowed to use "let" or "where".
Thank you for any help!
Sooo... I just followed the idea from #hugo to just first complete the task on the "normal" way, which works but is not allowed by our university correction tool:
alternative :: [Integer] -> Integer
alternative list = foldr (\ x y -> if x < 20 then 5*x -3 + y else y) 0 list
AND after try end error i got the solution:
alternative :: [Integer] -> Integer
alternative = foldr (\ x y -> if x < 20 then 5*x -3 + y else y) 0
A list like [1,4,2,5] is syntactical sugar for (:) 1 ((:) 4 ((:) 2 ((:) 5 []))). foldr f z basically replaces the (:) data constructor with f, and the empty list data constructor [] with z. So foldr f z will result in f 1 (f 4 (f 2 (f 5 z))).
Since you write original [] = 0, this thus means that for z, we can use 0. For f we can use if x < 20 then (+) (5*x-3) else id, since in case x < 20, we add 5*x-3 to the value, and otherwise, we do nothing with the recursively computed value.
We can thus make an alternative implementation that looks like:
alternative :: (Foldable f, Num a, Ord a) => f a -> a
alternative = foldr f 0
where f x ys | x < 20 = 5*x - 3 + ys
| otherwise = ys
or without the where clause with an inline lambda expression:
alternative :: (Foldable f, Num a, Ord a) => f a -> a
alternative = foldr (\x -> if x < 20 then (+) (5*x-3) else id) 0
When we fold a list with one or more elements inside as done below:
foldr (+) 0 [1,2,3]
We get:
foldr (+) 0 (1 : 2 : 3 : [])
foldr (+) 1 + (2 +(3 + 0)) // 6
Now when the list is empty:
foldr (+) 0 []
Result: foldr (+) 0 ([])
Since (+) is binary operator, it needs two arguments to complete but here we end up (+) 0. How does it result in 0 and not throwing error of partially applied function.
Short answer: you get the initial value z.
If you give foldl or foldr an empty list, then it returns the initial value. foldr :: (a -> b -> b) -> b -> t a -> b works like:
foldr f z [x1, x2, ..., xn] == x1 `f` (x2 `f` ... (xn `f` z)...)
So since there are no x1, ..., xn the function is never applied, and z is returned.
We can also inspect the source code:
foldr :: (a -> b -> b) -> b -> [a] -> b
-- foldr _ z [] = z
-- foldr f z (x:xs) = f x (foldr f z xs)
{-# INLINE [0] foldr #-}
-- Inline only in the final stage, after the foldr/cons rule has had a chance
-- Also note that we inline it when it has *two* parameters, which are the
-- ones we are keen about specialising!
foldr k z = go
where
go [] = z
go (y:ys) = y `k` go ys
So if we give foldr an empty list, then go will immediately work on that empty list, and return z, the initial value.
A cleaner syntax (and a bit less efficient, as is written in the comment of the function) would thus be:
foldr :: (a -> b -> b) -> b -> [a] -> b
foldr _ z [] = z
foldr f z (x:xs) = f x (foldr f z xs)
Note that - depending on the implementation of f - it is possible to foldr on infinite lists: if at some point f only looks at the initial value, and then returns a value, then the recursive part can be dropped.
The defined code is
fun foldl f e l = let
fun g(x, f'') = fn y => f''(f(x, y))
in foldr g (fn x => x) l e end
I don't understand how this works;
what is the purpose of g(x, f'')?
I also find a similar example in Haskell,
the definition is quite short
myFoldl f z xs = foldr step id xs z
where
step x g a = g (f a x)
Let's dissect the Haskell implementation of myFoldl and then take a look at the ocaml SML code. First, we'll look at some type signatures:
foldr :: (a -> b -> b) -- the step function
-> b -- the initial value of the accumulator
-> [a] -- the list to fold
-> b -- the result
It should be noted that although the foldr function accepts only three arguments we are applying it two four arguments:
foldr step id xs z
However, as you can see the second argument to foldr (i.e. the inital value of the accumulator) is id which is a function of the type x -> x. Therefore, the result is also of the type x -> x. Hence, it accepts four arguments.
Similarly, the step function is now of the type a -> (x -> x) -> x -> x. Hence, it accepts three arguments instead of two. The accumulator is an endofunction (i.e. a function whose domain and codomain is the same).
Endofunctions have a special property, they are composed from left to right instead of from right to left. For example, let's compose a bunch of Int -> Int functions:
inc :: Int -> Int
inc n = n + 1
dbl :: Int -> Int
dbl n = n * 2
The normal way to compose these functions is to use the function composition operator as follows:
incDbl :: Int -> Int
incDbl = inc . dbl
The incDbl function first doubles a number and then increments it. Note that this reads from right to left.
Another way to compose them is to use continuations (denoted by k):
inc' :: (Int -> Int) -> Int -> Int
inc' k n = k (n + 1)
dbl' :: (Int -> Int) -> Int -> Int
dbl' k n = k (n * 2)
Notice that the first argument is a continuation. If we want to recover the original functions then we can do:
inc :: Int -> Int
inc = inc' id
dbl :: Int -> Int
dbl = dbl' id
However, if we want to compose them then we do it as follows:
incDbl' :: (Int -> Int) -> Int -> Int
incDbl' = dbl' . inc'
incDbl :: Int -> Int
incDbl = incDbl' id
Notice that although we are still using the dot operator to compose the functions, it now reads from left to right.
This is the key behind making foldr behave as foldl. We fold the list from right to left but instead of folding it into a value, we fold it into an endofunction which when applied to an initial accumulator value actually folds the list from left to right.
Consider our incDbl function:
incDbl = incDbl' id
= (dbl' . inc') id
= dbl' (inc' id)
Now consider the definition of foldr:
foldr :: (a -> b -> b) -> b -> [a] -> b
foldr _ acc [] = acc
foldr fun acc (y:ys) = fun y (foldr fun acc ys)
In the basis case we simply return the accumulated value. However, in the inductive case we return fun y (foldr fun acc ys). Our step function is defined as follows:
step :: a -> (x -> x) -> x -> x
step x g a = g (f a x)
Here f is the reducer function of foldl and is of the type x -> a -> x. Notice that step x is an endofunction of the type (x -> x) -> x -> x which we know can be composed left to right.
Hence the folding operation (i.e. foldr step id) on a list [y1,y2..yn] looks like:
step y1 (step y2 (... (step yn id)))
-- or
(step y1 . step y2 . {dots} . step yn) id
Each step yx is an endofunction. Hence, this is equivalent to composing the endofunctions from left to right.
When this result is applied to an initial accumulator value then the list folds from left to right. Hence, myFoldl f z xs = foldr step id xs z.
Now consider the foldl function (which is written in Standard ML and not OCaml). It is defined as:
fun foldl f e l = let fun g (x, f'') = fn y => f'' (f (x, y))
in foldr g (fn x => x) l e end
The biggest difference between the foldr functions of Haskell and SML are:
In Haskell the reducer function has the type a -> b -> b.
In SML the reducer function has the type (a, b) -> b.
Both are correct. It's only a matter of preference. In SML instead of passing two separate arguments, you pass one single tuple which contains both arguments.
Now, the similarities:
The id function in Haskell is the anonymous fn x => x function in SML.
The step function in Haskell is the function g in SML which takes a tuple containing the first two arguments.
The step function is Haskell step x g a has been split into two functions in SML g (x, f'') = fn y => f'' (f (x, y)) for more clarity.
If we rewrite the SML function to use the same names as in Haskell then we have:
fun myFoldl f z xs = let step (x, g) = fn a => g (f (a, x))
in foldr step (fn x => x) xs z end
Hence, they are exactly the same function. The expression g (x, f'') simply applies the function g to the tuple (x, f''). Here f'' is a valid identifier.
Intuition
The foldl function traverses the list head to tail while operating elements with an accumulator:
(...(a⊗x1)⊗...⊗xn-1)⊗xn
And you want to define it via a foldr:
x1⊕(x2⊕...⊕(xn⊕e)...)
Rather unintuitive. The trick is that your foldr will not produce a value, but rather a function. The list traversal will operate the elements as to produce a function that, when applied to the accumulator, performs the computation you desire.
Lets see a simple example to illustrate how this works. Consider sum foldl (+) 0 [1,2,3] = ((0+1)+2)+3. We may calculate it via foldr as follows.
foldr ⊕ [1,2,3] id
-> 1⊕(2⊕(3⊕id))
-> 1⊕(2⊕(id.(+3))
-> 1⊕(id.(+3).(+2))
-> (id.(+3).(+2).(+1))
So when we apply this function to 0 we get
(id.(+3).(+2).(+1)) 0
= ((0+1)+2)+3
We began with the identity function and successively changed it as we traversed the list, using ⊕ where,
n ⊕ g = g . (+n)
Using this intuition, it isn't hard to define a sum with an accumulator via foldr. We built the computation for a given list via foldr ⊕ id xs. Then to calculate the sum we applied it to 0, foldr ⊕ id xs 0. So we have,
foldl (+) 0 xs = foldr ⊕ id xs 0
where n ⊕ g = g . (+n)
or equivalently, denoting n ⊕ g in prefix form by (⊕) n g and noting that (⊕) n g a = (g . (+n)) a = g (a+n),
foldl (+) 0 xs = foldr ⊕ id xs 0
where (⊕) n g a = g (a+n)
Note that the ⊕ is your step function, and that you can obtain the generic result you're looking for by substituting a function f for +, and accumulator a for 0.
Next let us show that the above really is correct.
Formal derivation
Moving on to a more formal approach. It is useful, for simplicity, to be aware of the following universal property of foldr.
h [] = e
h (x:xs) = f x (h xs)
iff
h = foldr f e
This means that rather than defining foldr directly, we may instead and more simply define a function h in the form above.
We want to define such an h so that,
h xs a = foldl f a xs
or equivalently,
h xs = \a -> foldl f a xs
So lets determine h. The empty case is simple:
h [] = \a -> foldl f a []
= \a -> a
= id
The non-empty case results in:
h (x:xs) = \a -> foldl f a (x:xs)
= \a -> foldl f (f a x) xs
= \a -> h xs (f a x)
= step x (h xs) where step x g = \a -> g (f a x)
= step x (h xs) where step x g a = g (f a x)
So we conclude that,
h [] = id
h (x:xs) = step x (h xs) where step x g a = g (f a x)
satisfies h xs a = foldl f a xs
And by the universal property above (noting that the f in the universal property formula corresponds to step here, and e to id) we know that h = foldr step id. Therefore,
h = foldr step id
h xs a = foldl f a xs
-----------------------
foldl f a xs = foldr step id xs a
where step x g a = g (f a x)
In the Haskell Wikibook, foldr is implemented as follows:
foldr :: (a -> b -> b) -> b -> [a] -> b
foldr f acc [] = acc
foldr f acc (x:xs) = f x (foldr f acc xs)
It is stated that the initial value of the accumulator is set as an argument. But as I understand it, acc is the identity value for the operation (e.g. 0 for sum or 1 for product) and its value does not change during the execution of the function. Why then is it referred to here and in other texts as an accumulator, implying that it changes or accumulates a value step by step?
I can see that an accumulator is relevant in a left fold, such as foldl, but is the wikibook explanation incorrect, and only for symmetry, in which case it is wrong?
Consider the evaluation of a simple foldr expression based on the (correct) definition you provided:
foldr (+) 0 [1,2,3,4]
= 1 + foldr (+) 0 [2,3,4]
= 1 + 2 + foldr (+) 0 [3,4]
= 1 + 2 + 3 + foldr (+) 0 [4]
= 1 + 2 + 3 + 4 + foldr (+) 0 []
= 1 + 2 + 3 + 4 + 0
= 10
So you are right: acc doesn't really "accumulate" anything. It never takes on a value other than 0.
Why is it called "acc" if it isn't an accumulator? Similarity to foldl? Hysterical raisins? A lie to children? I'm not sure.
Edit: I'll also point out that the GHC implementation of foldr uses z (presumably for zero) rather than acc.
acc doesn't really accumulate anything in the case of foldr as has been pointed out.
I'd add that without it, it's not clear what should happen when the input is an empty list.
It also changes the type signature of f, limiting the functions that can be used.
E.g:
foldr' :: (a -> a -> a) -> [a] -> a
foldr' f [] = error "empty list???"
foldr' f (x:[]) = x
foldr' f (x:xs) = f x (foldr' f xs)