permutation in a list with repetition ocaml - ocaml

I have a function that makes combinations of K distinct objects chosen from the N elements of a list, the problem is doesn't permute with repetition, for example:
extract 2 ["a";"b";"c";"d"];;
- : string list list =
[["a"; "b"]; ["a"; "c"]; ["a"; "d"]; ["b"; "c"]; ["b"; "d"]; ["c"; "d"]]
Here is my code:
# let rec extract k list =
if k <= 0 then [ [] ]
else match list with
| [] -> []
| h :: tl ->
let with_h = List.map (fun l -> h :: l) (extract (k-1) tl) in
let without_h = extract k tl in
with_h # without_h;;
I'm trying figure out how would do that, thanks for replying.

I have a function that makes combinations of K distinct objects
Indeed.
the problem is doesn't permute with repetition,
It's not much more difficult than combinations. There are two ways to go about your inductive step. If you want to permute k elements of a (non-empty) set A you can:
Method 1
For every element x in A, calculate the permutations of k-1 elements of A\{x} and put x in front of all those permutations. Then join the sets of solutions you obtained for each x.
Method 2
Choose one x in A, calculate the permutations of k-1 elements of A\{x}, then for each permutation found calculate new ones by inserting x in each possible position.

Related

Efficiently generating the cartesian product of lists with > 50 elements

I am trying to generate all possible combinations from a list over 50 elements, I was working in a previous implementation with this function:
let rec extract k list =
if k <= 0 then [ [] ]
else
match list with
| [] -> []
| h :: tl ->
let with_h = List.map (fun l -> h :: l) (extract (k-1) tl) in
let without_h = extract k tl in
with_h # without_h
And everything was OK, the point is that in my previous application the size of the list was small, between 5 and 10 items. However now my list will in some cases have more than 50 items, and this and other functions for generating combinations crash (with overflow memory messages in some cases). Is there a way to generate a function that can manage to generate combinations with very big lists?
A generic solution is to switch to sequences in order to lazily compute elements of the combination:
let rec extract k list () =
if k <= 0 then Seq.return [] ()
else match list with
| [] -> Seq.Nil
| h :: tl ->
let with_h = Seq.map (fun l -> h :: l) (extract (k-1) tl) in
let without_h = extract k tl in
Seq.append with_h without_h ()
let test = extract 20 (List.init 1_000_000 Fun.id)
let view = List.of_seq (Seq.take 20 l)
This solution avoids both the issue of stack overflows due to non-tailrecursive functions and the issue of eagerly generating a list of size n^k which would not fit in memory for large n.

Partitioning a list fairly

I'm sure this is a reasonably common thing but I can't find anything on it (my internet-search-fu is not strong).
I have a function that can group a list into a list of lists of N elements each, with the final sublist being smaller than N if the length of the list is not evenly divisible by N. Some examples:
groupEvery 2 [1,2,3,4] = [[1,2],[3,4]]
groupEvery 4 [1,2,3,4,5,6,7,8,9,10] = [[1,2,3,4], [5,6,7,8], [9,10]]
What I want is to take a list and a positive integer n (in the above examples n could be said to be 2 and 3) and partition it into a new list of n lists. It should work on a list of any type, and produce sublists whose sizes differ as little as possible.
So I would like to have:
fairPartition 3 [1,2,3,4,5,6,7,8,9,10] = [[1,2,3,4], [5,6,7], [8,9,10]]
Or any combination of sublists as long as there are two of length 3 and one of length 4.
A naive attempt using groupEvery:
fairPartition :: Int -> [a] -> [[a]]
fairPartition n xs = groupEvery ((length xs `div` n) + 1) xs
fairPartition 4 [1..10] = [[1,2,3],[4,5,6],[7,8,9],[10]]
but as you can see (3,3,3,1) is not a fair distribution of lengths, and for lists of smaller lengths it doesn't even return the right number of sublists:
# Haskell, at GHCi
*Main> let size = 4 in map (\l -> length . fairPartition 4 $ [1..l]) [size..25]
[2,3,3,4,3,3,4,4,3,4,4,4,4,4,4,4,4,4,4,4,4,4]
I would like a {pseudo,actual}-code function or explanation that is easily translatable to Haskell (the identity translation would be the best!).
Thanks.
You can use the split package's splitPlaces function for this.
import Data.List.Split
fairPartition n xs = case length xs `quotRem` n of
(q, r) -> splitPlaces (replicate r (q+1) ++ replicate (n-r) q) xs

Haskell: List Boundary

I have a list of doubles(myList), which I want to add to a new List (someList), but once the new list reaches a set size i.e. 25, I want to stop adding to it. I have tried implementing this function using sum but was unsuccessful. Example code below.
someList = [(a)| a <- myList, sum someList < 30]
The way #DanielFischer phrased the question is compatible with the Haskell way of thinking.
Do you want someList to be the longest prefix of myList that has a sum < 30?
Here's how I'd approach it: let's say our list is
>>> let list = [1..20]
we can find the "cumulative sums" using:
>>> let sums = tail . scanl (+) 0
>>> sums list
[1,3,6,10,15,21,28,36,45,55,66,78,91,105,120,136,153,171,190,210]
Now zip that with the original list to get pairs of elements with the sum up to that point
>>> zip list (sums list)
[(1,1),(2,3),(3,6),(4,10),(5,15),(6,21),(7,28),(8,36),
(9,45),(10,55),(11,66),(12,78),(13,91),(14,105),(15,120),
(16,136),(17,153),(18,171),(19,190),(20,210)]
Then we can takeWhile this list to get the prefix we want:
>>> takeWhile (\x -> snd x < 30) (zip list (sums list))
[(1,1),(2,3),(3,6),(4,10),(5,15),(6,21),(7,28)]
finally we can get rid of the cumulative sums that we used to perform this calculation:
>>> map fst (takeWhile (\x -> snd x < 30) (zip list (sums list)))
[1,2,3,4,5,6,7]
Note that because of laziness, this is as efficient as the recursive solutions -- only the sums up to the point where they fail the test need to be calculated. This can be seen because the solution works on infinite lists (because if we needed to calculate all the sums, we would never finish).
I'd probably abstract this and take the limit as a parameter:
>>> :{
... let initial lim list =
... map fst (takeWhile (\x -> snd x < lim) (zip list (sums list)))
... :}
This function has an obvious property it should satisfy, namely that the sum of a list should always be less than the limit (as long as the limit is greater than 0). So we can use QuickCheck to make sure we did it right:
>>> import Test.QuickCheck
>>> quickCheck (\lim list -> lim > 0 ==> sum (initial lim list) < lim)
+++ OK, passed 100 tests.
someList = makeList myList [] 0 where
makeList (x:xs) ys total = let newTot = total + x
in if newTot >= 25
then ys
else makeList xs (ys ++ [x]) newTot
This takes elements from myList as long as their sum is less than 25.
The logic takes place in makeList. It takes the first element of the input list and adds it to the running total, to see if it's greater than 25. If it is, we shouldn't add it to the output list, and we finish recursing. Otherwise, we put x on the end of the output list (ys) and keep going with the rest of the input list.
The behaviour you want is
ghci> appendWhileUnder 25 [1..5] [1..5]
[1,2,3,4,5,1,2,3]
because that sums to 21 and adding the 4 would bring it to 25.
OK, one way to go about this is by just appending them with ++ then taking the initial segment that's under 25.
appendWhileUnder n xs ys = takeWhileUnder n (xs++ys)
I don't want to keep summing intermediate lists, so I'll keep track with how much I'm allowed (n).
takeWhileUnder n [] = []
takeWhileUnder n (x:xs) | x < n = x:takeWhileUnder (n-x) xs
| otherwise = []
Here I allow x through if it doesn't take me beyond what's left of my allowance.
Possibly undesired side effect: it'll chop out bits of the original list if it sums to over 25. Workaround: use
appendWhileUnder' n xs ys = xs ++ takeWhileUnder (n - sum xs)
which keeps the entire xs whether it brings you over n or not.

OCaml code that works on 2 lists. Is there a better way of doing this

I have to iterate over 2 lists. One starts off as a list of empty sublists and the second one has the max length for each of the sublists that are in the first one.
Example; list1 = [[];[];[];]; list2 = [1;2;3]
I need to fill out the empty sublists in list1 ensuring that the length of the sublists never exceed the corresponding integer in list2. To that end, I wrote the following function, that given an element, elem and 2 two lists list and list, will fill out the sublists.
let mapfn elem list1 list2=
let d = ref 1 in
List.map2 (fun a b -> if ((List.length a) < b) && (!d=1)
then (incr d ; List.append a [elem])
else a )
list1 list2
;;
I can now call this function repeatedly on the elements of a list and get the final answer I need
This function works as expected. But I am little bothered by the need to use the int ref d.
Is there a better way for me to do this.
I always find it worthwhile to split the problem into byte-sized pieces that can be composed together to form a solution. You want to pad or truncate lists to a given length; this is easy to do in two steps, first pad, then truncate:
let all x = let rec xs = x :: xs in xs
let rec take n = function
| [] -> []
| _ when n = 0 -> []
| x :: xs -> x :: take (pred n) xs
all creates an infinite list by repeating a value, while take extracts the prefix sublist of at most the given length. With these two, padding and truncating is very straightforwad:
let pad_trim e n l = take n (l # all e)
(it might be a bit surprising that this actually works in a strict language like OCaml). With that defined, your required function is simply:
let mapfn elem list1 list2 = List.map2 (pad_trim elem) list2 list1
that is, taking the second list as a list of specified lengths, pad each of the lists in the first list to that length with the supplied padding element. For instance, mapfn 42 [[];[];[]] [1;2;3] gives [[42]; [42; 42]; [42; 42; 42]]. If this is not what you need, you can tweak the parts and their assembly to suit your requirements.
Are you looking for something like that?
let fill_list elem lengths =
let rec fill acc = function
| 0 -> acc
| n -> fill (elem :: acc) (n - 1) in
let accumulators = List.map (fun _ -> []) lengths in
List.map2 fill accumulators lengths
(* toplevel test *)
# let test = fill_list 42 [1; 3];;
val test : int list list = [[42]; [42; 42; 42]]
(I couldn't make sense of the first list of empty lists in your question, but I suspect it may be the accumulators for the tail-rec fill function.)

Split list into two equal lists in F#

I'm really new to F#, and I need a bit of help with an F# problem.
I need to implement a cut function that splits a list in half so that the output would be...
cut [1;2;3;4;5;6];;
val it : int list * int list = ([1; 2; 3], [4; 5; 6])
I can assume that the length of the list is even.
I'm also expected to define an auxiliary function gencut(n, xs) that cuts xs into two pieces, where n gives the size of the first piece:
gencut(2, [1;3;4;2;7;0;9]);;
val it : int list * int list = ([1; 3], [4; 2; 7; 0; 9])
I wouldn't normally ask for exercise help here, but I'm really at a loss as to where to even start. Any help, even if it's just a nudge in the right direction, would help.
Thanks!
Since your list has an even length, and you're cutting it cleanly in half, I recommend the following (psuedocode first):
Start with two pointers: slow and fast.
slow steps through the list one element at a time, fast steps two elements at a time.
slow adds each element to an accumulator variable, while fast moves foward.
When the fast pointer reaches the end of the list, the slow pointer will have only stepped half the number of elements, so its in the middle of the array.
Return the elements slow stepped over + the elements remaining. This should be two lists cut neatly in half.
The process above requires one traversal over the list and runs in O(n) time.
Since this is homework, I won't give a complete answer, but just to get you partway started, here's what it takes to cut the list cleanly in half:
let cut l =
let rec cut = function
| xs, ([] | [_]) -> xs
| [], _ -> []
| x::xs, y::y'::ys -> cut (xs, ys)
cut (l, l)
Note x::xs steps 1 element, y::y'::ys steps two.
This function returns the second half of the list. It is very easy to modify it so it returns the first half of the list as well.
You are looking for list slicing in F#. There was a great answer by #Juliet in this SO Thread: Slice like functionality from a List in F#
Basically it comes down to - this is not built in since there is no constant time index access in F# lists, but you can work around this as detailed. Her approach applied to your problem would yield a (not so efficient but working) solution:
let gencut(n, list) =
let firstList = list |> Seq.take n |> Seq.toList
let secondList = list |> Seq.skip n |> Seq.toList
(firstList, secondList)
(I didn't like my previous answer so I deleted it)
The first place to start when attacking list problems is to look at the List module which is filled with higher order functions which generalize many common problems and can give you succinct solutions. If you can't find anything suitable there, then you can look at the Seq module for solutions like #BrokenGlass demonstrated (but you can run into performance issues there). Next you'll want to consider recursion and pattern matching. There are two kinds of recursion you'll have to consider when processing lists: tail and non-tail. There are trade-offs. Tail-recursive solutions involve using an accumulator to pass state around, allowing you to place the recursive call in the tail position and avoid stack-overflows with large lists. But then you'll typically end up with a reversed list! For example,
Tail-recursive gencut solution:
let gencutTailRecursive n input =
let rec gencut cur acc = function
| hd::tl when cur < n ->
gencut (cur+1) (hd::acc) tl
| rest -> (List.rev acc), rest //need to reverse accumulator!
gencut 0 [] input
Non-tail-recursive gencut solution:
let gencutNonTailRecursive n input =
let rec gencut cur = function
| hd::tl when cur < n ->
let x, y = gencut (cur+1) tl //stackoverflow with big lists!
hd::x, y
| rest -> [], rest
gencut 0 input
Once you have your gencut solution, it's really easy to define cut:
let cut input = gencut ((List.length input)/2) input
Here's yet another way to do it using inbuilt library functions, which may or may not be easier to understand than some of the other answers. This solution also only requires one traversal across the input. My first thought after I looked at your problem was that you want something along the lines of List.partition, which splits a list into two lists based on a given predicate. However, in your case this predicate would be based on the index of the current element, which partition cannot handle, short of looking up the index for each element.
We can accomplish creating our own equivalent of this behavior using a fold or foldBack. I will use foldBack here as it means you won't have to reverse the lists afterward (see Stephens excellent answer). What we are going to do here is use the fold to provide our own index, along with the two output lists, all as the accumulator. Here is the generic function that will split your list into two lists based on n index:
let gencut n input =
//calculate the length of the list first so we can work out the index
let inputLength = input |> List.length
let results =
List.foldBack( fun elem acc->
let a,b,index = acc //decompose accumulator
if (inputLength - index) <= n then (elem::a,b,index+1)
else (a,elem::b,index+1) ) input ([],[],0)
let a,b,c = results
(a,b) //dump the index, leaving the two lists as output.
So here you see we start the foldBack with an initial accumulator value of ([],[],0). However, because we are starting at the end of the list, the 0 representing the current index needs to be subtracted from the total length of the list to get the actual index of the current element.
Then we simply check if the current index falls within the range of n. If it does, we update the accumulator by adding the current element to list a, leave list b alone, and increase the index by 1 : (elem::a,b,index+1). In all other cases, we do exactly the same but add the element to list b instead: (a,elem::b,index+1).
Now you can easily create your function that splits a list in half by creating another function over this one like so:
let cut input =
let half = (input |> List.length) / 2
input |> gencut half
I hope that can help you somewhat!
> cut data;;
val it : int list * int list = ([1; 2; 3], [4; 5; 6])
> gencut 5 data;;
val it : int list * int list = ([1; 2; 3; 4; 5], [6])
EDIT: you could avoid the index negation by supplying the length as the initial accumulator value and negating it on each cycle instead of increasing it - probably simpler that way :)
let gencut n input =
let results =
List.foldBack( fun elem acc->
let a,b,index = acc //decompose accumulator
if index <= n then (elem::a,b,index-1)
else (a,elem::b,index-1) ) input ([],[],List.length input)
let a,b,c = results
(a,b) //dump the index, leaving the two lists as output.
I have the same Homework, this was my solution. I'm just a student and new in F#
let rec gencut(n, listb) =
let rec cut n (lista : int list) (listb : int list) =
match (n , listb ) with
| 0, _ -> lista, listb
| _, [] -> lista, listb
| _, b :: listb -> cut (n - 1) (List.rev (b :: lista )) listb
cut n [] listb
let cut xs = gencut((List.length xs) / 2, xs)
Probably is not the best recursive solution, but it works. I think
You can use List.nth for random access and list comprehensions to generate a helper function:
let Sublist x y data = [ for z in x..(y - 1) -> List.nth data z ]
This will return items [x..y] from data. Using this you can easily generate gencut and cut functions (remember to check bounds on x and y) :)
check this one out:
let gencut s xs =
([for i in 0 .. s - 1 -> List.nth xs i], [for i in s .. (List.length xs) - 1 -> List.nth xs i])
the you just call
let cut xs =
gencut ((List.length xs) / 2) xs
with n durationn only one iteration split in two