I coded a function to enumerate all permutations of a given list. What do you think of the code below?
def interleave(x:Int, l:List[Int]):List[List[Int]] = {
l match {
case Nil => List(List(x))
case (head::tail) =>
(x :: head :: tail) :: interleave(x, tail).map(head :: _)
}
}
def permutations(l:List[Int]):List[List[Int]] = {
l match {
case Nil => List(List())
case (head::tail) =>
for(p0 <- permutations(tail); p1 <- interleave(head, p0)) yield p1
}
}
Given a Seq, one can already have permutations by invoking the permutations method.
scala> List(1,2,3).permutations.mkString("\n")
res3: String =
List(1, 2, 3)
List(1, 3, 2)
List(2, 1, 3)
List(2, 3, 1)
List(3, 1, 2)
List(3, 2, 1)
Furthermore there is also a method for combinations:
scala> List(1,2,3).combinations(2).mkString("\n")
res4: String =
List(1, 2)
List(1, 3)
List(2, 3)
Regarding your implementation I would say three things:
(1) Its good looking
(2) Provide an iterator (which is the std collections approach that allows to discard elements). Otherwise, you can get lists with 1000! elements which may not fit in memory.
scala> val longList = List((1 to 1000):_*)
longList: List[Int] = List(1, 2, 3,...
scala> permutations(longList)
java.lang.OutOfMemoryError: Java heap space
at scala.collection.immutable.List.$colon$colon(List.scala:67)
at .interleave(<console>:11)
at .interleave(<console>:11)
at .interleave(<console>:11)
(3) You should remove duplicated permutations (as observed by Luigi), since :
scala> permutations(List(1,1,3))
res4: List[List[Int]] = List(List(1, 1, 3), List(1, 1, 3), List(1, 3, 1), List(1, 3, 1), List(3, 1, 1), List(3, 1, 1))
scala> List(1,1,3).permutations.toList
res5: List[List[Int]] = List(List(1, 1, 3), List(1, 3, 1), List(3, 1, 1))
Consider the difference here: your version
scala> permutations(List(1,1,2)) foreach println
List(1, 1, 2)
List(1, 1, 2)
List(1, 2, 1)
List(1, 2, 1)
List(2, 1, 1)
List(2, 1, 1)
The reference version:
scala> List(1,1,2).permutations foreach println
List(1, 1, 2)
List(1, 2, 1)
List(2, 1, 1)
Maybe this thread is already well-saturated, but I thought I'd throw my solution into the mix:
Assuming no repeat elements:
def permList(l: List[Int]): List[List[Int]] = l match {
case List(ele) => List(List(ele))
case list =>
for {
i <- List.range(0, list.length)
p <- permList(list.slice(0, i) ++ list.slice(i + 1, list.length))
} yield list(i) :: p
}
With repeat elements, preventing duplicates (not as pretty):
def permList(l: List[Int]): List[List[Int]] = l match {
case List(ele) => List(List(ele))
case list =>
for {
i <- List.range(0, list.length)
val traversedList = list.slice(0, i)
val nextEle = list(i)
if !(traversedList contains nextEle)
p <- permList(traversedList ++ list.slice(i + 1, list.length))
} yield list(i) :: p
}
It's potentially not the most "list-y", given that it uses slice and an index on the list, but it's rather concise and a slightly different way of looking at it. It works by singling out each element in the list and computing the permutations of what's remaining, and then concatenating the single element to each of those permutations. If there's a more idiomatic way to do this, I'd love to hear about it.
I think such a function already exists in the standard library: Seq.permutations. So why reinventing the wheel again?
Here is a version based on span.
def perms[T](xs: List[T]): List[List[T]] = xs match {
case List(_) => List(xs)
case _ => for ( x <- xs
; val (l, r) = xs span { x!= }
; ys <- perms(l ++ r.tail)
) yield x :: ys
}
I got the following approach from SICP. It takes me fare amount of time to understand it. But it's worth and beautiful to see. How recursion works behind the scenes.
def permutations(list: List[Int]): List[List[Int]] = list match {
case Nil => Nil
case List(x) => List(List(x))
case _ => list
.flatMap(x =>
permutations(list.filterNot(_==x))
.map(p => x :: p))
}
The above solution can be translated in for loop as follows:
def perms(list: List[Int]): List[List[Int]] = {
if (list.size == 1) List(list)
else for {
x <- list
y <- perms(list.filterNot(_ == x))
} yield x :: y
}
I guess you are practicing your Scala programming skill. Here is another one, where the idea is to take different elements as head of sequence and repeats are removed through filter. Complexity of the code shall be fine, since O(n)+O(n or maybe n^2)+O(n)*P(n-1) is dominated by O(n)*P(n-1), where P(n) is the permutation number and cannot be improved, .
def permute(xs:List[Int]):List[List[Int]] = xs match {
case Nil => List(List())
case head::tail => {
val len = xs.length
val tps = (0 to len-1).map(xs.splitAt(_)).toList.filter(tp => !tp._1.contains(tp._2.head))
tps.map(tp => permute(tp._1:::tp._2.tail).map(tp._2.head :: _)).flatten
}
}
I think that mine solution is better than others
def withReplacements(chars: String, n: Int) {
def internal(path: String, acc: List[String]): List[String] = {
if (path.length == n) path :: acc else
chars.toList.flatMap {c => internal(path + c, acc)}
}
val res = internal("", Nil)
println("there are " + res.length + " " + n + "-permutations with replacement for " + chars + " = " + res)
} //> withReplacements: (chars: String, n: Int)Unit
def noReplacements(chars: String, n: Int) {
//val set = chars.groupBy(c => c).map {case (c, list) => (c -> list.length)}.toList
import scala.collection.immutable.Queue
type Set = Queue[Char]
val set = Queue[Char](chars.toList: _*)
type Result = Queue[String]
// The idea is that recursions will scan the set with one element excluded.
// Queue was chosen to implement the set to enable excluded element to bubble through it.
def internal(set: Set, path: String, acc: Result): Result = {
if (path.length == n) acc.enqueue(path)
else
set.foldLeft(acc, set.dequeue){case ((acc, (consumed_el, q)), e) =>
(internal(q, consumed_el + path, acc), q.enqueue(consumed_el).dequeue)
}. _1
}
val res = internal(set, "", Queue.empty)
println("there are " + res.length + " " + n + "-permutations without replacement for " + set + " = " + res)
} //> noReplacements: (chars: String, n: Int)Unit
withReplacements("abc", 2) //> there are 9 2-permutations with replacement for abc = List(aa, ab, ac, ba,
//| bb, bc, ca, cb, cc)
noReplacements("abc", 2) //> there are 6 2-permutations without replacement for Queue(a, b, c) = Queue(b
//| a, ca, cb, ab, ac, bc)
noReplacements("abc", 3) //> there are 6 3-permutations without replacement for Queue(a, b, c) = Queue(c
//| ba, bca, acb, cab, bac, abc)
withReplacements("abc", 3) //> there are 27 3-permutations with replacement for abc = List(aaa, aab, aac,
//| aba, abb, abc, aca, acb, acc, baa, bab, bac, bba, bbb, bbc, bca, bcb, bcc,
//| caa, cab, cac, cba, cbb, cbc, cca, ccb, ccc)
// you can run with replacements (3 chars, n = 4) but noReplacements will fail for obvious reason -- you cannont combine 3 chars to produce 4
withReplacements("abc", 4) //> there are 81 4-permutations with replacement for abc = List(aaaa, aaab, aaa
//| c, aaba, aabb, aabc, aaca, aacb, aacc, abaa, abab, abac, abba, abbb, abbc,
//| abca, abcb, abcc, acaa, acab, acac, acba, acbb, acbc, acca, accb, accc, baa
//| a, baab, baac, baba, babb, babc, baca, bacb, bacc, bbaa, bbab, bbac, bbba,
//| bbbb, bbbc, bbca, bbcb, bbcc, bcaa, bcab, bcac, bcba, bcbb, bcbc, bcca, bcc
//| b, bccc, caaa, caab, caac, caba, cabb, cabc, caca, cacb, cacc, cbaa, cbab,
//| cbac, cbba, cbbb, cbbc, cbca, cbcb, cbcc, ccaa, ccab, ccac, ccba, ccbb, ccb
//| c, ccca, cccb, cccc)
(1 to 3) foreach (u => noReplacements("aab", u))//> there are 3 1-permutations without replacement for Queue(a, a, b) = Queue(a
//| , a, b)
//| there are 6 2-permutations without replacement for Queue(a, a, b) = Queue(a
//| a, ba, ba, aa, ab, ab)
//| there are 6 3-permutations without replacement for Queue(a, a, b) = Queue(b
//| aa, aba, aba, baa, aab, aab)
These are the same 3 lines of code but variable permutation lengths are supported and list concatenations are eliminated.
I have made the second more ideomatic (so that flat-map merges of accumulator are prevented, which also makes it more tail-recursive) and extended into multiset permutations, so that you can say that "aab", "aba" and "baa" are permutations (of each other). The idea is that letter "a" is replacable two times instead infinitly (with replacement case) or availble only one time (without replacements). So, you need a counter, which tells you how many times every letter is avaiable for replacement.
// Rewrite with replacement a bit to eliminate flat-map merges.
def norep2(chars: String, n: Int/* = chars.length*/) {
import scala.collection.immutable.Queue
type Set = Queue[Char]
val set = Queue[Char](chars.toList: _*)
type Result = Queue[String]
def siblings(set: (Char, Set), offset: Int, path: String, acc: Result): Result = set match {case (bubble, queue) =>
val children = descend(queue, path + bubble, acc) // bubble was used, it is not available for children that will produce combinations in other positions
if (offset == 0) children else siblings(queue.enqueue(bubble).dequeue, offset - 1, path, children) // siblings will produce different chars at the same position, fetch next char for them
}
def descend(set: Set, path: String, acc: Result): Result = {
if (path.length == n) acc.enqueue(path) else siblings(set.dequeue, set.size-1, path, acc)
}
val res = descend(set, "", Queue.empty)
println("there are " + res.length + " " + n + "-permutations without replacement for " + set + " = " + res)
} //> norep2: (chars: String, n: Int)Unit
assert(norep2("abc", 2) == noReplacements("abc", 2))
//> there are 6 2-permutations without replacement for Queue(a, b, c) = Queue(a
//| b, ac, bc, ba, ca, cb)
//| there are 6 2-permutations without replacement for Queue(a, b, c) = Queue(b
//| a, ca, cb, ab, ac, bc)
assert(norep2("abc", 3) == noReplacements("abc", 3))
//> there are 6 3-permutations without replacement for Queue(a, b, c) = Queue(a
//| bc, acb, bca, bac, cab, cba)
//| there are 6 3-permutations without replacement for Queue(a, b, c) = Queue(c
//| ba, bca, acb, cab, bac, abc)
def multisets(chars: String, n: Int/* = chars.length*/) {
import scala.collection.immutable.Queue
type Set = Queue[Bubble]
type Bubble = (Char, Int)
type Result = Queue[String]
def siblings(set: (Bubble, Set), offset: Int, path: String, acc: Result): Result = set match {case ((char, avail), queue) =>
val children = descend(if (avail - 1 == 0) queue else queue.enqueue(char -> {avail-1}), path + char, acc) // childern can reuse the symbol while if it is available
if (offset == 0) children else siblings(queue.enqueue((char, avail)).dequeue, offset - 1, path, children)
}
def descend(set: Set, path: String, acc: Result): Result = {
if (path.length == n) acc.enqueue(path) else siblings(set.dequeue, set.size-1, path, acc)
}
val set = Queue[Bubble]((chars.toList groupBy (c => c) map {case (k, v) => (k, v.length)}).toList: _*)
val res = descend(set, "", Queue.empty)
println("there are " + res.length + " multiset " + n + "-permutations for " + set + " = " + res)
} //> multisets: (chars: String, n: Int)Unit
assert(multisets("abc", 2) == norep2("abc", 2)) //> there are 6 multiset 2-permutations for Queue((b,1), (a,1), (c,1)) = Queue(
//| ba, bc, ac, ab, cb, ca)
//| there are 6 2-permutations without replacement for Queue(a, b, c) = Queue(a
//| b, ac, bc, ba, ca, cb)
assert(multisets("abc", 3) == norep2("abc", 3)) //> there are 6 multiset 3-permutations for Queue((b,1), (a,1), (c,1)) = Queue(
//| bac, bca, acb, abc, cba, cab)
//| there are 6 3-permutations without replacement for Queue(a, b, c) = Queue(a
//| bc, acb, bca, bac, cab, cba)
assert (multisets("aaab", 2) == multisets2("aaab".toList, 2) )
//> there are 3 multiset 2-permutations for Queue((b,1), (a,3)) = Queue(ba, ab,
//| aa)
//| there are 3 multiset 2-permutations for Queue((b,1), (a,3)) = List(List(a,
//| a), List(b, a), List(a, b))
multisets("aab", 2) //> there are 3 multiset 2-permutations for Queue((b,1), (a,2)) = Queue(ba, ab,
//| aa)
multisets("aab", 3) //> there are 3 multiset 3-permutations for Queue((b,1), (a,2)) = Queue(baa, ab
//| a, aab)
norep2("aab", 3) //> there are 6 3-permutations without replacement for Queue(a, a, b) = Queue(a
//| ab, aba, aba, aab, baa, baa)
As generalizaiton, you can obtain with/without replacements using multisets funciton. For instance,
//take far more letters than resulting permutation length to emulate withReplacements
assert(multisets("aaaaabbbbbccccc", 3) == withReplacements("abc", 3))
//> there are 27 multiset 3-permutations for Queue((b,5), (a,5), (c,5)) = Queue
//| (bac, bab, baa, bcb, bca, bcc, bba, bbc, bbb, acb, aca, acc, aba, abc, abb,
//| aac, aab, aaa, cba, cbc, cbb, cac, cab, caa, ccb, cca, ccc)
//| there are 27 3-permutations with replacement for abc = List(aaa, aab, aac,
//| aba, abb, abc, aca, acb, acc, baa, bab, bac, bba, bbb, bbc, bca, bcb, bcc,
//| caa, cab, cac, cba, cbb, cbc, cca, ccb, ccc)
//take one letter of each to emulate withoutReplacements
assert(multisets("aaaaabbbbbccccc", 3) == noReplacements("abc", 3))
//> there are 27 multiset 3-permutations for Queue((b,5), (a,5), (c,5)) = Queue
//| (bac, bab, baa, bcb, bca, bcc, bba, bbc, bbb, acb, aca, acc, aba, abc, abb,
//| aac, aab, aaa, cba, cbc, cbb, cac, cab, caa, ccb, cca, ccc)
//| there are 6 3-permutations without replacement for Queue(a, b, c) = Queue(c
//| ba, bca, acb, cab, bac, abc)
If you are more interested about permutations, you may look at
def permutator[T](list: List[T]): List[List[T]] = {
def _p(total: Int, list: List[T]): List[List[T]] = {
if (total == 0) {
// End of the recursion tree
list.map(List(_))
} else {
// Controlled combinatorial
// explosion while total > 0
for (i <- list;
j <- _p(total - 1, list))
yield { i :: j }
// It is a recursion tree to generate the
// permutations of the elements
// --------------------------------------
// total = 0 => _p returns 3 elements (A, B, C)
// total = 1 => _p returns 3 * 3 List(List(A, A)...
// total = 2 => _p returns 3 * 3 * 3 elements List(List(A, A, A)...
}
}
_p(list.length - 1, list)
}
permutator(List("A", "B", "C"))
// output:
List(A, A, A),List(A, A, B),List(A, A, C),List(A, B, A),List(A, B, B),
List(A, B, C),List(A, C, A),List(A, C, B),List(A, C, C),List(B, A, A),
List(B, A, B),List(B, A, C),List(B, B, A),List(B, B, B),List(B, B, C),
List(B, C, A),List(B, C, B),List(B, C, C),List(C, A, A),List(C, A, B),
List(C, A, C),List(C, B, A),List(C, B, B),List(C, B, C),List(C, C, A),
List(C, C, B),List(C, C, C)
def perms(in:List[Int]):List[List[Int]] = {
def perms0(in: List[Int], tmp: List[Int]): List[List[Int]] =
if (in.isEmpty) List(tmp)
else in.foldLeft(Nil: List[List[Int]])((acc, el) => perms0(in.filter(en => en != el) , el :: tmp) ++ acc)
perms0(in, Nil)
}
I inspired solution by the idea shown on the picture below showing building graph of partial permutations starting from empty (graph root) toward all permutations (graph leaves): https://medium.com/algorithms-and-leetcode/backtracking-e001561b9f28
This is an implementation which is based on the concept of cycle and a trivial implementation of permute with two elements. It does not handle duplicates and stack overflow aspect in the permute method
object ImmuPermute extends App {
def nextCycle(nums: List[Int]): List[Int] = {
nums match {
case Nil => Nil
case head :: tail => tail :+ head
}
}
def cycles(nums: List[Int]): List[List[Int]] = {
def loop(l: List[Int], acc: List[List[Int]]): List[List[Int]] = {
if (acc.size == nums.size)
acc
else {
val next = nextCycle(l)
loop(next, next :: acc)
}
}
loop(nums, List(nums))
}
def permute(nums: List[Int]): List[List[Int]] = {
nums match {
case Nil => Nil
case head :: Nil => List(List(head))
case first :: second :: Nil => List(List(first, second), List(second, first))
case _ => {
val cycledList = cycles(nums)
cycledList.flatMap { cycle =>
val h = cycle.head
val t = cycle.tail
val permutedList = permute(t)
permutedList map { pList =>
h :: pList
}
}
}
}
}
val l = permute(List(1, 2, 3, 4))
l foreach println
println(l.size)
}
One liner:
List(1,2,3).permutations.toList
res1: List[List[Int]] = List(List(1, 2, 3), List(1, 3, 2), List(2, 1, 3), List(2, 3, 1), List(3, 1, 2), List(3, 2, 1))
