Say I have a list of n elements, I know there are n! possible ways to order these elements. What is an algorithm to generate all possible orderings of this list? Example, I have list [a, b, c]. The algorithm would return [[a, b, c], [a, c, b,], [b, a, c], [b, c, a], [c, a, b], [c, b, a]].
I'm reading this here http://en.wikipedia.org/wiki/Permutation#Algorithms_to_generate_permutations
But Wikipedia has never been good at explaining. I don't understand much of it.
Basically, for each item from left to right, all the permutations of the remaining items are generated (and each one is added with the current elements). This can be done recursively (or iteratively if you like pain) until the last item is reached at which point there is only one possible order.
So with the list [1,2,3,4] all the permutations that start with 1 are generated, then all the permutations that start with 2, then 3 then 4.
This effectively reduces the problem from one of finding permutations of a list of four items to a list of three items. After reducing to 2 and then 1 item lists, all of them will be found.
Example showing process permutations using 3 coloured balls:
(from https://en.wikipedia.org/wiki/Permutation#/media/File:Permutations_RGB.svg - https://commons.wikimedia.org/wiki/File:Permutations_RGB.svg)
Here is an algorithm in Python that works by in place on an array:
def permute(xs, low=0):
if low + 1 >= len(xs):
yield xs
else:
for p in permute(xs, low + 1):
yield p
for i in range(low + 1, len(xs)):
xs[low], xs[i] = xs[i], xs[low]
for p in permute(xs, low + 1):
yield p
xs[low], xs[i] = xs[i], xs[low]
for p in permute([1, 2, 3, 4]):
print p
You can try the code out for yourself here: http://repl.it/J9v
There is already plenty of good solutions here, but I would like to share how I solved this problem on my own and hope that this might be helpful for somebody who would also like to derive his own solution.
After some pondering about the problem I have come up with two following conclusions:
L of size n there will be equal number of solutions starting with L1, L2 ... Ln elements of the list. Since in total there are n! permutations of the list of size n, we get n! / n = (n-1)! permutations in each group.[a,b] and [b,a]. Using these two simple ideas I have derived the following algorithm:
permute array
if array is of size 2
return first and second element as new array
return second and first element as new array
else
for each element in array
new subarray = array with excluded element
return element + permute subarray
Here is how I implemented this in C#:
public IEnumerable<List<T>> Permutate<T>(List<T> input)
{
if (input.Count == 2) // this are permutations of array of size 2
{
yield return new List<T>(input);
yield return new List<T> {input[1], input[0]};
}
else
{
foreach(T elem in input) // going through array
{
var rlist = new List<T>(input); // creating subarray = array
rlist.Remove(elem); // removing element
foreach(List<T> retlist in Permutate(rlist))
{
retlist.Insert(0,elem); // inserting the element at pos 0
yield return retlist;
}
}
}
}
Wikipedia's answer for "lexicographic order" seems perfectly explicit in cookbook style to me. It cites a 14th century origin for the algorithm!
I've just written a quick implementation in Java of Wikipedia's algorithm as a check and it was no trouble. But what you have in your Q as an example is NOT "list all permutations", but "a LIST of all permutations", so wikipedia won't be a lot of help to you. You need a language in which lists of permutations are feasibly constructed. And believe me, lists a few billion long are not usually handled in imperative languages. You really want a non-strict functional programming language, in which lists are a first-class object, to get out stuff while not bringing the machine close to heat death of the Universe.
That's easy. In standard Haskell or any modern FP language:
-- perms of a list
perms :: [a] -> [ [a] ]
perms (a:as) = [bs ++ a:cs | perm <- perms as, (bs,cs) <- splits perm]
perms [] = [ [] ]
and
-- ways of splitting a list into two parts
splits :: [a] -> [ ([a],[a]) ]
splits [] = [ ([],[]) ]
splits (a:as) = ([],a:as) : [(a:bs,cs) | (bs,cs) <- splits as]
As WhirlWind said, you start at the beginning.
You swap cursor with each remaining value, including cursor itself, these are all new instances (I used an int[] and array.clone() in the example).
Then perform permutations on all these different lists, making sure the cursor is one to the right.
When there are no more remaining values (cursor is at the end), print the list. This is the stop condition.
public void permutate(int[] list, int pointer) {
if (pointer == list.length) {
//stop-condition: print or process number
return;
}
for (int i = pointer; i < list.length; i++) {
int[] permutation = (int[])list.clone();.
permutation[pointer] = list[i];
permutation[i] = list[pointer];
permutate(permutation, pointer + 1);
}
}
Recursive always takes some mental effort to maintain. And for big numbers, factorial is easily huge and stack overflow will easily be a problem.
For small numbers (3 or 4, which is mostly encountered), multiple loops are quite simple and straight forward. It is unfortunate answers with loops didn't get voted up.
Let's start with enumeration (rather than permutation). Simply read the code as pseudo perl code.
$foreach $i1 in @list
$foreach $i2 in @list
$foreach $i3 in @list
print "$i1, $i2, $i3\n"
Enumeration is more often encountered than permutation, but if permutation is needed, just add the conditions:
$foreach $i1 in @list
$foreach $i2 in @list
$if $i2==$i1
next
$foreach $i3 in @list
$if $i3==$i1 or $i3==$i2
next
print "$i1, $i2, $i3\n"
Now if you really need general method potentially for big lists, we can use radix method. First, consider the enumeration problem:
$n=@list
my @radix
$for $i=0:$n
$radix[$i]=0
$while 1
my @temp
$for $i=0:$n
push @temp, $list[$radix[$i]]
print join(", ", @temp), "\n"
$call radix_increment
subcode: radix_increment
$i=0
$while 1
$radix[$i]++
$if $radix[$i]==$n
$radix[$i]=0
$i++
$else
last
$if $i>=$n
last
Radix increment is essentially number counting (in the base of number of list elements).
Now if you need permutaion, just add the checks inside the loop:
subcode: check_permutation
my @check
my $flag_dup=0
$for $i=0:$n
$check[$radix[$i]]++
$if $check[$radix[$i]]>1
$flag_dup=1
last
$if $flag_dup
next
Edit: The above code should work, but for permutation, radix_increment could be wasteful. So if time is a practical concern, we have to change radix_increment into permute_inc:
subcode: permute_init
$for $i=0:$n
$radix[$i]=$i
subcode: permute_inc
$max=-1
$for $i=$n:0
$if $max<$radix[$i]
$max=$radix[$i]
$else
$for $j=$n:0
$if $radix[$j]>$radix[$i]
$call swap, $radix[$i], $radix[$j]
break
$j=$i+1
$k=$n-1
$while $j<$k
$call swap, $radix[$j], $radix[$k]
$j++
$k--
break
$if $i<0
break
Of course now this code is logically more complex, I'll leave for reader's exercise.
If anyone wonders how to be done in permutation in javascript.
Idea/pseudocode
for example. 'a'+ permute(bc). permute of bc would be bc & cb. Now add these two will give abc, acb. similarly, pick b + permute (ac) will provice bac, bca...and keep going.
now look at the code
function permutations(arr){
var len = arr.length,
perms = [],
rest,
picked,
restPerms,
next;
//for one or less item there is only one permutation
if (len <= 1)
return [arr];
for (var i=0; i<len; i++)
{
//copy original array to avoid changing it while picking elements
rest = Object.create(arr);
//splice removed element change array original array(copied array)
//[1,2,3,4].splice(2,1) will return [3] and remaining array = [1,2,4]
picked = rest.splice(i, 1);
//get the permutation of the rest of the elements
restPerms = permutations(rest);
// Now concat like a+permute(bc) for each
for (var j=0; j<restPerms.length; j++)
{
next = picked.concat(restPerms[j]);
perms.push(next);
}
}
return perms;
}
Take your time to understand this. I got this code from (pertumation in JavaScript)
I have written this recursive solution in ANSI C. Each execution of the Permutate function provides one different permutation until all are completed. Global variables can also be used for variables fact and count.
#include <stdio.h>
#define SIZE 4
void Rotate(int vec[], int size)
{
int i, j, first;
first = vec[0];
for(j = 0, i = 1; i < size; i++, j++)
{
vec[j] = vec[i];
}
vec[j] = first;
}
int Permutate(int *start, int size, int *count)
{
static int fact;
if(size > 1)
{
if(Permutate(start + 1, size - 1, count))
{
Rotate(start, size);
}
fact *= size;
}
else
{
(*count)++;
fact = 1;
}
return !(*count % fact);
}
void Show(int vec[], int size)
{
int i;
printf("%d", vec[0]);
for(i = 1; i < size; i++)
{
printf(" %d", vec[i]);
}
putchar('\n');
}
int main()
{
int vec[] = { 1, 2, 3, 4, 5, 6 }; /* Only the first SIZE items will be permutated */
int count = 0;
do
{
Show(vec, SIZE);
} while(!Permutate(vec, SIZE, &count));
putchar('\n');
Show(vec, SIZE);
printf("\nCount: %d\n\n", count);
return 0;
}
Another one in Python, it's not in place as @cdiggins's, but I think it's easier to understand
def permute(num):
if len(num) == 2:
# get the permutations of the last 2 numbers by swapping them
yield num
num[0], num[1] = num[1], num[0]
yield num
else:
for i in range(0, len(num)):
# fix the first number and get the permutations of the rest of numbers
for perm in permute(num[0:i] + num[i+1:len(num)]):
yield [num[i]] + perm
for p in permute([1, 2, 3, 4]):
print p
I was thinking of writing a code for getting the permutations of any given integer of any size, i.e., providing a number 4567 we get all possible permutations till 7654...So i worked on it and found an algorithm and finally implemented it, Here is the code written in "c". You can simply copy it and run on any open source compilers. But some flaws are waiting to be debugged. Please appreciate.
Code:
#include <stdio.h>
#include <conio.h>
#include <malloc.h>
//PROTOTYPES
int fact(int); //For finding the factorial
void swap(int*,int*); //Swapping 2 given numbers
void sort(int*,int); //Sorting the list from the specified path
int imax(int*,int,int); //Finding the value of imax
int jsmall(int*,int); //Gives position of element greater than ith but smaller than rest (ahead of imax)
void perm(); //All the important tasks are done in this function
int n; //Global variable for input OR number of digits
void main()
{
int c=0;
printf("Enter the number : ");
scanf("%d",&c);
perm(c);
getch();
}
void perm(int c){
int *p; //Pointer for allocating separate memory to every single entered digit like arrays
int i, d;
int sum=0;
int j, k;
long f;
n = 0;
while(c != 0) //this one is for calculating the number of digits in the entered number
{
sum = (sum * 10) + (c % 10);
n++; //as i told at the start of loop
c = c / 10;
}
f = fact(n); //It gives the factorial value of any number
p = (int*) malloc(n*sizeof(int)); //Dynamically allocation of array of n elements
for(i=0; sum != 0 ; i++)
{
*(p+i) = sum % 10; //Giving values in dynamic array like 1234....n separately
sum = sum / 10;
}
sort(p,-1); //For sorting the dynamic array "p"
for(c=0 ; c<f/2 ; c++) { //Most important loop which prints 2 numbers per loop, so it goes upto 1/2 of fact(n)
for(k=0 ; k<n ; k++)
printf("%d",p[k]); //Loop for printing one of permutations
printf("\n");
i = d = 0;
i = imax(p,i,d); //provides the max i as per algo (i am restricted to this only)
j = i;
j = jsmall(p,j); //provides smallest i val as per algo
swap(&p[i],&p[j]);
for(k=0 ; k<n ; k++)
printf("%d",p[k]);
printf("\n");
i = d = 0;
i = imax(p,i,d);
j = i;
j = jsmall(p,j);
swap(&p[i],&p[j]);
sort(p,i);
}
free(p); //Deallocating memory
}
int fact (int a)
{
long f=1;
while(a!=0)
{
f = f*a;
a--;
}
return f;
}
void swap(int *p1,int *p2)
{
int temp;
temp = *p1;
*p1 = *p2;
*p2 = temp;
return;
}
void sort(int*p,int t)
{
int i,temp,j;
for(i=t+1 ; i<n-1 ; i++)
{
for(j=i+1 ; j<n ; j++)
{
if(*(p+i) > *(p+j))
{
temp = *(p+i);
*(p+i) = *(p+j);
*(p+j) = temp;
}
}
}
}
int imax(int *p, int i , int d)
{
while(i<n-1 && d<n-1)
{
if(*(p+d) < *(p+d+1))
{
i = d;
d++;
}
else
d++;
}
return i;
}
int jsmall(int *p, int j)
{
int i,small = 32767,k = j;
for (i=j+1 ; i<n ; i++)
{
if (p[i]<small && p[i]>p[k])
{
small = p[i];
j = i;
}
}
return j;
}
void permutate(char[] x, int i, int n){
x=x.clone();
if (i==n){
System.out.print(x);
System.out.print(" ");
counter++;}
else
{
for (int j=i; j<=n;j++){
// System.out.print(temp); System.out.print(" "); //Debugger
swap (x,i,j);
// System.out.print(temp); System.out.print(" "+"i="+i+" j="+j+"\n");// Debugger
permutate(x,i+1,n);
// swap (temp,i,j);
}
}
}
void swap (char[] x, int a, int b){
char temp = x[a];
x[a]=x[b];
x[b]=temp;
}
I created this one. based on research too permutate(qwe, 0, qwe.length-1); Just so you know, you can do it with or without backtrack
Here's a toy Ruby method that works like #permutation.to_a that might be more legible to crazy people. It's hella slow, but also 5 lines.
def permute(ary)
return [ary] if ary.size <= 1
ary.collect_concat.with_index do |e, i|
rest = ary.dup.tap {|a| a.delete_at(i) }
permute(rest).collect {|a| a.unshift(e) }
end
end
Here is the code in Python to print all possible permutations of a list:
def next_perm(arr):
# Find non-increasing suffix
i = len(arr) - 1
while i > 0 and arr[i - 1] >= arr[i]:
i -= 1
if i <= 0:
return False
# Find successor to pivot
j = len(arr) - 1
while arr[j] <= arr[i - 1]:
j -= 1
arr[i - 1], arr[j] = arr[j], arr[i - 1]
# Reverse suffix
arr[i : ] = arr[len(arr) - 1 : i - 1 : -1]
print arr
return True
def all_perm(arr):
a = next_perm(arr)
while a:
a = next_perm(arr)
arr = raw_input()
arr.split(' ')
arr = map(int, arr)
arr.sort()
print arr
all_perm(arr)
I have used a lexicographic order algorithm to get all possible permutations, but a recursive algorithm is more efficient. You can find the code for recursive algorithm here: Python recursion permutations
Java version
/**
* @param uniqueList
* @param permutationSize
* @param permutation
* @param only Only show the permutation of permutationSize,
* else show all permutation of less than or equal to permutationSize.
*/
public static void my_permutationOf(List<Integer> uniqueList, int permutationSize, List<Integer> permutation, boolean only) {
if (permutation == null) {
assert 0 < permutationSize && permutationSize <= uniqueList.size();
permutation = new ArrayList<>(permutationSize);
if (!only) {
System.out.println(Arrays.toString(permutation.toArray()));
}
}
for (int i : uniqueList) {
if (permutation.contains(i)) {
continue;
}
permutation.add(i);
if (!only) {
System.out.println(Arrays.toString(permutation.toArray()));
} else if (permutation.size() == permutationSize) {
System.out.println(Arrays.toString(permutation.toArray()));
}
if (permutation.size() < permutationSize) {
my_permutationOf(uniqueList, permutationSize, permutation, only);
}
permutation.remove(permutation.size() - 1);
}
}
E.g.
public static void main(String[] args) throws Exception {
my_permutationOf(new ArrayList<Integer>() {
{
add(1);
add(2);
add(3);
}
}, 3, null, true);
}
output:
[1, 2, 3]
[1, 3, 2]
[2, 1, 3]
[2, 3, 1]
[3, 1, 2]
[3, 2, 1]
in PHP
$set=array('A','B','C','D');
function permutate($set) {
$b=array();
foreach($set as $key=>$value) {
if(count($set)==1) {
$b[]=$set[$key];
}
else {
$subset=$set;
unset($subset[$key]);
$x=permutate($subset);
foreach($x as $key1=>$value1) {
$b[]=$value.' '.$value1;
}
}
}
return $b;
}
$x=permutate($set);
var_export($x);
this is a java version for permutation
public class Permutation {
static void permute(String str) {
permute(str.toCharArray(), 0, str.length());
}
static void permute(char [] str, int low, int high) {
if (low == high) {
System.out.println(str);
return;
}
for (int i=low; i<high; i++) {
swap(str, i, low);
permute(str, low+1, high);
swap(str, low, i);
}
}
static void swap(char [] array, int i, int j) {
char t = array[i];
array[i] = array[j];
array[j] = t;
}
}
public class PermutationGenerator
{
private LinkedList<List<int>> _permutationsList;
public void FindPermutations(List<int> list, int permutationLength)
{
_permutationsList = new LinkedList<List<int>>();
foreach(var value in list)
{
CreatePermutations(value, permutationLength);
}
}
private void CreatePermutations(int value, int permutationLength)
{
var node = _permutationsList.First;
var last = _permutationsList.Last;
while (node != null)
{
if (node.Value.Count < permutationLength)
{
GeneratePermutations(node.Value, value, permutationLength);
}
if (node == last)
{
break;
}
node = node.Next;
}
List<int> permutation = new List<int>();
permutation.Add(value);
_permutationsList.AddLast(permutation);
}
private void GeneratePermutations(List<int> permutation, int value, int permutationLength)
{
if (permutation.Count < permutationLength)
{
List<int> copyOfInitialPermutation = new List<int>(permutation);
copyOfInitialPermutation.Add(value);
_permutationsList.AddLast(copyOfInitialPermutation);
List<int> copyOfPermutation = new List<int>();
copyOfPermutation.AddRange(copyOfInitialPermutation);
int lastIndex = copyOfInitialPermutation.Count - 1;
for (int i = lastIndex;i > 0;i--)
{
int temp = copyOfPermutation[i - 1];
copyOfPermutation[i - 1] = copyOfPermutation[i];
copyOfPermutation[i] = temp;
List<int> perm = new List<int>();
perm.AddRange(copyOfPermutation);
_permutationsList.AddLast(perm);
}
}
}
public void PrintPermutations(int permutationLength)
{
int count = _permutationsList.Where(perm => perm.Count() == permutationLength).Count();
Console.WriteLine("The number of permutations is " + count);
}
}
In Scala
def permutazione(n: List[Int]): List[List[Int]] = permutationeAcc(n, Nil)
def permutationeAcc(n: List[Int], acc: List[Int]): List[List[Int]] = {
var result: List[List[Int]] = Nil
for (i ← n if (!(acc contains (i))))
if (acc.size == n.size-1)
result = (i :: acc) :: result
else
result = result ::: permutationeAcc(n, i :: acc)
result
}
Here's an implementation for ColdFusion (requires CF10 because of the merge argument to ArrayAppend() ):
public array function permutateArray(arr){
if (not isArray(arguments.arr) ) {
return ['The ARR argument passed to the permutateArray function is not of type array.'];
}
var len = arrayLen(arguments.arr);
var perms = [];
var rest = [];
var restPerms = [];
var rpLen = 0;
var next = [];
//for one or less item there is only one permutation
if (len <= 1) {
return arguments.arr;
}
for (var i=1; i <= len; i++) {
// copy the original array so as not to change it and then remove the picked (current) element
rest = arraySlice(arguments.arr, 1);
arrayDeleteAt(rest, i);
// recursively get the permutation of the rest of the elements
restPerms = permutateArray(rest);
rpLen = arrayLen(restPerms);
// Now concat each permutation to the current (picked) array, and append the concatenated array to the end result
for (var j=1; j <= rpLen; j++) {
// for each array returned, we need to make a fresh copy of the picked(current) element array so as to not change the original array
next = arraySlice(arguments.arr, i, 1);
arrayAppend(next, restPerms[j], true);
arrayAppend(perms, next);
}
}
return perms;
}
Based on KhanSharp's js solution above.
In the following Java solution we take advantage over the fact that Strings are immutable in order to avoid cloning the result-set upon every iteration.
The input will be a String, say "abc", and the output will be all the possible permutations:
abc
acb
bac
bca
cba
cab
public static void permute(String s) {
permute(s, 0);
}
private static void permute(String str, int left){
if(left == str.length()-1) {
System.out.println(str);
} else {
for(int i = left; i < str.length(); i++) {
String s = swap(str, left, i);
permute(s, left+1);
}
}
}
private static String swap(String s, int left, int right) {
if (left == right)
return s;
String result = s.substring(0, left);
result += s.substring(right, right+1);
result += s.substring(left+1, right);
result += s.substring(left, left+1);
result += s.substring(right+1);
return result;
}
Same approach can be applied to arrays (instead of a string):
public static void main(String[] args) {
int[] abc = {1,2,3};
permute(abc, 0);
}
public static void permute(int[] arr, int index) {
if (index == arr.length) {
System.out.println(Arrays.toString(arr));
} else {
for (int i = index; i < arr.length; i++) {
int[] permutation = arr.clone();
permutation[index] = arr[i];
permutation[i] = arr[index];
permute(permutation, index + 1);
}
}
}
I know this a very very old and even off-topic in today's stackoverflow but I still wanted to contribute a friendly javascript answer for the simple reason that it runs in your browser.
I've also added the debugger directive breakpoint so you can step through the code (chrome required) to see how this algorithm works. Open up your dev console in chrome (F12 in windows or CMD + OPTION + I on mac) and then click "Run code snippet". This implements the same exact algorithm that @WhirlWind presented in his answer.
Your browser should pause execution at the debugger directive. Use F8 to continue code execution.
function permute(rest, prefix = []) {
if (rest.length === 0) {
return [prefix];
}
return (rest
.map((x, index) => {
const oldRest = rest;
const oldPrefix = prefix;
// the `...` destructures the array into single values flattening it
const newRest = [...rest.slice(0, index), ...rest.slice(index + 1)];
const newPrefix = [...prefix, x];
debugger;
const result = permute(newRest, newPrefix);
return result;
})
// this step flattens the array of arrays returned by calling permute
.reduce((flattened, arr) => [...flattened, ...arr], [])
);
}
console.log(permute([1, 2, 3]));
It's my solution on Java:
public class CombinatorialUtils {
public static void main(String[] args) {
List<String> alphabet = new ArrayList<>();
alphabet.add("1");
alphabet.add("2");
alphabet.add("3");
alphabet.add("4");
for (List<String> strings : permutations(alphabet)) {
System.out.println(strings);
}
System.out.println("-----------");
for (List<String> strings : combinations(alphabet)) {
System.out.println(strings);
}
}
public static List<List<String>> combinations(List<String> alphabet) {
List<List<String>> permutations = permutations(alphabet);
List<List<String>> combinations = new ArrayList<>(permutations);
for (int i = alphabet.size(); i > 0; i--) {
final int n = i;
combinations.addAll(permutations.stream().map(strings -> strings.subList(0, n)).distinct().collect(Collectors.toList()));
}
return combinations;
}
public static <T> List<List<T>> permutations(List<T> alphabet) {
ArrayList<List<T>> permutations = new ArrayList<>();
if (alphabet.size() == 1) {
permutations.add(alphabet);
return permutations;
} else {
List<List<T>> subPerm = permutations(alphabet.subList(1, alphabet.size()));
T addedElem = alphabet.get(0);
for (int i = 0; i < alphabet.size(); i++) {
for (List<T> permutation : subPerm) {
int index = i;
permutations.add(new ArrayList<T>(permutation) {{
add(index, addedElem);
}});
}
}
}
return permutations;
}
}
You can't really talk about solving a permultation problem in recursion without posting an implementation in a (dialect of) language that pioneered the idea. So, for the sake of completeness, here is one of the ways that can be done in Scheme.
(define (permof wd)
(cond ((null? wd) '())
((null? (cdr wd)) (list wd))
(else
(let splice ([l '()] [m (car wd)] [r (cdr wd)])
(append
(map (lambda (x) (cons m x)) (permof (append l r)))
(if (null? r)
'()
(splice (cons m l) (car r) (cdr r))))))))
calling (permof (list "foo" "bar" "baz")) we'll get:
'(("foo" "bar" "baz")
("foo" "baz" "bar")
("bar" "foo" "baz")
("bar" "baz" "foo")
("baz" "bar" "foo")
("baz" "foo" "bar"))
I won't go into the algorithm details because it's been explained enough in other posts. The idea is the same.
However, recursive problems tend to be much harder to model and think about in destructive medium like Python, C, and Java, while in Lisp or ML it can be concisely expressed.
I came up with these algorithms (one recursive and the other iterative) when I was in my first year of engineering and implemented them in python (I apologize for the random naming scheme):
def permute(x):
m = [ord(i) for i in x]
m.sort()
count = 1
print(count,''.join([chr(i) for i in m]))
while (True):
l = -1
lgth = len(m)
while (m[l] <= m[l-1]):
l -= 1
if l == -lgth:
break
if l == -lgth:
break
z = sorted(m[l-1:])
k = m[l-1]
z2 = z[:]
z.reverse()
b = z[z.index(k)-1]
z2.remove(b)
v = [b] + z2
m[l-1:] = v
count += 1
print(count, ''.join([chr(i) for i in m]))
"""A Less-powerful solution: """
def permute_recursively(inp_list):
if (len(inp_list) == 1):
return [inp_list]
else:
x = permute_recursively(inp_list[1:])
a = inp_list[0]
z = []
for i in x:
for k in range(len(i) + 1):
z.append(i[:k] + [a] + i[k:])
return(z)
""" main section """
x = input("Enter the string: ")
#call to permute
permute(x)
print("****** End of permute ******\n")
#call to permute_recursively
count = 1
for i in permute_recursively(sorted(list(x))):
print(count, ''.join(i))
count += 1
Characteristics of the first algorithm:
The second algorithm is recursive:
Here is a recursive solution in PHP. WhirlWind's post accurately describes the logic. It's worth mentioning that generating all permutations runs in factorial time, so it might be a good idea to use an iterative approach instead.
public function permute($sofar, $input){
for($i=0; $i < strlen($input); $i++){
$diff = strDiff($input,$input[$i]);
$next = $sofar.$input[$i]; //next contains a permutation, save it
$this->permute($next, $diff);
}
}
The strDiff function takes two strings, s1 and s2, and returns a new string with everything in s1 without elements in s2 (duplicates matter). So, strDiff('finish','i') => 'fnish' (the second 'i' is not removed).
Here is an algorithm in R, in case anyone needs to avoid loading additional libraries like I had to.
permutations <- function(n){
if(n==1){
return(matrix(1))
} else {
sp <- permutations(n-1)
p <- nrow(sp)
A <- matrix(nrow=n*p,ncol=n)
for(i in 1:n){
A[(i-1)*p+1:p,] <- cbind(i,sp+(sp>=i))
}
return(A)
}
}
Example usage:
> matrix(letters[permutations(3)],ncol=3)
[,1] [,2] [,3]
[1,] "a" "b" "c"
[2,] "a" "c" "b"
[3,] "b" "a" "c"
[4,] "b" "c" "a"
[5,] "c" "a" "b"
[6,] "c" "b" "a"
#!/usr/bin/env python
import time
def permutations(sequence):
# print sequence
unit = [1, 2, 1, 2, 1]
if len(sequence) >= 4:
for i in range(4, (len(sequence) + 1)):
unit = ((unit + [i - 1]) * i)[:-1]
# print unit
for j in unit:
temp = sequence[j]
sequence[j] = sequence[0]
sequence[0] = temp
yield sequence
else:
print 'You can use PEN and PAPER'
# s = [1,2,3,4,5,6,7,8,9,10]
s = [x for x in 'PYTHON']
print s
z = permutations(s)
try:
while True:
# time.sleep(0.0001)
print next(z)
except StopIteration:
print 'Done'
['P', 'Y', 'T', 'H', 'O', 'N']
['Y', 'P', 'T', 'H', 'O', 'N']
['T', 'P', 'Y', 'H', 'O', 'N']
['P', 'T', 'Y', 'H', 'O', 'N']
['Y', 'T', 'P', 'H', 'O', 'N']
['T', 'Y', 'P', 'H', 'O', 'N']
['H', 'Y', 'P', 'T', 'O', 'N']
['Y', 'H', 'P', 'T', 'O', 'N']
['P', 'H', 'Y', 'T', 'O', 'N']
['H', 'P', 'Y', 'T', 'O', 'N']
['Y', 'P', 'H', 'T', 'O', 'N']
['P', 'Y', 'H', 'T', 'O', 'N']
['T', 'Y', 'H', 'P', 'O', 'N']
['Y', 'T', 'H', 'P', 'O', 'N']
['H', 'T', 'Y', 'P', 'O', 'N']
['T', 'H', 'Y', 'P', 'O', 'N']
['Y', 'H', 'T', 'P', 'O', 'N']
['H', 'Y', 'T', 'P', 'O', 'N']
['P', 'Y', 'T', 'H', 'O', 'N']
.
.
.
['Y', 'T', 'N', 'H', 'O', 'P']
['N', 'T', 'Y', 'H', 'O', 'P']
['T', 'N', 'Y', 'H', 'O', 'P']
['Y', 'N', 'T', 'H', 'O', 'P']
['N', 'Y', 'T', 'H', 'O', 'P']
Just to be complete, C++
#include <iostream>
#include <algorithm>
#include <string>
std::string theSeq = "abc";
do
{
std::cout << theSeq << endl;
}
while (std::next_permutation(theSeq.begin(), theSeq.end()));
...
abc
acb
bac
bca
cab
cba
Here is a non-recursive solution in C++ that provides the next permutation in ascending order, similarly to the functionality provided by std::next_permutation:
void permute_next(vector<int>& v)
{
if (v.size() < 2)
return;
if (v.size() == 2)
{
int tmp = v[0];
v[0] = v[1];
v[1] = tmp;
return;
}
// Step 1: find first ascending-ordered pair from right to left
int i = v.size()-2;
while(i>=0)
{
if (v[i] < v[i+1])
break;
i--;
}
if (i<0) // vector fully sorted in descending order (last permutation)
{
//resort in ascending order and return
sort(v.begin(), v.end());
return;
}
// Step 2: swap v[i] with next higher element of remaining elements
int pos = i+1;
int val = v[pos];
for(int k=i+2; k<v.size(); k++)
if(v[k] < val && v[k] > v[i])
{
pos = k;
val = v[k];
}
v[pos] = v[i];
v[i] = val;
// Step 3: sort remaining elements from i+1 ... end
sort(v.begin()+i+1, v.end());
}
In C, create a single matrix (unsigned char) to quickly and easily access all permutations from 1 to 6. Based on code from https://www.geeksforgeeks.org/heaps-algorithm-for-generating-permutations/.
void swap(unsigned char* a, unsigned char* b)
{
unsigned char t;
t = *b;
*b = *a;
*a = t;
}
void print_permutations(unsigned char a[], unsigned char n)
{
// can't rely on sizeof(a[6]) == 6, such as with MSVC 2019
for (int i = 0; i < n; i++)
{
assert(a[i] < n);
printf("%d ", a[i]);
}
printf("\n");
}
// Generating permutation using Heap Algorithm
void generate_permutations(unsigned char (**permutations)[6], unsigned char a[], int size, int n)
{
// can't rely on sizeof(a[6]) == 6, such as with MSVC 2019
// if size becomes 1 then prints the obtained permutation
if (size == 1)
{
memcpy(*permutations, a, n);
*permutations += 1;
}
else
{
for (int i = 0; i < size; i++)
{
generate_permutations(permutations, a, size - 1, n);
// if size is odd, swap first and last element
if (size & 1)
swap(a, a + size - 1);
// If size is even, swap ith and last element
else
swap(a + i, a + size - 1);
}
}
}
int main()
{
unsigned char permutations[720][6]; // easily access all permutations from 1 to 6
unsigned char suit_length_indexes[] = { 0, 1, 2, 3, 4, 5 };
assert(sizeof(suit_length_indexes) == sizeof(permutations[0]));
unsigned char(*p)[sizeof(suit_length_indexes)] = permutations;
generate_permutations(&p, suit_length_indexes, sizeof(suit_length_indexes), sizeof(suit_length_indexes));
for (int i = 0; i < sizeof(permutations) / sizeof(permutations[0]); i++)
print_permutations(permutations[i], sizeof(suit_length_indexes));
return 0;
}
Below are two classic kotlin implementations of recursive and non recursive Heap algorithm implementations:
Non recursive:
fun <T> permutationsHeapNonRecursive(list: List<T>): List<List<T>> {
val result = mutableListOf<List<T>>()
val c = Array(list.size) {0}
result.add(list.toList())
val tempList = list.toMutableList()
var i = 1
while (i < list.size) {
if (c[i] < i) {
if (i % 2 == 0)
tempList[0] = tempList[i].also { tempList[i] = tempList[0] }
else
tempList[c[i]] = tempList[i].also { tempList[i] = tempList[c[i]] }
result.add(tempList.toList())
c[i] += 1
i = 1
}
else {
c[i] = 0
i += 1
}
}
return result
}
and, recursive:
private fun <T> permutationsHeapRecursiveInternal(k: Int, list: MutableList<T>, outList: MutableList<List<T>>) {
if (k == 1) {
outList.add(List<T>(list.size) {list[it]})
}
else {
permutationsHeapRecursiveInternal(k - 1, list, outList)
for (i in 0 until k-1) {
if (k % 2 == 0)
list[i] = list[k-1].also{ list[k-1] = list[i] }
else
list[0] = list[k-1].also{ list[k-1] = list[0] }
permutationsHeapRecursiveInternal(k - 1, list, outList)
}
}
}
fun <T> permutationsHeapRecursive(list: List<T>): List<List<T>> {
val result = mutableListOf<List<T>>()
if (list.isNotEmpty()) {
val tempList = MutableList<T>(list.size) { i -> list[i] }
permutationsHeapRecursiveInternal(tempList.size, tempList, result)
}
return result
}
I have profiled non recursive version and after some tweaking with limiting memory allocations it is faster than recursive one.
As a C# Extension Method:
Usage:
List<List<string>> permutations = new List<string> { "A", "BB", "CCC" }.Permutations();
Debug.WriteLine(String.Join("; ", permutations.Select(p => string.Join(", ", p))));
Output:
A, BB, CCC; A, CCC, BB; BB, A, CCC; BB, CCC, A; CCC, A, BB; CCC, BB, A
Method:
static public List<List<T>> Permutations<T>(this List<T> list, List<T> prefillOpt = null)
{
List<List<T>> result = new List<List<T>>();
if (list.Count == 0)
{
if (prefillOpt?.Any() == true)
result.Add(prefillOpt.ToList()); // make a copy, owned by caller
}
else
{
prefillOpt = prefillOpt ?? new List<T>();
for (int i = 0; i < list.Count; i++)
{
prefillOpt.Add(list[i]);
int j = 0;
result.AddRange(Permutations(list.Where(moot => j++ != i).ToList(), prefillOpt));
prefillOpt.RemoveAt(prefillOpt.Count - 1);
}
}
return result;
}
The following is the generic implementation of Heap's algorithm (works with int, char, etc):
void generate<T>(int k, List<T> A, List<List<T>> result)
{
if (k == 1)
{
result.Add(new List<T>(A));
return;
}
for (int i = 0; i < k; i++)
{
generate(k - 1, A, result);
if (k % 2 == 0)
(A[i], A[k - 1]) = (A[k - 1], A[i]);
else
(A[0], A[k - 1]) = (A[k - 1], A[0]);
}
}
To use it:
var permutations = new List<List<char>>();
List<char> A = new List<char>() { 'a', 'b', 'c', 'd'};
generate<char>(A.Count, A, permutations);
// The output is in permutations
int c = 0;
foreach (var item in permutations)
{
c++;
Console.Write(c + ". ");
foreach (var element in item)
{
Console.Write(element);
}
Console.WriteLine();
}
Bourne shell solution - in a total of four lines (without test for no param case):
test $# -eq 1 && echo "$1" && exit
for i in $*; do
$0 `echo "$*" | sed -e "s/$i//"` | sed -e "s/^/$i /"
done
This produces them one-at-time without making a list - the same end result as Marios Choudary's answer (or simply calling C++'s nextPermute, as Anders answer notes). But this is Heap's algorithm (the non-recursive version) re-arranged and a class to save context. Used as:
P5=new genPermutes_t(5); // P5.P is now [0,1,2,3,4]
while(!P5.isDone()) {
// use P5.P here
P5.next();
}
Code is in C# without being an endorsement. Variables are as-is from Heap's pseudocode, to which the comments also refer:
public class genPermutes_t {
public int[] P; // the current permuation
private int n, i; // vars from the original algorithm
private int[] c; // ditto
public genPermutes_t(int count) {
// init algorithm:
n=count;
i=0;
c=new int[n];
for(int j=0;j<n;j++) c[j]=0;
// start current permutation as 0,1 ... n-1:
P=new int[n];
for(int j=0;j<n;j++) P[j]=j;
}
public bool isDone() {
return i>=n; // condition on the original while loop
}
public void next() {
// the part of the loop that spins until done or ready for next permute:
while(i<n && c[i]>=i) {
c[i]=0;
i++;
}
// pulled from inside loop -- the part that makes next permute:
if(i<n) { // if not done
if(i%2==0) swap(0,i);
else swap(c[i], i);
// "print P" removed. User will simply examine it
c[i]+=1;
i=0;
}
}
private void swap(int i1, int i2) {int tmp=P[i1]; P[i1]=P[i2]; P[i2]=tmp;}
}
the simplest way I can think to explain this is by using some pseudo code
so
list of 1, 2 ,3
for each item in list
templist.Add(item)
for each item2 in list
if item2 is Not item
templist.add(item)
for each item3 in list
if item2 is Not item
templist.add(item)
end if
Next
end if
Next
permanentListofPermutaitons,add(templist)
tempList.Clear()
Next
Now obviously this is not the most flexible way to do this, and doing it recursively would be a lot more functional by my tired sunday night brain doesn't want to think about that at this moment. If no ones put up a recursive version by the morning I'll do one.
