如何生成列表的所有排列？

How do you generate all the permutations of a list in Python, independently of the type of elements in that list?

For example:

permutations([])
[]

permutations([1])
[1]

permutations([1, 2])
[1, 2]
[2, 1]

permutations([1, 2, 3])
[1, 2, 3]
[1, 3, 2]
[2, 1, 3]
[2, 3, 1]
[3, 1, 2]
[3, 2, 1]

Starting with Python 2.6 (and if you’re on Python 3) you have a standard-library tool for this: itertools.permutations.

import itertools
list(itertools.permutations([1, 2, 3]))

If you’re using an older Python (<2.6) for some reason or are just curious to know how it works, here’s one nice approach, taken from http://code.activestate.com/recipes/252178/:

def all_perms(elements):
    if len(elements) <=1:
        yield elements
    else:
        for perm in all_perms(elements[1:]):
            for i in range(len(elements)):
                # nb elements[0:1] works in both string and list contexts
                yield perm[:i] + elements[0:1] + perm[i:]

A couple of alternative approaches are listed in the documentation of itertools.permutations. Here’s one:

def permutations(iterable, r=None):
    # permutations('ABCD', 2) --> AB AC AD BA BC BD CA CB CD DA DB DC
    # permutations(range(3)) --> 012 021 102 120 201 210
    pool = tuple(iterable)
    n = len(pool)
    r = n if r is None else r
    if r > n:
        return
    indices = range(n)
    cycles = range(n, n-r, -1)
    yield tuple(pool[i] for i in indices[:r])
    while n:
        for i in reversed(range(r)):
            cycles[i] -= 1
            if cycles[i] == 0:
                indices[i:] = indices[i+1:] + indices[i:i+1]
                cycles[i] = n - i
            else:
                j = cycles[i]
                indices[i], indices[-j] = indices[-j], indices[i]
                yield tuple(pool[i] for i in indices[:r])
                break
        else:
            return

And another, based on itertools.product:

def permutations(iterable, r=None):
    pool = tuple(iterable)
    n = len(pool)
    r = n if r is None else r
    for indices in product(range(n), repeat=r):
        if len(set(indices)) == r:
            yield tuple(pool[i] for i in indices)

And in Python 2.6 onwards:

import itertools
itertools.permutations([1,2,3])

(returned as a generator. Use list(permutations(l)) to return as a list.)

The following code with Python 2.6 and above ONLY

First, import itertools:

import itertools

Permutation (order matters):

print list(itertools.permutations([1,2,3,4], 2))
[(1, 2), (1, 3), (1, 4),
(2, 1), (2, 3), (2, 4),
(3, 1), (3, 2), (3, 4),
(4, 1), (4, 2), (4, 3)]

Combination (order does NOT matter):

print list(itertools.combinations('123', 2))
[('1', '2'), ('1', '3'), ('2', '3')]

Cartesian product (with several iterables):

print list(itertools.product([1,2,3], [4,5,6]))
[(1, 4), (1, 5), (1, 6),
(2, 4), (2, 5), (2, 6),
(3, 4), (3, 5), (3, 6)]

Cartesian product (with one iterable and itself):

print list(itertools.product([1,2], repeat=3))
[(1, 1, 1), (1, 1, 2), (1, 2, 1), (1, 2, 2),
(2, 1, 1), (2, 1, 2), (2, 2, 1), (2, 2, 2)]

def permutations(head, tail=''):
    if len(head) == 0: print tail
    else:
        for i in range(len(head)):
            permutations(head[0:i] + head[i+1:], tail+head[i])

called as:

permutations('abc')

#!/usr/bin/env python

def perm(a, k=0):
   if k == len(a):
      print a
   else:
      for i in xrange(k, len(a)):
         a[k], a[i] = a[i] ,a[k]
         perm(a, k+1)
         a[k], a[i] = a[i], a[k]

perm([1,2,3])

Output:

[1, 2, 3]
[1, 3, 2]
[2, 1, 3]
[2, 3, 1]
[3, 2, 1]
[3, 1, 2]

As I’m swapping the content of the list it’s required a mutable sequence type as input. E.g. perm(list("ball")) will work and perm("ball") won’t because you can’t change a string.

This Python implementation is inspired by the algorithm presented in the book Computer Algorithms by Horowitz, Sahni and Rajasekeran.

This solution implements a generator, to avoid holding all the permutations on memory:

def permutations (orig_list):
    if not isinstance(orig_list, list):
        orig_list = list(orig_list)

    yield orig_list

    if len(orig_list) == 1:
        return

    for n in sorted(orig_list):
        new_list = orig_list[:]
        pos = new_list.index(n)
        del(new_list[pos])
        new_list.insert(0, n)
        for resto in permutations(new_list[1:]):
            if new_list[:1] + resto <> orig_list:
                yield new_list[:1] + resto

In a functional style

def addperm(x,l):
    return [ l[0:i] + [x] + l[i:]  for i in range(len(l)+1) ]

def perm(l):
    if len(l) == 0:
        return [[]]
    return [x for y in perm(l[1:]) for x in addperm(l[0],y) ]

print perm([ i for i in range(3)])

The result:

[[0, 1, 2], [1, 0, 2], [1, 2, 0], [0, 2, 1], [2, 0, 1], [2, 1, 0]]

The following code is an in-place permutation of a given list, implemented as a generator. Since it only returns references to the list, the list should not be modified outside the generator. The solution is non-recursive, so uses low memory. Work well also with multiple copies of elements in the input list.

def permute_in_place(a):
    a.sort()
    yield list(a)

    if len(a) <= 1:
        return

    first = 0
    last = len(a)
    while 1:
        i = last - 1

        while 1:
            i = i - 1
            if a[i] < a[i+1]:
                j = last - 1
                while not (a[i] < a[j]):
                    j = j - 1
                a[i], a[j] = a[j], a[i] # swap the values
                r = a[i+1:last]
                r.reverse()
                a[i+1:last] = r
                yield list(a)
                break
            if i == first:
                a.reverse()
                return

if __name__ == '__main__':
    for n in range(5):
        for a in permute_in_place(range(1, n+1)):
            print a
        print

    for a in permute_in_place([0, 0, 1, 1, 1]):
        print a
    print

A quite obvious way in my opinion might be also:

def permutList(l):
    if not l:
            return [[]]
    res = []
    for e in l:
            temp = l[:]
            temp.remove(e)
            res.extend([[e] + r for r in permutList(temp)])

    return res

list2Perm = [1, 2.0, 'three']
listPerm = [[a, b, c]
            for a in list2Perm
            for b in list2Perm
            for c in list2Perm
            if ( a != b and b != c and a != c )
            ]
print listPerm

Output:

[
    [1, 2.0, 'three'], 
    [1, 'three', 2.0], 
    [2.0, 1, 'three'], 
    [2.0, 'three', 1], 
    ['three', 1, 2.0], 
    ['three', 2.0, 1]
]

I used an algorithm based on the factorial number system– For a list of length n, you can assemble each permutation item by item, selecting from the items left at each stage. You have n choices for the first item, n-1 for the second, and only one for the last, so you can use the digits of a number in the factorial number system as the indices. This way the numbers 0 through n!-1 correspond to all possible permutations in lexicographic order.

from math import factorial
def permutations(l):
    permutations=[]
    length=len(l)
    for x in xrange(factorial(length)):
        available=list(l)
        newPermutation=[]
        for radix in xrange(length, 0, -1):
            placeValue=factorial(radix-1)
            index=x/placeValue
            newPermutation.append(available.pop(index))
            x-=index*placeValue
        permutations.append(newPermutation)
    return permutations

permutations(range(3))

output:

[[0, 1, 2], [0, 2, 1], [1, 0, 2], [1, 2, 0], [2, 0, 1], [2, 1, 0]]

This method is non-recursive, but it is slightly slower on my computer and xrange raises an error when n! is too large to be converted to a C long integer (n=13 for me). It was enough when I needed it, but it’s no itertools.permutations by a long shot.

Note that this algorithm has an n factorial time complexity, where n is the length of the input list

Print the results on the run:

global result
result = [] 

def permutation(li):
if li == [] or li == None:
    return

if len(li) == 1:
    result.append(li[0])
    print result
    result.pop()
    return

for i in range(0,len(li)):
    result.append(li[i])
    permutation(li[:i] + li[i+1:])
    result.pop()    

Example:

permutation([1,2,3])

Output:

[1, 2, 3]
[1, 3, 2]
[2, 1, 3]
[2, 3, 1]
[3, 1, 2]
[3, 2, 1]

One can indeed iterate over the first element of each permutation, as in tzwenn’s answer. It is however more efficient to write this solution this way:

def all_perms(elements):
    if len(elements) <= 1:
        yield elements  # Only permutation possible = no permutation
    else:
        # Iteration over the first element in the result permutation:
        for (index, first_elmt) in enumerate(elements):
            other_elmts = elements[:index]+elements[index+1:]
            for permutation in all_perms(other_elmts): 
                yield [first_elmt] + permutation

This solution is about 30 % faster, apparently thanks to the recursion ending at len(elements) <= 1 instead of 0. It is also much more memory-efficient, as it uses a generator function (through yield), like in Riccardo Reyes’s solution.

This is inspired by the Haskell implementation using list comprehension:

def permutation(list):
    if len(list) == 0:
        return [[]]
    else:
        return [[x] + ys for x in list for ys in permutation(delete(list, x))]

def delete(list, item):
    lc = list[:]
    lc.remove(item)
    return lc

Regular implementation (no yield – will do everything in memory):

def getPermutations(array):
    if len(array) == 1:
        return [array]
    permutations = []
    for i in range(len(array)): 
        # get all perm's of subarray w/o current item
        perms = getPermutations(array[:i] + array[i+1:])  
        for p in perms:
            permutations.append([array[i], *p])
    return permutations

Yield implementation:

def getPermutations(array):
    if len(array) == 1:
        yield array
    else:
        for i in range(len(array)):
            perms = getPermutations(array[:i] + array[i+1:])
            for p in perms:
                yield [array[i], *p]

The basic idea is to go over all the elements in the array for the 1st position, and then in 2nd position go over all the rest of the elements without the chosen element for the 1st, etc. You can do this with recursion, where the stop criteria is getting to an array of 1 element – in which case you return that array.

For performance, a numpy solution inspired by Knuth, (p22) :

from numpy import empty, uint8
from math import factorial

def perms(n):
    f = 1
    p = empty((2*n-1, factorial(n)), uint8)
    for i in range(n):
        p[i, :f] = i
        p[i+1:2*i+1, :f] = p[:i, :f]  # constitution de blocs
        for j in range(i):
            p[:i+1, f*(j+1):f*(j+2)] = p[j+1:j+i+2, :f]  # copie de blocs
        f = f*(i+1)
    return p[:n, :]

Copying large blocs of memory saves time – it’s 20x faster than list(itertools.permutations(range(n)) :

In [1]: %timeit -n10 list(permutations(range(10)))
10 loops, best of 3: 815 ms per loop

In [2]: %timeit -n100 perms(10) 
100 loops, best of 3: 40 ms per loop

from __future__ import print_function

def perm(n):
    p = []
    for i in range(0,n+1):
        p.append(i)
    while True:
        for i in range(1,n+1):
            print(p[i], end=' ')
        print("")
        i = n - 1
        found = 0
        while (not found and i>0):
            if p[i]<p[i+1]:
                found = 1
            else:
                i = i - 1
        k = n
        while p[i]>p[k]:
            k = k - 1
        aux = p[i]
        p[i] = p[k]
        p[k] = aux
        for j in range(1,(n-i)/2+1):
            aux = p[i+j]
            p[i+j] = p[n-j+1]
            p[n-j+1] = aux
        if not found:
            break

perm(5)

Here is an algorithm that works on a list without creating new intermediate lists similar to Ber’s solution at https://stackoverflow.com/a/108651/184528.

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

The beauty of recursion:

>>> import copy
>>> def perm(prefix,rest):
...      for e in rest:
...              new_rest=copy.copy(rest)
...              new_prefix=copy.copy(prefix)
...              new_prefix.append(e)
...              new_rest.remove(e)
...              if len(new_rest) == 0:
...                      print new_prefix + new_rest
...                      continue
...              perm(new_prefix,new_rest)
... 
>>> perm([],['a','b','c','d'])
['a', 'b', 'c', 'd']
['a', 'b', 'd', 'c']
['a', 'c', 'b', 'd']
['a', 'c', 'd', 'b']
['a', 'd', 'b', 'c']
['a', 'd', 'c', 'b']
['b', 'a', 'c', 'd']
['b', 'a', 'd', 'c']
['b', 'c', 'a', 'd']
['b', 'c', 'd', 'a']
['b', 'd', 'a', 'c']
['b', 'd', 'c', 'a']
['c', 'a', 'b', 'd']
['c', 'a', 'd', 'b']
['c', 'b', 'a', 'd']
['c', 'b', 'd', 'a']
['c', 'd', 'a', 'b']
['c', 'd', 'b', 'a']
['d', 'a', 'b', 'c']
['d', 'a', 'c', 'b']
['d', 'b', 'a', 'c']
['d', 'b', 'c', 'a']
['d', 'c', 'a', 'b']
['d', 'c', 'b', 'a']

This algorithm is the most effective one, it avoids of array passing and manipulation in recursive calls, works in Python 2, 3:

def permute(items):
    length = len(items)
    def inner(ix=[]):
        do_yield = len(ix) == length - 1
        for i in range(0, length):
            if i in ix: #avoid duplicates
                continue
            if do_yield:
                yield tuple([items[y] for y in ix + [i]])
            else:
                for p in inner(ix + [i]):
                    yield p
    return inner()

Usage:

for p in permute((1,2,3)):
    print(p)

(1, 2, 3)
(1, 3, 2)
(2, 1, 3)
(2, 3, 1)
(3, 1, 2)
(3, 2, 1)

def pzip(c, seq):
    result = []
    for item in seq:
        for i in range(len(item)+1):
            result.append(item[i:]+c+item[:i])
    return result


def perm(line):
    seq = [c for c in line]
    if len(seq) <=1 :
        return seq
    else:
        return pzip(seq[0], perm(seq[1:]))

ANOTHER APPROACH (without libs)

def permutation(input):
    if len(input) == 1:
        return input if isinstance(input, list) else [input]

    result = []
    for i in range(len(input)):
        first = input[i]
        rest = input[:i] + input[i + 1:]
        rest_permutation = permutation(rest)
        for p in rest_permutation:
            result.append(first + p)
    return result

Input can be a string or a list

print(permutation('abcd'))
print(permutation(['a', 'b', 'c', 'd']))

Disclaimer: shapeless plug by package author. :)

The trotter package is different from most implementations in that it generates pseudo lists that don’t actually contain permutations but rather describe mappings between permutations and respective positions in an ordering, making it possible to work with very large ‘lists’ of permutations, as shown in this demo which performs pretty instantaneous operations and look-ups in a pseudo-list ‘containing’ all the permutations of the letters in the alphabet, without using more memory or processing than a typical web page.

In any case, to generate a list of permutations, we can do the following.

import trotter

my_permutations = trotter.Permutations(3, [1, 2, 3])

print(my_permutations)

for p in my_permutations:
    print(p)

Output:

A pseudo-list containing 6 3-permutations of [1, 2, 3].
[1, 2, 3]
[1, 3, 2]
[3, 1, 2]
[3, 2, 1]
[2, 3, 1]
[2, 1, 3]

Generate all possible permutations

I’m using python3.4:

def calcperm(arr, size):
    result = set([()])
    for dummy_idx in range(size):
        temp = set()
        for dummy_lst in result:
            for dummy_outcome in arr:
                if dummy_outcome not in dummy_lst:
                    new_seq = list(dummy_lst)
                    new_seq.append(dummy_outcome)
                    temp.add(tuple(new_seq))
        result = temp
    return result

Test Cases:

lst = [1, 2, 3, 4]
#lst = ["yellow", "magenta", "white", "blue"]
seq = 2
final = calcperm(lst, seq)
print(len(final))
print(final)

To save you folks possible hours of searching and experimenting, here’s the non-recursive permutaions solution in Python which also works with Numba (as of v. 0.41):

@numba.njit()
def permutations(A, k):
    r = [[i for i in range(0)]]
    for i in range(k):
        r = [[a] + b for a in A for b in r if (a in b)==False]
    return r
permutations([1,2,3],3)
[[1, 2, 3], [1, 3, 2], [2, 1, 3], [2, 3, 1], [3, 1, 2], [3, 2, 1]]

To give an impression about performance:

%timeit permutations(np.arange(5),5)

243 µs ± 11.1 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)
time: 406 ms

%timeit list(itertools.permutations(np.arange(5),5))
15.9 µs ± 8.61 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
time: 12.9 s

So use this version only if you have to call it from njitted function, otherwise prefer itertools implementation.

I see a lot of iteration going on inside these recursive functions, not exactly pure recursion…

so for those of you who cannot abide by even a single loop, here’s a gross, totally unnecessary fully recursive solution

def all_insert(x, e, i=0):
    return [x[0:i]+[e]+x[i:]] + all_insert(x,e,i+1) if i<len(x)+1 else []

def for_each(X, e):
    return all_insert(X[0], e) + for_each(X[1:],e) if X else []

def permute(x):
    return [x] if len(x) < 2 else for_each( permute(x[1:]) , x[0])


perms = permute([1,2,3])

Another solution:

def permutation(flag, k =1 ):
    N = len(flag)
    for i in xrange(0, N):
        if flag[i] != 0:
            continue
        flag[i] = k 
        if k == N:
            print flag
        permutation(flag, k+1)
        flag[i] = 0

permutation([0, 0, 0])

My Python Solution:

def permutes(input,offset):
    if( len(input) == offset ):
        return [''.join(input)]

    result=[]        
    for i in range( offset, len(input) ):
         input[offset], input[i] = input[i], input[offset]
         result = result + permutes(input,offset+1)
         input[offset], input[i] = input[i], input[offset]
    return result

# input is a "string"
# return value is a list of strings
def permutations(input):
    return permutes( list(input), 0 )

# Main Program
print( permutations("wxyz") )

def permutation(word, first_char=None):
    if word == None or len(word) == 0: return []
    if len(word) == 1: return [word]

    result = []
    first_char = word[0]
    for sub_word in permutation(word[1:], first_char):
        result += insert(first_char, sub_word)
    return sorted(result)

def insert(ch, sub_word):
    arr = [ch + sub_word]
    for i in range(len(sub_word)):
        arr.append(sub_word[i:] + ch + sub_word[:i])
    return arr


assert permutation(None) == []
assert permutation('') == []
assert permutation('1')  == ['1']
assert permutation('12') == ['12', '21']

print permutation('abc')

Output: [‘abc’, ‘acb’, ‘bac’, ‘bca’, ‘cab’, ‘cba’]

Using Counter

from collections import Counter

def permutations(nums):
    ans = [[]]
    cache = Counter(nums)

    for idx, x in enumerate(nums):
        result = []
        for items in ans:
            cache1 = Counter(items)
            for id, n in enumerate(nums):
                if cache[n] != cache1[n] and items + [n] not in result:
                    result.append(items + [n])

        ans = result
    return ans
permutations([1, 2, 2])
> [[1, 2, 2], [2, 1, 2], [2, 2, 1]]