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]
This question is related to
python
algorithm
permutation
combinatorics
python-2.5
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
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.
in case anyone fancies this ugly one-liner (works only for strings though):
def p(a):
return a if len(a) == 1 else [[a[i], *j] for i in range(len(a)) for j in p(a[:i] + a[i + 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']))
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.
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]]
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])
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
Anyway we could use sympy library , also support for multiset permutations
import sympy
from sympy.utilities.iterables import multiset_permutations
t = [1,2,3]
p = list(multiset_permutations(t))
print(p)
# [[1, 2, 3], [1, 3, 2], [2, 1, 3], [2, 3, 1], [3, 1, 2], [3, 2, 1]]
Answer is highly inspired by Get all permutations of a numpy array
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.)
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
from typing import List
import time, random
def measure_time(func):
def wrapper_time(*args, **kwargs):
start_time = time.perf_counter()
res = func(*args, **kwargs)
end_time = time.perf_counter()
return res, end_time - start_time
return wrapper_time
class Solution:
def permute(self, nums: List[int], method: int = 1) -> List[List[int]]:
perms = []
perm = []
if method == 1:
_, time_perm = self._permute_recur(nums, 0, len(nums) - 1, perms)
elif method == 2:
_, time_perm = self._permute_recur_agian(nums, perm, perms)
print(perm)
return perms, time_perm
@measure_time
def _permute_recur(self, nums: List[int], l: int, r: int, perms: List[List[int]]):
# base case
if l == r:
perms.append(nums.copy())
for i in range(l, r + 1):
nums[l], nums[i] = nums[i], nums[l]
self._permute_recur(nums, l + 1, r , perms)
nums[l], nums[i] = nums[i], nums[l]
@measure_time
def _permute_recur_agian(self, nums: List[int], perm: List[int], perms_list: List[List[int]]):
"""
The idea is similar to nestedForLoops visualized as a recursion tree.
"""
if nums:
for i in range(len(nums)):
# perm.append(nums[i]) mistake, perm will be filled with all nums's elements.
# Method1 perm_copy = copy.deepcopy(perm)
# Method2 add in the parameter list using + (not in place)
# caveat: list.append is in-place , which is useful for operating on global element perms_list
# Note that:
# perms_list pass by reference. shallow copy
# perm + [nums[i]] pass by value instead of reference.
self._permute_recur_agian(nums[:i] + nums[i+1:], perm + [nums[i]], perms_list)
else:
# Arrive at the last loop, i.e. leaf of the recursion tree.
perms_list.append(perm)
if __name__ == "__main__":
array = [random.randint(-10, 10) for _ in range(3)]
sol = Solution()
# perms, time_perm = sol.permute(array, 1)
perms2, time_perm2 = sol.permute(array, 2)
print(perms2)
# print(perms, perms2)
# print(time_perm, time_perm2)
```
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)
for Python we can use itertools and import both permutations and combinations to solve your problem
from itertools import product, permutations
A = ([1,2,3])
print (list(permutations(sorted(A),2)))
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]
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.
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.
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:]))
def permutations(head, tail=''):
if len(head) == 0:
print(tail)
else:
for i in range(len(head)):
permutations(head[: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 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
import java.util.Arrays;
public class Permutation {
/* runtime -O(n) for generating nextPermutaion
* and O(n*n!) for generating all n! permutations with increasing sorted array as start
* return true, if there exists next lexicographical sequence
* e.g [1,2,3],3-> true, modifies array to [1,3,2]
* e.g [3,2,1],3-> false, as it is largest lexicographic possible */
public static boolean nextPermutation(int[] seq, int len) {
// 1
if (len <= 1)
return false;// no more perm
// 2: Find last j such that seq[j] <= seq[j+1]. Terminate if no such j exists
int j = len - 2;
while (j >= 0 && seq[j] >= seq[j + 1]) {
--j;
}
if (j == -1)
return false;// no more perm
// 3: Find last l such that seq[j] <= seq[l], then exchange elements j and l
int l = len - 1;
while (seq[j] >= seq[l]) {
--l;
}
swap(seq, j, l);
// 4: Reverse elements j+1 ... count-1:
reverseSubArray(seq, j + 1, len - 1);
// return seq, add store next perm
return true;
}
private static void swap(int[] a, int i, int j) {
int temp = a[i];
a[i] = a[j];
a[j] = temp;
}
private static void reverseSubArray(int[] a, int lo, int hi) {
while (lo < hi) {
swap(a, lo, hi);
++lo;
--hi;
}
}
public static void main(String[] args) {
int[] array = {1,2,3,4,5,6,7};
int cnt=0;
do {
System.out.println(Arrays.toString(array));
cnt++;
}while(nextPermutation(array, array.length));
System.out.println(cnt);//5040=7!
}
}
Algorithm for generating all permutations(Java). It generates the next lexicographic sequence, modifies the array in place.
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]]
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)
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 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
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])
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']
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)
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]
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
def permuteArray (arr):
arraySize = len(arr)
permutedList = []
if arraySize == 1:
return [arr]
i = 0
for item in arr:
for elem in permuteArray(arr[:i] + arr[i + 1:]):
permutedList.append([item] + elem)
i = i + 1
return permutedList
I intended to not exhaust every possibility in a new line to make it somewhat unique.
The following code with Python 2.6 and above ONLY
First, import itertools
:
import itertools
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)]
print list(itertools.combinations('123', 2))
[('1', '2'), ('1', '3'), ('2', '3')]
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)]
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)]
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") )
Source: Stackoverflow.com