为了节省您可能的搜索和实验时间，下面是Python中的非递归置换解决方案，它也适用于Numba（从0.41版开始）：

@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]]

要给人留下绩效印象：

%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

因此，只有在必须从njit函数调用它时才使用此版本，否则更倾向于itertools实现。

这是受Haskell实现使用列表理解的启发：

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

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)
```

对于Python 2.6及以上版本：

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

这将作为生成器返回。使用列表（排列（xs））作为列表返回。

生成所有可能的排列

我正在使用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

测试用例：

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

人们确实可以对每个排列的第一个元素进行迭代，正如tzwen的答案。然而，这样编写此解决方案更有效：

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

这个解决方案大约快了30%，显然是因为递归以len（元素）<=1而不是0结尾。它的内存效率也高得多，因为它使用了一个生成器函数（通过yield），就像Riccardo Reyes的解决方案一样。