以下代码是给定列表的就地排列，作为生成器实现。由于它只返回对列表的引用，因此不应在生成器外部修改列表。该解决方案是非递归的，因此使用了低内存。还可以很好地处理输入列表中元素的多个副本。

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

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)
```

如果有人喜欢这个丑陋的单行线（虽然只适用于字符串）：

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:])]

我使用了一种基于阶乘数系统的算法——对于长度为n的列表，您可以逐项组装每个排列，从每个阶段留下的项目中进行选择。第一项有n个选项，第二项有n-1个选项，最后一项只有一个选项，因此可以使用阶乘数系统中数字的数字作为索引。这是数字0到n-1对应于词典顺序中的所有可能的排列。

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))

输出：

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

此方法是非递归的，但在我的计算机上速度稍慢，xrange在n！太大，无法转换为C长整数（我的n=13）。当我需要它的时候，它已经足够了，但它远没有itertools.permutations。

递归之美：

>>> 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']

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。