您有 k 个子列表。在每次迭代中,找到第一个元素最小的子列表;将该元素附加到结果列表中;在该子列表中推进一个,而不在其他子列表中推进。
如果您有一个函数 arg_min 或 min_with_index 可以更容易地为您提供最小元素及其索引(这样您就知道它来自哪个子列表)。
这里有两种等效的编写函数min_with_index的方法,使用python的内置函数min获取最小值,enumerate获取索引:
def min_with_index(it):
return min(enumerate(it), key=lambda p:p[1])
import operator
def min_with_index(it):
return min(enumerate(it), key=operator.itemgetter(1))
# >>> min_with_index([14,16,13,15])
# (2, 13)
这是为了合并。以下是使用列表切片的两种不同拆分方式:
def split_kway_1(l, k):
return [l[i::k] for i in range(k)]
def split_kway_2(l, k):
j = (len(l)-1) // k + 1
return [l[i:i+j] for i in range(0,len(l),j)]
def split_kway_3(l, k):
j = len(l) // k
result = [l[i:i+j] for i in range(0, j*(k-1), j)]
result.append(l[j*(k-1):])
return result
# >>> split_kway_1(list(range(10)), 3)
# [[0, 3, 6, 9], [1, 4, 7], [2, 5, 8]]
# >>> split_kway_2(list(range(10)), 3)
# [[0, 1, 2, 3], [4, 5, 6, 7], [8, 9]]
# >>> split_kway_3(list(range(10)), 3)
# [[0, 1, 2], [3, 4, 5], [6, 7, 8, 9]]
# versions 2 and 3 differ only when the length of the list is not a multiple of k
现在我们可以结合拆分和合并来编写归并排序:
import operator
def split_kway(l, k):
return [l[i::k] for i in range(k)]
def min_with_index(it):
return min(enumerate(it), key=operator.itemgetter(1))
def merge_kway(list_of_sublists):
result = []
list_of_sublists = [l for l in list_of_sublists if len(l) > 0]
while list_of_sublists:
i,v = min_with_index(l[0] for l in list_of_sublists)
result.append(v)
if len(list_of_sublists[i]) > 1:
list_of_sublists[i].pop(0) # advance in sublist i
else:
list_of_sublists.pop(i) # remove sublist i which is now empty
return result
def merge_sort_kway(l, k):
if len(l) > 1:
list_of_sublists = split_kway(l, k)
list_of_sublists = [merge_sort_kway(l, k) for l in list_of_sublists]
return merge_kway(list_of_sublists)
else:
return list(l)
另请参阅:Wikipedia on k-way merge