Python 排序与查找算法收集
Python 语言实现几种不同的排序算法,代码来自于老男孩Python全栈开发,学习教程!
import random
import time
import copy
import sys
def cal_time(func):
def wrapper(*args, **kwargs):
t1 = time.time()
result = func(*args, **kwargs)
t2 = time.time()
print("%s running time: %s secs." % (func.__name__, t2 - t1))
return result
return wrapper
@cal_time
def bubble_sort(li):
for i in range(len(li) - 1):
for j in range(len(li) - i - 1):
if li[j] > li[j+1]:
li[j], li[j+1] = li[j+1], li[j]
@cal_time
def bubble_sort_1(li):
for i in range(len(li) - 1):
exchange = False
for j in range(len(li) - i - 1):
if li[j] > li[j+1]:
li[j], li[j+1] = li[j+1], li[j]
exchange = True
if not exchange:
break
def select_sort(li):
for i in range(len(li) - 1):
min_loc = i
for j in range(i+1,len(li)):
if li[j] < li[min_loc]:
min_loc = j
li[i], li[min_loc] = li[min_loc], li[i]
def insert_sort(li):
for i in range(1, len(li)):
tmp = li[i]
j = i - 1
while j >= 0 and li[j] > tmp:
li[j+1]=li[j]
j = j - 1
li[j + 1] = tmp
def quick_sort_x(data, left, right):
if left < right:
mid = partition(data, left, right)
quick_sort_x(data, left, mid - 1)
quick_sort_x(data, mid + 1, right)
def partition(data, left, right):
tmp = data[left]
while left < right:
while left < right and data[right] >= tmp:
right -= 1
data[left] = data[right]
while left < right and data[left] <= tmp:
left += 1
data[right] = data[left]
data[left] = tmp
return left
@cal_time
def quick_sort(data):
return quick_sort_x(data, 0, len(data) - 1)
@cal_time
def sys_sort(data):
return data.sort()
def sift(data, low, high):
i = low
j = 2 * i + 1
tmp = data[i]
while j <= high: #只要没到子树的最后
if j < high and data[j] < data[j + 1]:
j += 1
if tmp < data[j]:#如果领导不能干
data[i] = data[j] #小领导上位
i = j
j = 2 * i + 1
else:
break
data[i] = tmp
def heap_sort(data):
n = len(data)
for i in range(n // 2 - 1, -1, -1):
sift(data, i, n - 1)
for i in range(n - 1, -1, -1):
data[0], data[i] = data[i], data[0]
sift(data, 0, i - 1)
sys.setrecursionlimit(100000)
data = list(range(1000, 1, -1))
data.sort()
#random.shuffle(data)
data1 = copy.deepcopy(data)
data2 = copy.deepcopy(data)
data3 = copy.deepcopy(data)
bubble_sort(data1)
quick_sort(data2)
sys_sort(data3)
Python语言实现的几种,不同的查找算法。
import time
import random
def cal_time(func):
def wrapper(*args, **kwargs):
t1 = time.time()
result = func(*args, **kwargs)
t2 = time.time()
print("%s running time: %s secs." % (func.__name__, t2 - t1))
return result
return wrapper
@cal_time
def bin_search(data_set, val):
low = 0
high = len(data_set) - 1
while low <= high:
mid = (low+high)//2
if data_set[mid]['id'] == val:
return mid
elif data_set[mid]['id'] < val:
low = mid + 1
else:
high = mid - 1
return
def binary_search(dataset, find_num):
if len(dataset) > 1:
mid = int(len(dataset) / 2)
if dataset[mid] == find_num:
#print("Find it")
return dataset[mid]
elif dataset[mid] > find_num:
return binary_search(dataset[0:mid], find_num)
else:
return binary_search(dataset[mid + 1:], find_num)
else:
if dataset[0] == find_num:
#print("Find it")
return dataset[0]
else:
pass
#print("Cannot find it.")
@cal_time
def binary_search_alex(data_set, val):
return binary_search(data_set, val)
def random_list(n):
result = []
ids = list(range(1001,1001+n))
a1 = ['zhao','qian','sun','li']
a2 = ['li','hao','','']
a3 = ['qiang','guo']
for i in range(n):
age = random.randint(18,60)
id = ids[i]
name = random.choice(a1)+random.choice(a2)+random.choice(a3)
data = list(range(100000000))
print(bin_search(data, 173320))
print(binary_search_alex(data, 173320))
文章出处:https://www.cnblogs.com/LyShark/p/11983037.html
本博客所有文章除特别声明外,均采用 BY-NC-SA 许可协议。转载请注明出处!
本博客所有文章除特别声明外,均采用 BY-NC-SA 许可协议。转载请注明出处!