mirror of
https://github.com/TheAlgorithms/Python.git
synced 2024-11-23 21:11:08 +00:00
parent
832610ab1d
commit
3dbafd3f0d
|
@ -1,62 +0,0 @@
|
|||
from random import randint
|
||||
from tempfile import TemporaryFile
|
||||
|
||||
import numpy as np
|
||||
|
||||
|
||||
def _in_place_quick_sort(a, start, end):
|
||||
count = 0
|
||||
if start < end:
|
||||
pivot = randint(start, end)
|
||||
temp = a[end]
|
||||
a[end] = a[pivot]
|
||||
a[pivot] = temp
|
||||
|
||||
p, count = _in_place_partition(a, start, end)
|
||||
count += _in_place_quick_sort(a, start, p - 1)
|
||||
count += _in_place_quick_sort(a, p + 1, end)
|
||||
return count
|
||||
|
||||
|
||||
def _in_place_partition(a, start, end):
|
||||
count = 0
|
||||
pivot = randint(start, end)
|
||||
temp = a[end]
|
||||
a[end] = a[pivot]
|
||||
a[pivot] = temp
|
||||
new_pivot_index = start - 1
|
||||
for index in range(start, end):
|
||||
count += 1
|
||||
if a[index] < a[end]: # check if current val is less than pivot value
|
||||
new_pivot_index = new_pivot_index + 1
|
||||
temp = a[new_pivot_index]
|
||||
a[new_pivot_index] = a[index]
|
||||
a[index] = temp
|
||||
|
||||
temp = a[new_pivot_index + 1]
|
||||
a[new_pivot_index + 1] = a[end]
|
||||
a[end] = temp
|
||||
return new_pivot_index + 1, count
|
||||
|
||||
|
||||
outfile = TemporaryFile()
|
||||
p = 100 # 1000 elements are to be sorted
|
||||
|
||||
|
||||
mu, sigma = 0, 1 # mean and standard deviation
|
||||
X = np.random.normal(mu, sigma, p)
|
||||
np.save(outfile, X)
|
||||
print("The array is")
|
||||
print(X)
|
||||
|
||||
|
||||
outfile.seek(0) # using the same array
|
||||
M = np.load(outfile)
|
||||
r = len(M) - 1
|
||||
z = _in_place_quick_sort(M, 0, r)
|
||||
|
||||
print(
|
||||
"No of Comparisons for 100 elements selected from a standard normal distribution"
|
||||
"is :"
|
||||
)
|
||||
print(z)
|
Loading…
Reference in New Issue
Block a user