You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
166 lines
3.3 KiB
166 lines
3.3 KiB
=begin
|
|
File: time_complexity.rb
|
|
Created Time: 2024-03-30
|
|
Author: Xuan Khoa Tu Nguyen (ngxktuzkai2000@gmail.com)
|
|
=end
|
|
|
|
### 常数阶 ###
|
|
def constant(n)
|
|
count = 0
|
|
size = 100000
|
|
|
|
(0...size).each { count += 1 }
|
|
|
|
count
|
|
end
|
|
|
|
### 线性阶 ###
|
|
def linear(n)
|
|
count = 0
|
|
(0...n).each { count += 1 }
|
|
count
|
|
end
|
|
|
|
### 线性阶(遍历数组)###
|
|
def array_traversal(nums)
|
|
count = 0
|
|
|
|
# 循环次数与数组长度成正比
|
|
for num in nums
|
|
count += 1
|
|
end
|
|
|
|
count
|
|
end
|
|
|
|
### 平方阶 ###
|
|
def quadratic(n)
|
|
count = 0
|
|
|
|
# 循环次数与数据大小 n 成平方关系
|
|
for i in 0...n
|
|
for j in 0...n
|
|
count += 1
|
|
end
|
|
end
|
|
|
|
count
|
|
end
|
|
|
|
### 平方阶(冒泡排序)###
|
|
def bubble_sort(nums)
|
|
count = 0 # 计数器
|
|
|
|
# 外循环:未排序区间为 [0, i]
|
|
for i in (nums.length - 1).downto(0)
|
|
# 内循环:将未排序区间 [0, i] 中的最大元素交换至该区间的最右端
|
|
for j in 0...i
|
|
if nums[j] > nums[j + 1]
|
|
# 交换 nums[j] 与 nums[j + 1]
|
|
tmp = nums[j]
|
|
nums[j] = nums[j + 1]
|
|
nums[j + 1] = tmp
|
|
count += 3 # 元素交换包含 3 个单元操作
|
|
end
|
|
end
|
|
end
|
|
|
|
count
|
|
end
|
|
|
|
### 指数阶(循环实现)###
|
|
def exponential(n)
|
|
count, base = 0, 1
|
|
|
|
# 细胞每轮一分为二,形成数列 1, 2, 4, 8, ..., 2^(n-1)
|
|
(0...n).each do
|
|
(0...base).each { count += 1 }
|
|
base *= 2
|
|
end
|
|
|
|
# count = 1 + 2 + 4 + 8 + .. + 2^(n-1) = 2^n - 1
|
|
count
|
|
end
|
|
|
|
### 指数阶(递归实现)###
|
|
def exp_recur(n)
|
|
return 1 if n == 1
|
|
exp_recur(n - 1) + exp_recur(n - 1) + 1
|
|
end
|
|
|
|
### 对数阶(循环实现)###
|
|
def logarithmic(n)
|
|
count = 0
|
|
|
|
while n > 1
|
|
n /= 2
|
|
count += 1
|
|
end
|
|
|
|
count
|
|
end
|
|
|
|
### 对数阶(递归实现)###
|
|
def log_recur(n)
|
|
return 0 unless n > 1
|
|
log_recur(n / 2) + 1
|
|
end
|
|
|
|
### 线性对数阶 ###
|
|
def linear_log_recur(n)
|
|
return 1 unless n > 1
|
|
|
|
count = linear_log_recur(n / 2) + linear_log_recur(n / 2)
|
|
(0...n).each { count += 1 }
|
|
|
|
count
|
|
end
|
|
|
|
### 阶乘阶(递归实现)###
|
|
def factorial_recur(n)
|
|
return 1 if n == 0
|
|
|
|
count = 0
|
|
# 从 1 个分裂出 n 个
|
|
(0...n).each { count += factorial_recur(n - 1) }
|
|
|
|
count
|
|
end
|
|
|
|
### Driver Code ###
|
|
if __FILE__ == $0
|
|
# 可以修改 n 运行,体会一下各种复杂度的操作数量变化趋势
|
|
n = 8
|
|
puts "输入数据大小 n = #{n}"
|
|
|
|
count = constant(n)
|
|
puts "常数阶的操作数量 = #{count}"
|
|
|
|
count = linear(n)
|
|
puts "线性阶的操作数量 = #{count}"
|
|
count = array_traversal(Array.new(n, 0))
|
|
puts "线性阶(遍历数组)的操作数量 = #{count}"
|
|
|
|
count = quadratic(n)
|
|
puts "平方阶的操作数量 = #{count}"
|
|
nums = Array.new(n) { |i| n - i } # [n, n-1, ..., 2, 1]
|
|
count = bubble_sort(nums)
|
|
puts "平方阶(冒泡排序)的操作数量 = #{count}"
|
|
|
|
count = exponential(n)
|
|
puts "指数阶(循环实现)的操作数量 = #{count}"
|
|
count = exp_recur(n)
|
|
puts "指数阶(递归实现)的操作数量 = #{count}"
|
|
|
|
count = logarithmic(n)
|
|
puts "对数阶(循环实现)的操作数量 = #{count}"
|
|
count = log_recur(n)
|
|
puts "对数阶(递归实现)的操作数量 = #{count}"
|
|
|
|
count = linear_log_recur(n)
|
|
puts "线性对数阶(递归实现)的操作数量 = #{count}"
|
|
|
|
count = factorial_recur(n)
|
|
puts "阶乘阶(递归实现)的操作数量 = #{count}"
|
|
end
|