如何优化Python按键模拟代码 缩小执行时间与延迟总和差值？

问题原因 #

总执行时间远大于两次延迟总和，核心来自这些额外开销：

• pyautogui.keyDown()/keyUp()的跨平台封装开销：pyautogui为兼容多平台做了大量上层逻辑，底层调用的额外耗时明显
• random_delay函数中的print操作：IO打印是典型的耗时操作，会增加非必要执行时间
• Python函数调用的层级开销：多次函数调用会累积额外执行时间
• time.sleep()的精度限制：系统调度机制会让实际休眠时间略长于设定值，小延迟场景下误差占比更高

优化步骤 #

1. 移除不必要的IO操作 #

直接删除random_delay中的打印语句，仅在调试阶段临时启用，避免IO耗时拖慢整体执行。

2. 简化函数结构，减少调用层级 #

把延迟计算逻辑直接内联到主函数中，避免额外的函数调用开销。

3. 替换轻量输入模拟库 #

用pynput替代pyautogui，pynput的底层实现更简洁，跨平台兼容的同时保持更低的调用开销；Windows平台还可以直接调用系统API，进一步砍掉中间层开销。

4. 实现精确延迟控制 #

用time.perf_counter()手动循环等待，代替time.sleep()，减少系统调度带来的延迟误差。

优化后的代码示例 #

方案一：优化pyautogui版本 #

import time
import random
import pyautogui

def key_sim(key, min_delay_ms, max_delay_ms):
    start_time = time.perf_counter()

    # 按下按键并执行精确延迟
    press_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    pyautogui.keyDown(key)
    target_time = time.perf_counter() + press_delay
    while time.perf_counter() < target_time:
        pass

    # 释放按键并执行精确延迟
    release_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    pyautogui.keyUp(key)
    target_time = time.perf_counter() + release_delay
    while time.perf_counter() < target_time:
        pass

    return time.perf_counter() - start_time

# 执行测试
execution_time = key_sim('a', 30, 60)
print("Total execution time: {:.2f} ms".format(execution_time*1000))

方案二：使用pynput（更高效） #

import time
import random
from pynput.keyboard import Controller

keyboard = Controller()

def key_sim(key, min_delay_ms, max_delay_ms):
    start_time = time.perf_counter()

    press_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    keyboard.press(key)
    target_time = time.perf_counter() + press_delay
    while time.perf_counter() < target_time:
        pass

    release_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    keyboard.release(key)
    target_time = time.perf_counter() + release_delay
    while time.perf_counter() < target_time:
        pass

    return time.perf_counter() - start_time

# 执行测试
execution_time = key_sim('a', 30, 60)
print("Total execution time: {:.2f} ms".format(execution_time*1000))

方案三：Windows平台直接调用系统API（极致优化） #

import time
import random
import ctypes

user32 = ctypes.WinDLL('user32', use_last_error=True)
KEYEVENTF_KEYUP = 0x0002
VK_A = 0x41  # 'A'键的虚拟键码

def keybd_event(vk_code, scan_code=0, flags=0, extra_info=0):
    user32.keybd_event(vk_code, scan_code, flags, extra_info)

def key_sim(min_delay_ms, max_delay_ms):
    start_time = time.perf_counter()

    press_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    keybd_event(VK_A)  # 按下按键
    target_time = time.perf_counter() + press_delay
    while time.perf_counter() < target_time:
        pass

    release_delay = random.randint(min_delay_ms, max_delay_ms) / 1000.0
    keybd_event(VK_A, 0, KEYEVENTF_KEYUP)  # 释放按键
    target_time = time.perf_counter() + release_delay
    while time.perf_counter() < target_time:
        pass

    return time.perf_counter() - start_time

# 执行测试
execution_time = key_sim(30, 60)
print("Total execution time: {:.2f} ms".format(execution_time*1000))

效果说明 #

优化后总执行时间会大幅接近两次延迟的总和，其中直接调用Windows系统API的版本额外开销最小，pynput版本次之，优化后的pyautogui版本也能显著缩小总时间与延迟总和的差值。

内容的提问来源于stack exchange，提问作者Камаль