1
def extract_json_by_stack(html_content):
2
    """
3
    核心算法：通过数大括号的方式提取核心数据
4
    解决正则匹配不完整导致的 JSON 解析错误
5
    """
6
    # 1. 定位变量起点
7
    start_marker = "var dataToReturn ="
8
    start_idx = html_content.find(start_marker)
9
    if start_idx == -1: return {}
10

11
    # 2. 找到第一个左括号
12
    cursor = html_content.find("{", start_idx)
13
    if cursor == -1: return {}
14

15
    # 3. 堆栈计数
16
    brace_count = 0
17
    json_str = ""
18
    found_end = False
19

20
    # 截取缓冲区，避免全页遍历
21
    buffer = html_content[cursor : cursor + 500000]
22

23
    for char in buffer:
24
        json_str += char
25
        if char == '{':
26
            brace_count += 1
27
        elif char == '}':
28
            brace_count -= 1
29
            # 闭合检测
30
            if brace_count == 0:
31
                found_end = True
32
                break
33

34
    if found_end:
35
        try:
36
            # 数据清洗与加载
37
            return json.loads(json_str)
38
        except json.JSONDecodeError:
39
            pass # 容错处理
40
    return {}

1
async def download_video_ffmpeg(url, path, semaphore):
2
    # 检查文件是否已存在且有效
3
    if os.path.exists(path) and os.path.getsize(path) > 10240:
4
        return
5

6
    async with semaphore:
7
        print(f"🎬 [FFmpeg启动] 正在下载: {os.path.basename(path)}")
8

9
        # 构造命令：-copy 模式不转码，速度极快
10
        # -bsf:a aac_adtstoasc 用于修复音频流格式
11
        cmd = [
12
            "ffmpeg", "-y", "-v", "error",
13
            "-protocol_whitelist", "file,http,https,tcp,tls,crypto",
14
            "-i", url,
15
            "-c", "copy",
16
            "-bsf:a", "aac_adtstoasc",
17
            path
18
        ]
19

20
        try:
21
            # 异步调用子进程，不阻塞主线程
22
            process = await asyncio.create_subprocess_exec(*cmd)
23
            await process.wait()
24
        except Exception as e:
25
            print(f"❌ 下载失败: {e}")

1
# 获取页面总高度
2
page_height = await page.evaluate("document.body.scrollHeight")
3
scroll_step = 200  # 每次滚动200像素
4
total_steps = max(1, int(page_height / scroll_step))  # 至少1步
5
scroll_duration = 5  # 总共5秒
6
delay_per_step = scroll_duration / total_steps  # 每步延迟时间
7

8
# 从上到下慢慢滚动
9
current_scroll = 0
10
while current_scroll < page_height:
11
    await page.evaluate(f"window.scrollTo(0, {current_scroll})")
12
    current_scroll += scroll_step
13
    await asyncio.sleep(delay_per_step)
14

15
# 确保滚动到底部
16
await page.evaluate("window.scrollTo(0, document.body.scrollHeight)")
17
await asyncio.sleep(1)
18

19
# 扫描视频容器（其实这一段代码没啥用，根本定位不到，但是我还是保留了，小伙伴们使用时可以删掉这个）
20
print("   -> 扫描视频板块...")
21
try:
22
    video_loc = page.locator("#vse-related-videos, .vse-related-videos-container, #aplus").first
23
    if await video_loc.count() > 0:
24
        await video_loc.scroll_into_view_if_needed()
25
        await asyncio.sleep(2)
26
except: pass