338 lines
10 KiB
Python
338 lines
10 KiB
Python
"""Playwright 页面抓取工具。
|
|
|
|
对外暴露三个动作：
|
|
- open_page(url, wait_ms): 打开页面，返回 page_id
|
|
- query(page_id, selector): 选中节点并提取文字、图片 URL 与属性
|
|
- close_page(page_id): 关闭页面，释放浏览器上下文
|
|
|
|
实现上内部常驻一个 Browser；每次 open_page 创建独立 BrowserContext 和 Page。
|
|
Playwright 用 async API 跑在一个后台事件循环线程里，多个调用方线程可以真正并发
|
|
（页面加载彼此重叠，不会互相排队）；外层 API 仍保持同步。
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import atexit
|
|
import threading
|
|
import uuid
|
|
from collections.abc import Coroutine
|
|
from dataclasses import dataclass
|
|
from typing import Any
|
|
|
|
from playwright.async_api import Browser, BrowserContext, Page, async_playwright
|
|
|
|
# 页面打开超时（毫秒）
|
|
_GOTO_TIMEOUT = 30000
|
|
|
|
# Chromium 启动参数：
|
|
# --no-sandbox 容器内以 root 运行时 Chrome 沙箱不可用，不加会直接启动失败
|
|
# --disable-dev-shm-usage 容器 /dev/shm 偏小时改用磁盘，避免渲染进程崩溃
|
|
_LAUNCH_ARGS = ["--no-sandbox", "--disable-dev-shm-usage"]
|
|
|
|
|
|
@dataclass
|
|
class NodeSnapshot:
|
|
"""页面节点快照。"""
|
|
|
|
tag: str
|
|
text: str
|
|
image_urls: list[str]
|
|
attrs: dict[str, str]
|
|
|
|
|
|
@dataclass
|
|
class _PageSession:
|
|
context: BrowserContext
|
|
page: Page
|
|
|
|
|
|
# 节点提取脚本：取标签名、可见文字、图片地址（含懒加载）与全部属性
|
|
_EXTRACT_JS = r"""(elements) => elements.map((el) => {
|
|
const attrs = {};
|
|
for (const attr of Array.from(el.attributes || [])) {
|
|
attrs[attr.name] = attr.value;
|
|
}
|
|
|
|
const imageUrls = [];
|
|
const addImageUrl = (url) => {
|
|
if (!url) return;
|
|
try {
|
|
const absUrl = new URL(url, location.href).href;
|
|
if (
|
|
absUrl.startsWith('data:') ||
|
|
absUrl.startsWith('blob:') ||
|
|
absUrl === 'about:blank'
|
|
) {
|
|
return;
|
|
}
|
|
imageUrls.push(absUrl);
|
|
} catch {
|
|
if (
|
|
url.startsWith('data:') ||
|
|
url.startsWith('blob:') ||
|
|
url === 'about:blank'
|
|
) {
|
|
return;
|
|
}
|
|
imageUrls.push(url);
|
|
}
|
|
};
|
|
|
|
const resolveImageUrl = (img) => {
|
|
const candidates = [
|
|
img.getAttribute('data-src'),
|
|
img.getAttribute('data-original'),
|
|
img.getAttribute('data-croporisrc'),
|
|
img.currentSrc,
|
|
img.src,
|
|
img.getAttribute('src'),
|
|
];
|
|
|
|
const srcset = img.getAttribute('srcset');
|
|
if (srcset) {
|
|
for (const part of srcset.split(',')) {
|
|
const candidate = part.trim().split(/\s+/)[0];
|
|
candidates.push(candidate);
|
|
}
|
|
}
|
|
|
|
for (const candidate of candidates) {
|
|
if (!candidate) continue;
|
|
if (
|
|
candidate.startsWith('data:') ||
|
|
candidate.startsWith('blob:') ||
|
|
candidate === 'about:blank'
|
|
) {
|
|
continue;
|
|
}
|
|
try {
|
|
const absUrl = new URL(candidate, location.href).href;
|
|
if (
|
|
absUrl.startsWith('data:') ||
|
|
absUrl.startsWith('blob:') ||
|
|
absUrl === 'about:blank'
|
|
) {
|
|
continue;
|
|
}
|
|
return absUrl;
|
|
} catch {
|
|
return candidate;
|
|
}
|
|
}
|
|
|
|
return null;
|
|
};
|
|
|
|
if (el.tagName === 'IMG') {
|
|
addImageUrl(resolveImageUrl(el));
|
|
}
|
|
|
|
for (const img of Array.from(el.querySelectorAll('img'))) {
|
|
addImageUrl(resolveImageUrl(img));
|
|
}
|
|
|
|
const text = (el.innerText || el.textContent || '').trim();
|
|
|
|
return {
|
|
tag: (el.tagName || '').toLowerCase(),
|
|
text,
|
|
image_urls: Array.from(new Set(imageUrls)),
|
|
attrs,
|
|
};
|
|
})"""
|
|
|
|
|
|
class _BrowserRuntime:
|
|
"""在后台事件循环线程中管理 Playwright 生命周期。
|
|
|
|
所有 Playwright 调用都以协程形式提交到同一个事件循环，
|
|
因此多个业务线程的页面操作可以并发交叠执行。
|
|
"""
|
|
|
|
def __init__(self) -> None:
|
|
self._thread: threading.Thread | None = None
|
|
self._loop: asyncio.AbstractEventLoop | None = None
|
|
self._launch_lock: asyncio.Lock | None = None
|
|
self._ready = threading.Event()
|
|
self._lock = threading.RLock()
|
|
self._playwright: Any = None
|
|
self._browser: Browser | None = None
|
|
self._pages: dict[str, _PageSession] = {}
|
|
atexit.register(self.close)
|
|
|
|
# ──────────── 事件循环线程 ────────────
|
|
|
|
def _thread_main(self) -> None:
|
|
loop = asyncio.new_event_loop()
|
|
asyncio.set_event_loop(loop)
|
|
self._loop = loop
|
|
# 在循环所属线程内创建，避免多协程各自新建锁
|
|
self._launch_lock = asyncio.Lock()
|
|
self._ready.set()
|
|
try:
|
|
loop.run_forever()
|
|
finally:
|
|
loop.close()
|
|
|
|
def _ensure_loop(self) -> asyncio.AbstractEventLoop:
|
|
with self._lock:
|
|
if self._thread is None or not self._thread.is_alive():
|
|
self._ready.clear()
|
|
self._thread = threading.Thread(
|
|
target=self._thread_main,
|
|
name="offerpie-playwright",
|
|
daemon=True,
|
|
)
|
|
self._thread.start()
|
|
self._ready.wait()
|
|
|
|
loop = self._loop
|
|
if loop is None:
|
|
raise RuntimeError("Browser runtime loop is not ready.")
|
|
return loop
|
|
|
|
def _submit(self, coro: Coroutine[Any, Any, Any]) -> Any:
|
|
"""把协程提交到事件循环并同步等待结果。"""
|
|
loop = self._ensure_loop()
|
|
return asyncio.run_coroutine_threadsafe(coro, loop).result()
|
|
|
|
# ──────────── Playwright 操作 ────────────
|
|
|
|
async def _ensure_browser(self) -> Browser:
|
|
if self._browser is not None:
|
|
return self._browser
|
|
|
|
assert self._launch_lock is not None
|
|
async with self._launch_lock:
|
|
if self._browser is None:
|
|
self._playwright = await async_playwright().start()
|
|
self._browser = await self._playwright.chromium.launch(
|
|
headless=True, args=_LAUNCH_ARGS
|
|
)
|
|
return self._browser
|
|
|
|
async def _open_page_impl(self, url: str, wait_ms: int) -> str:
|
|
browser = await self._ensure_browser()
|
|
context = await browser.new_context()
|
|
page = await context.new_page()
|
|
|
|
try:
|
|
await page.goto(url, wait_until="domcontentloaded", timeout=_GOTO_TIMEOUT)
|
|
if wait_ms > 0:
|
|
await page.wait_for_timeout(wait_ms)
|
|
except BaseException:
|
|
# 打开失败也要释放上下文，避免泄漏
|
|
await self._dispose(context, page)
|
|
raise
|
|
|
|
page_id = uuid.uuid4().hex
|
|
self._pages[page_id] = _PageSession(context=context, page=page)
|
|
return page_id
|
|
|
|
async def _query_impl(self, page_id: str, selector: str) -> list[NodeSnapshot]:
|
|
session = self._pages.get(page_id)
|
|
if session is None:
|
|
raise KeyError(f"Unknown page_id: {page_id}")
|
|
|
|
locator = session.page.locator(selector)
|
|
payload = await locator.evaluate_all(_EXTRACT_JS)
|
|
return [
|
|
NodeSnapshot(
|
|
tag=item.get("tag", ""),
|
|
text=item.get("text", ""),
|
|
image_urls=list(item.get("image_urls", [])),
|
|
attrs=dict(item.get("attrs", {})),
|
|
)
|
|
for item in payload
|
|
]
|
|
|
|
@staticmethod
|
|
async def _dispose(context: BrowserContext, page: Page | None) -> None:
|
|
if page is not None:
|
|
try:
|
|
await page.close()
|
|
except Exception:
|
|
pass
|
|
try:
|
|
await context.close()
|
|
except Exception:
|
|
pass
|
|
|
|
async def _close_page_impl(self, page_id: str) -> None:
|
|
session = self._pages.pop(page_id, None)
|
|
if session is None:
|
|
return
|
|
await self._dispose(session.context, session.page)
|
|
|
|
async def _shutdown_impl(self) -> None:
|
|
for page_id in list(self._pages):
|
|
await self._close_page_impl(page_id)
|
|
|
|
if self._browser is not None:
|
|
try:
|
|
await self._browser.close()
|
|
except Exception:
|
|
pass
|
|
self._browser = None
|
|
|
|
if self._playwright is not None:
|
|
try:
|
|
await self._playwright.stop()
|
|
except Exception:
|
|
pass
|
|
self._playwright = None
|
|
|
|
# ──────────── 生命周期 ────────────
|
|
|
|
def close(self) -> None:
|
|
"""关闭全部页面与浏览器，停止事件循环。"""
|
|
with self._lock:
|
|
thread = self._thread
|
|
loop = self._loop
|
|
if thread is None or not thread.is_alive() or loop is None:
|
|
return
|
|
|
|
try:
|
|
asyncio.run_coroutine_threadsafe(
|
|
self._shutdown_impl(), loop
|
|
).result(timeout=15)
|
|
except Exception:
|
|
pass
|
|
|
|
loop.call_soon_threadsafe(loop.stop)
|
|
thread.join(timeout=10)
|
|
self._thread = None
|
|
self._loop = None
|
|
self._launch_lock = None
|
|
|
|
|
|
_runtime = _BrowserRuntime()
|
|
|
|
|
|
def open_page(url: str, wait_ms: int = 8000) -> str:
|
|
"""打开页面并返回 page_id。"""
|
|
return _runtime._submit(_runtime._open_page_impl(url, wait_ms))
|
|
|
|
|
|
def query(page_id: str, selector: str) -> list[NodeSnapshot]:
|
|
"""按选择器抓取节点，提取文本、图片 URL 与属性。"""
|
|
items = _runtime._submit(_runtime._query_impl(page_id, selector))
|
|
seen_urls: set[str] = set()
|
|
|
|
for item in items:
|
|
unique_urls: list[str] = []
|
|
for url in item.image_urls:
|
|
if url in seen_urls:
|
|
continue
|
|
seen_urls.add(url)
|
|
unique_urls.append(url)
|
|
item.image_urls = unique_urls
|
|
|
|
return items
|
|
|
|
|
|
def close_page(page_id: str) -> None:
|
|
"""关闭页面，释放对应的浏览器上下文。"""
|
|
_runtime._submit(_runtime._close_page_impl(page_id))
|