|
| 1 | +import asyncio |
| 2 | +from playwright.async_api import async_playwright |
| 3 | +from urllib.parse import urljoin, urldefrag |
| 4 | +from collections import deque |
| 5 | +import aiofiles |
| 6 | + |
| 7 | +BASE_URL = "http://localhost:8000" |
| 8 | +LOG_FILE = "visited.log" |
| 9 | +BROKEN_FILE = "broken-links.txt" |
| 10 | +CONTENTS_FILE = "contents.txt" |
| 11 | + |
| 12 | + |
| 13 | +async def check_page(page, source_url): |
| 14 | + url = page.url |
| 15 | + html = await page.content() |
| 16 | + |
| 17 | + # Save page content |
| 18 | + async with aiofiles.open(CONTENTS_FILE, "a") as f: |
| 19 | + await f.write(html) |
| 20 | + |
| 21 | + # Check for 404 |
| 22 | + if "<h1>404 - Not found</h1>" in html: |
| 23 | + async with aiofiles.open(BROKEN_FILE, "a") as f: |
| 24 | + await f.write(f"[404] {url} (linked from {source_url})\n") |
| 25 | + |
| 26 | + # Broken image sources |
| 27 | + bad_imgs = await page.eval_on_selector_all( |
| 28 | + "img[src]", |
| 29 | + """els => els |
| 30 | + .map(e => ({ src: e.src, complete: e.complete, naturalWidth: e.naturalWidth })) |
| 31 | + .filter(e => !e.complete || e.naturalWidth === 0) |
| 32 | + .map(e => e.src)""", |
| 33 | + ) |
| 34 | + if bad_imgs: |
| 35 | + async with aiofiles.open(BROKEN_FILE, "a") as f: |
| 36 | + for img in bad_imgs: |
| 37 | + await f.write(f"[Broken IMG] {img} on {url}\n") |
| 38 | + |
| 39 | + # Check anchor hrefs |
| 40 | + anchors = await page.eval_on_selector_all("a[href]", "els => els.map(e => e.href)") |
| 41 | + for link in anchors: |
| 42 | + link_url, _ = urldefrag(urljoin(url, link)) |
| 43 | + if not link_url.startswith(BASE_URL): |
| 44 | + continue |
| 45 | + try: |
| 46 | + res = await page.context.request.get(link_url) |
| 47 | + if res.status >= 400: |
| 48 | + async with aiofiles.open(BROKEN_FILE, "a") as f: |
| 49 | + await f.write( |
| 50 | + f"[Broken LINK] {link_url} on {url} (status {res.status})\n" |
| 51 | + ) |
| 52 | + except Exception as e: |
| 53 | + async with aiofiles.open(BROKEN_FILE, "a") as f: |
| 54 | + await f.write(f"[Broken LINK] {link_url} on {url} (error {e})\n") |
| 55 | + |
| 56 | + |
| 57 | +async def crawl(base_url): |
| 58 | + visited = set() |
| 59 | + queue = deque([(base_url, "(start)")]) |
| 60 | + |
| 61 | + async with async_playwright() as p: |
| 62 | + browser = await p.chromium.launch() |
| 63 | + context = await browser.new_context() |
| 64 | + page = await context.new_page() |
| 65 | + |
| 66 | + async with aiofiles.open(LOG_FILE, "w") as log_file: |
| 67 | + while queue: |
| 68 | + raw_url, source_url = queue.popleft() |
| 69 | + url, _ = urldefrag(raw_url) |
| 70 | + |
| 71 | + if url in visited or not url.startswith(base_url): |
| 72 | + continue |
| 73 | + visited.add(url) |
| 74 | + |
| 75 | + try: |
| 76 | + await page.goto(url) |
| 77 | + await check_page(page, source_url) |
| 78 | + await log_file.write(url + "\n") |
| 79 | + await log_file.flush() |
| 80 | + |
| 81 | + links = await page.eval_on_selector_all( |
| 82 | + "a[href]", "els => els.map(e => e.href)" |
| 83 | + ) |
| 84 | + for link in links: |
| 85 | + abs_link, _ = urldefrag(urljoin(url, link)) |
| 86 | + if abs_link.startswith(base_url) and abs_link not in visited: |
| 87 | + queue.append((abs_link, url)) |
| 88 | + |
| 89 | + except Exception as e: |
| 90 | + print(f"Error visiting {url}: {e}") |
| 91 | + |
| 92 | + await browser.close() |
| 93 | + |
| 94 | + |
| 95 | +if __name__ == "__main__": |
| 96 | + asyncio.run(crawl(BASE_URL)) |
0 commit comments