diff --git a/开心影院.py b/开心影院.py index 570ae74..efca0b1 100644 --- a/开心影院.py +++ b/开心影院.py @@ -247,7 +247,16 @@ class Spider(Spider): return {'list': []} html = rsp.text - data = pq(html) + + # ===== 清洗HTML,移除非法字符 ===== + html = html.replace('\x00', '') + html = re.sub(r'[\x01-\x08\x0b\x0c\x0e-\x1f]', '', html) + + try: + data = pq(html, parser='html') + except: + html = re.sub(r'[^\x09\x0a\x0d\x20-\x7e\x80-\xff]', '', html) + data = pq(html, parser='html') # ===== 标题 ===== name = data('h1').text().strip()