与它内部的

，导致同一内容重复（容器那份常还被 get_text(strip=True) 拍平）。判据：某段被列表内另一段完全包含 → 它是容器段 → 剔除。保护：剔除后为空则原样返回。 """ if not parts: return parts ps = [p for p in parts if isinstance(p, str)] if len(ps) != len(parts): return parts # ① 先按值去重 —— 否则两段完全相同时 a in b 与 b in a 双向成立， # 会把两段都判成「容器段」剔光，再触发下面的保护而原样返回（等于没修）。 seen, uniq = set(), [] for p in parts: if p not in seen: seen.add(p); uniq.append(p) # ② 再剔容器段（被其它段完全包含的那个） keep = [a for a in uniq if not (len(a) >= 40 and any(b is not a and b and b in a for b in uniq))] return keep if keep else uniq def log(msg): print(msg, file=sys.stderr, flush=True) def fetch(url, retries=3): for attempt in range(retries): try: r = requests.get(url, headers=HEADERS, timeout=30) r.encoding = "utf-8" return r.text except Exception as e: log(f" [WARN] {url} 请求失败({attempt+1}/{retries}): {e}") time.sleep(2) return "" def crawl_list(pg): """抓列表页, 返回 [{title, url, date}]""" if pg <= 1: url = LIST_DIR + "/index.html" else: url = f"{LIST_DIR}/pages/{pg}.html" html = fetch(url) if not html: return [] soup = BeautifulSoup(html, "html.parser") info = soup.find("div", class_="submiInfor") if not info: return [] items = [] for li in info.find_all("li"): a = li.find("a") if not a: continue href = a.get("href", "") if not href.startswith("http"): href = urljoin(BASE, href) title = a.get_text(strip=True) # strip 实体前缀 title = re.sub(r"^(?:·|·| |\s)+", "", title).strip() # 日期 dm = re.search(r"(20\d{2}-\d{2}-\d{2})", li.get_text()) date = dm.group(1) if dm else "" items.append({"title": title, "url": href, "date": date}) return items def extract_detail(url): """返回 (title, date, content)""" html = fetch(url) if not html: return None, None, None soup = BeautifulSoup(html, "html.parser") # 标题 title = "" mt = soup.find("meta", attrs={"name": "ArticleTitle"}) if mt and mt.get("content"): title = mt["content"].strip() if not title: sc = soup.find("div", class_="subjectCen") if sc: # 跳过面包屑, 取标题行 for line in sc.stripped_strings: if "首页" not in line and "通知公告" not in line: title = line.strip() break # 日期 date = "" md = soup.find("meta", attrs={"name": "PubDate"}) if md and md.get("content"): m = re.search(r"(20\d{2}-\d{2}-\d{2})", md["content"]) if m: date = m.group(1) # 正文: div#div_content content = "" dc = soup.find("div", id="div_content") if dc: # 附件: a[href] 指向文件 + span.edui-pdf data-pdf attachments = [] seen_att = set() for a in dc.find_all("a", href=True): href = a["href"] if re.search(r"\.(pdf|doc|docx|xls|xlsx|zip|rar|wps|txt)$", href, re.I): if not href.startswith("http"): href = urljoin(BASE, href) name = a.get_text(strip=True) or "附件" if href not in seen_att: seen_att.add(href) attachments.append({"name": name, "url": href}) for sp in dc.find_all("span", class_="edui-pdf"): pdf = sp.get("data-pdf", "") if pdf: if not pdf.startswith("http"): pdf = urljoin(BASE, pdf) name = sp.get_text(strip=True) or "附件" if pdf not in seen_att: seen_att.add(pdf) attachments.append({"name": name, "url": pdf}) parts = [] # 排除相关阅读/二维码/分享区块 + 附件下载重复区(filelink) for node in dc.find_all(["p", "div", "table"], recursive=True): # 关键: 跳过表格内部节点(表格已整体保留, 内部 p/div 不拍平, 避免重复) if node.name != "table" and node.find_parent("table"): continue if node.name == "div" and node.get("id") in ("xgwdP", "qrcode"): continue if node.find_parent("div", id="xgwdP"): continue # 跳过 filelink 附件下载重复区 if node.find_parent(class_="filelink") or (node.get("class") and "filelink" in node.get("class")): continue # 跳过含已提取附件的p(避免正文+附件区重复) if node.name == "p": def _att_url(el): """提取元素的文件URL(相对+绝对归一化)""" u = "" if el.name == "a": u = el.get("href", "") elif el.name == "span" and "edui-pdf" in (el.get("class") or []): u = el.get("data-pdf", "") if u and not u.startswith("http"): u = urljoin(BASE, u) return u has_att = any( _att_url(el) in seen_att for el in node.find_all(["a", "span"]) if _att_url(el)) if has_att: continue html_str = str(node).strip() if not html_str: continue # 去嵌套: p 内再套 p 只取最内层文本段落 if node.name == "p": # 外层p包含表格则跳过(表格内部节点独立处理, 避免表格内容被p包裹提取) if node.find("table"): continue inner_p = node.find("p", recursive=False) if inner_p: # 保留内层 p 的HTML, 去掉外层包装 html_str = str(inner_p).strip() if node.name == "div" and node.find_parent("div") and not node.find(["p", "table"]): parts.append(html_str) elif node.name in ("p", "table"): parts.append(html_str) parts = _drop_container_parts(parts) content = "\n\n".join(parts).strip() # 去重: 相同段落只保留一次 + 过滤空/纯
段落 uniq_parts = [] for part in content.split("\n\n"): part_stripped = part.strip() if not part_stripped: continue # 剥标签后无文本(仅空白/br) → 过滤 if not re.sub(r"<[^>]+>", "", part_stripped).strip(): continue if part_stripped not in uniq_parts: uniq_parts.append(part_stripped) content = "\n\n".join(uniq_parts).strip() if attachments: att_links = "\n".join(f'

' for att in attachments) content = content + "\n\n" + att_links if content else att_links return title, date, content def main(): import argparse ap = argparse.ArgumentParser() ap.add_argument("--pages", type=int, default=1) args = ap.parse_args() pages = max(1, min(args.pages, MAX_PAGES)) conn = sqlite3.connect(DB, timeout=15) conn.execute("PRAGMA busy_timeout=10000") c = conn.cursor() c.execute("CREATE TABLE IF NOT EXISTS gov_raw (id INTEGER PRIMARY KEY, site_name TEXT, source_url TEXT, page_url TEXT, title TEXT, publish_date TEXT, date_rank INTEGER DEFAULT 0, summary TEXT, status TEXT, category TEXT DEFAULT '', visits INTEGER DEFAULT 0, content TEXT DEFAULT '', tags TEXT DEFAULT '', industry TEXT DEFAULT 'other', attachments TEXT DEFAULT '', group_name TEXT, has_table INTEGER DEFAULT 0, script_name TEXT DEFAULT '')") c.execute("CREATE VIRTUAL TABLE IF NOT EXISTS gov_search USING fts5(title, site_name, summary, tokenize='trigram')") total_new = total_dup = total_skip = 0 seen = set() for pg in range(1, pages + 1): log(f"--- 第{pg}页 ---") items = crawl_list(pg) if not items: log("列表为空, 停止翻页") break log(f"找到{len(items)}条") for item in items: if item["url"] in seen: continue seen.add(item["url"]) c.execute("SELECT id FROM gov_raw WHERE page_url=?", (item["url"],)) if c.fetchone(): total_dup += 1 continue log(f" [{item['date']}] {item['title'][:45]}") title, date, content = extract_detail(item["url"]) if not title: title = item["title"] if not date: date = item["date"] if not content or len(content) < 50: log(f" 正文过短({len(content) if content else 0}), 跳过") total_skip += 1 continue summary = re.sub(r"<[^>]+>", "", content)[:200] summary = re.sub(r"\s+", " ", summary).strip() has_table = 1 if re.search(r"