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 crawl_list(pg): """抓列表页, 返回 [{title, url, date}]""" if pg == 1: url = BASE + LIST_DIR + "index.shtml" else: url = BASE + LIST_DIR + f"index_{pg-1}.shtml" try: r = requests.get(url, headers=HEADERS, timeout=30) r.encoding = "utf-8" except Exception as e: log(f" [WARN] 第{pg}页请求失败: {e}") return [] soup = BeautifulSoup(r.text, "html.parser") items = [] for li in soup.find_all("li"): time_el = li.find("span", class_="time") p_el = li.find("p", class_="text") if not time_el or not p_el: continue a = p_el.find("a", href=True) if not a: continue href = a["href"].strip() if not href or href == "#" or "javascript" in href: continue if not href.startswith("http"): href = urljoin(BASE + LIST_DIR, href) title = a.get_text(strip=True) date = "" dm = re.search(r"(20\d{2}-\d{2}-\d{2})", time_el.get_text()) if dm: date = dm.group(1) items.append({"title": title, "url": href, "date": date}) return items def extract_detail(url): """返回 (title, date, content)""" try: r = requests.get(url, headers=HEADERS, timeout=30) r.encoding = "utf-8" html = r.text except Exception as e: log(f" [WARN] 详情失败 {url}: {e}") 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: h = soup.find("h1") if h: title = h.get_text(strip=True) # 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 content = "" dc = soup.find("div", class_="trs_editor_view") if not dc: dc = soup.find("div", class_="news-content") if dc: # 附件(正文内 a[appendix=true] 或附件扩展名) attachments = [] seen_att = set() for a in dc.find_all("a", href=True): href = a["href"].strip() if ATTACH_RE.search(href) or a.get("appendix") == "true" or a.get("data-appendix") == "true": # 优先 OLDSRC 绝对路径(相对路径可能失效) abs_url = a.get("OLDSRC") or href if not abs_url.startswith("http"): abs_url = urljoin(BASE, abs_url) name = a.get("title", "").strip() or a.get_text(strip=True) or "附件" if abs_url not in seen_att: seen_att.add(abs_url) attachments.append({"name": name, "url": abs_url}) parts = [] for node in dc.find_all(["p", "div", "table"], recursive=True): # 跳过表格内部节点(表格整体保留) if node.name != "table" and node.find_parent("table"): continue # 跳过含已提取附件的p(避免正文+附件区重复) if node.name == "p": has_att = False for a in node.find_all(["a"], href=True): h = a.get("href", "").strip() if h in seen_att or a.get("OLDSRC") in seen_att: has_att = True break if has_att: continue html_str = str(node).strip() if not html_str: continue # 去嵌套: 外层p包含表格则跳过; p内再套p只取最内层 if node.name == "p": if node.find("table"): continue inner_p = node.find("p", recursive=False) if inner_p: html_str = str(inner_p).strip() if node.name == "div" and node.find_parent("div") and not node.find(["p", "table"]): continue if node.name in ("p", "table"): parts.append(html_str) parts = _drop_container_parts(parts) content = "\n\n".join(parts).strip() # 去重 + 过滤空段落 uniq_parts = [] for part in content.split("\n\n"): part_stripped = part.strip() if not part_stripped: continue 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(re.sub(r"<[^>]+>", "", content).strip()) < 10: 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"