#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
crawl_dafeng_sthj.py - 盐城市大丰区人民政府 (www.dafeng.gov.cn)
https://www.dafeng.gov.cn/col/col25839/index.html (生态环境)
CMS: Hanweb 大汉网络 jpage dataproxy (GET)
列表: GET /module/web/jpage/dataproxy.jsp?page=N&col=1&appid=1&webid=52&path=/
        &columnid=25839&sourceContentType=1&unitid=66892&webname=盐城市大丰区人民政府&permissiontype=0
  响应 XML: <datastore><totalrecord>253</totalrecord>...
            <record><![CDATA[<li><a href="/art/2026/9/11/art_25839_ID.html"
              title="完整标题">显示标题</a><span>2026-09-11</span></li>]]></record>
  ⚠️ page=1 即返回全量 253 条（page2/3 是子集，按 URL 去重）；totalpage=3 虚报
详情: /art/YYYY/M/D/art_{columnid}_{id}.html
  meta ArticleTitle(完整标题) / PubDate(带时分) / ContentSource(来源)
  正文: <meta name="ContentStart"> … <meta name="ContentEnd"> 区间
        （落在 div.main-txt#zoom 内；表格 4 个为正文项目情况表，需保留 HTML）
附件: /module/download/downfile.jsp?classid=0&filename=<hash>.pdf
        <a href="..."><img src="/module/jslib/icons/acrobat.png"/>中文文件名.pdf</a>
        图标 img 随 <a> 替换自动丢弃；附件名取 a 内文本

与 crawl_dongtai_hjsp.py 同 CMS（东台同属盐城市），差异:
  ① 列表 span 是完整 YYYY-MM-DD（东台为 MM-DD）
  ② 列表 a 带 title 属性（完整标题，天然免截断）
  ③ 正文用 ContentStart/ContentEnd 标记（东台为 div.detail_texts div.content）
  ④ 无加速乐 WAF（详情可带 Referer）
"""
import sys, os, re, time, json, html as html_lib, urllib.parse
from datetime import datetime, timedelta

try:
    from bs4 import BeautifulSoup
except ImportError:
    BeautifulSoup = None

JSONL_PATH = os.environ.get("JSONL_PATH", "")
ENC = "utf-8"

SITE_NAME = "大丰区-生态环境"
GROUP_NAME = "江苏"
SCRIPT_NAME = "crawl_dafeng_sthj.py"
CATEGORY = "生态环境"
BASE_URL = "https://www.dafeng.gov.cn"
LIST_URL = "https://www.dafeng.gov.cn/col/col25839/index.html"
PROXY_URL = "https://www.dafeng.gov.cn/module/web/jpage/dataproxy.jsp"
JQ_PARAMS = {
    "col": "1",
    "appid": "1",
    "webid": "52",
    "path": "/",
    "columnid": "25839",
    "sourceContentType": "1",
    "unitid": "66892",
    "webname": "盐城市大丰区人民政府",
    "permissiontype": "0",
}
# 3 年截断（用户标准：CUTOFF 3 年）
CUTOFF = (datetime.now() - timedelta(days=365 * 3)).strftime("%Y-%m-%d")

HEADERS = {
    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
                  "(KHTML, like Gecko) Chrome/120.0 Safari/537.36",
    "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
    "Accept-Language": "zh-CN,zh;q=0.9",
}

_MAX_PAGES = 1
for i, a in enumerate(sys.argv):
    if a.startswith("--pages="):
        try:
            _MAX_PAGES = int(a.split("=", 1)[1])
        except ValueError:
            pass
    elif a == "--pages" and i + 1 < len(sys.argv):
        try:
            _MAX_PAGES = int(sys.argv[i + 1])
        except ValueError:
            pass
_PAGES = _MAX_PAGES if _MAX_PAGES >= 1 else 1


def clean_title(t):
    if not t:
        return ""
    t = html_lib.unescape(t)
    t = t.replace("\u200b", "").replace("\u200c", "").replace("\u200d", "").replace("\ufeff", "")
    t = re.sub(r"^\s*(?:&middot;|·|\u00b7)?\s*(?:&nbsp;|\u00a0)?\s*", "", t)
    t = re.sub(r"^[•·]\s*", "", t)
    t = re.sub(r"\s+", " ", t)
    return t.strip()


def fetch(url, timeout=30):
    h = dict(HEADERS)
    last_err = None
    for attempt in range(3):
        try:
            import requests, urllib3
            urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
            r = requests.get(url, headers=h, timeout=timeout, verify=False)
            if r.status_code == 200:
                r.encoding = ENC
                return r.text
            last_err = "HTTP %s" % r.status_code
        except Exception as e:
            last_err = str(e)[:80]
        time.sleep(2)
    print("  [fetch fail] %s -> %s" % (url[:90], last_err), flush=True)
    return None


def html_to_text(content, page_url):
    """正文 HTML 清洗：保留 <p>/<table>，附件转 <p><a>绝对URL</a></p> 段落"""
    if not content:
        return "", 0, []
    content = re.sub(r"<!--.*?-->", "", content, flags=re.S)
    content = re.sub(r"<meta[^>]*>", "", content, flags=re.I)
    content = re.sub(r"<script[^>]*>.*?</script>", "", content, flags=re.S | re.I)
    content = re.sub(r"<style[^>]*>.*?</style>", "", content, flags=re.S | re.I)

    attachments = []
    link_protect = {}

    content = re.sub(r"【字号：[^】]*】", "", content)

    # ① 附件/图片 <a> 保护（在表格保护之前，防表格正则吞掉紧跟的附件链接）
    if BeautifulSoup:
        try:
            _soup = BeautifulSoup(content, "html.parser")
            for a in _soup.find_all("a", href=True):
                href = a["href"]
                if (a.get("appendix") or a.get("data-appendix")
                        or re.search(r"\.(pdf|docx?|xlsx?|zip|rar|wps|et|ofd)", href, re.I)
                        or "/module/download/" in href):
                    real_href = a.get("oldsrc") or href
                    txt = (a.get_text(strip=True) or a.get("_title") or a.get("title")
                           or os.path.basename(real_href.split("?")[0]) or "附件")
                    abs_url = urllib.parse.urljoin(page_url, real_href)
                    attachments.append((abs_url, txt))
                    key = "__ATTACH__%d__" % len(link_protect)
                    link_protect[key] = '<a href="%s" target="_blank">%s</a>' % (abs_url, txt)
                    a.replace_with(key)
            # 附件图标 img（acrobat.png / fileTypeImages/icon_*.gif）清理
            for img in _soup.find_all("img"):
                src = (img.get("src") or img.get("oldsrc") or "")
                if re.search(r"(icons?/|fileTypeImages/icon_)", src, re.I):
                    img.decompose()
            # 纯图片段落 → 图片链接段
            for p in _soup.find_all("p"):
                imgs = p.find_all("img")
                if imgs and not p.get_text(strip=True):
                    srcs = []
                    for img in imgs:
                        src = img.get("src") or img.get("oldsrc") or ""
                        if not src:
                            continue
                        abs_url = urllib.parse.urljoin(page_url, src)
                        txt = (img.get("alt") or img.get("title") or "").strip() \
                            or os.path.basename(src.split("?")[0]) or "图片"
                        srcs.append('<a href="%s" target="_blank">%s</a>' % (abs_url, txt))
                    if srcs:
                        key = "__IMG__%d__" % len(link_protect)
                        link_protect[key] = "<p>" + "<br/>".join(srcs) + "</p>"
                        p.replace_with(key)
            # 其余裸 img（正文插图）→ <p><a>图</a></p>
            for img in _soup.find_all("img"):
                src = img.get("src") or img.get("oldsrc") or ""
                if not src:
                    continue
                abs_url = urllib.parse.urljoin(page_url, src)
                txt = (img.get("alt") or img.get("title") or "").strip() \
                    or os.path.basename(src.split("?")[0]) or "图片"
                key = "__IMG__%d__" % len(link_protect)
                link_protect[key] = '<p><a href="%s" target="_blank">%s</a></p>' % (abs_url, txt)
                img.replace_with(key)
            content = str(_soup)
        except Exception:
            pass

    # ② 残余 <a>（相对路径/无属性）统一绝对化并入 attachments
    #    ⚠️ mailto:/tel:/javascript:/#锚点 **不是附件** — 直接原样保留（正文里「邮箱：xxx@163.com」
    #    这类链接若当附件处理会污染 attachments 列 + 生成假附件段，122/248 行中招）
    def _link_repl(m):
        href = m.group(1)
        if re.match(r"^(?:javascript:|mailto:|tel:|s?ftp:|#)", href, re.I):
            return "" if href.lower().startswith("javascript:") else m.group(0)
        inner = m.group(2)
        txt = re.sub(r"<[^>]+>", "", inner)
        txt = html_lib.unescape(txt).strip()
        txt = re.sub(r"^附件[:：]\s*", "", txt).strip()
        if not txt:
            txt = os.path.basename(href.split("?")[0]) or "附件"
        abs_url = urllib.parse.urljoin(page_url, href)
        attachments.append((abs_url, txt))
        key = "__LINK__%d__" % len(link_protect)
        link_protect[key] = '<a href="%s" target="_blank">%s</a>' % (abs_url, txt)
        return key
    content = re.sub(r'<a\s[^>]*href="([^"]+)"[^>]*>(.*?)</a>', _link_repl,
                     content, flags=re.S | re.I)

    # ③ 清属性 + 解 div / 删空 div
    if BeautifulSoup:
        try:
            _soup = BeautifulSoup(content, "html.parser")
            for tag in _soup.find_all(True):
                for attr in ("style", "class", "lang", "dir", "align", "valign", "width",
                             "height", "border", "cellpadding", "cellspacing"):
                    tag.attrs.pop(attr, None)
            for d in _soup.find_all("div"):
                if not d.get_text(strip=True) and not d.find("table") \
                        and not d.find("img") and not d.find("a"):
                    d.decompose()
                else:
                    d.unwrap()
            content = str(_soup)
        except Exception:
            pass

    # ④ 表格保护（支持嵌套：由内向外剥，占位符还原时天然嵌套正确）
    table_protect = []

    def _tbl_repl(m):
        table_protect.append(m.group(0))
        return "__TBL__%d__" % (len(table_protect) - 1)

    has_table = 1 if re.search(r"<table[^>]*>", content, re.I) else 0
    prev = None
    while prev != content:
        prev = content
        content = re.sub(r"<table(?:(?!<table).)*?</table>", _tbl_repl,
                         content, flags=re.S | re.I)

    # ⑤ 分段
    content = re.sub(r"</p>", "</p>\n\n", content, flags=re.I)
    content = re.sub(r"<br\s*/?>", "\n", content, flags=re.I)
    content = re.sub(r"</?(?:span|font|o:p|st1?:[a-z]+|strong|em|b|i|u)\b[^>]*>", "",
                     content, flags=re.I)

    parts = []
    for block in content.split("\n\n"):
        b = block.strip()
        if not b:
            continue
        b = re.sub(r"(?<=>)\s*[\r\n\t]+\s*", "", b)
        b = re.sub(r"\s*[\r\n\t]+\s*(?=<)", "", b)
        b = b.replace("\r", "").replace("\t", " ")
        b = re.sub(r"[ \t]{2,}", " ", b)
        plain = re.sub(r"<[^>]+>", "", b)
        plain = html_lib.unescape(plain)
        plain = plain.replace("\xa0", " ").replace("&nbsp;", " ")
        plain = re.sub(r"[\s\u200b\u200c\u200d\ufeff]+", "", plain)
        if not plain:
            continue
        for k, v in link_protect.items():
            b = b.replace(k, v)
        for i, tbl in enumerate(table_protect):
            b = b.replace("__TBL__%d__" % i, tbl)
        # 裸链接块 → 包 <p>
        if not re.search(r"<p[ >]", b, re.I) and re.search(r"<a[ >]", b, re.I):
            b = "<p>%s</p>" % b
        parts.append(b.strip())

    out = "\n\n".join(parts)
    for k, v in link_protect.items():
        out = out.replace(k, v)
    for i, tbl in enumerate(table_protect):
        out = out.replace("__TBL__%d__" % i, tbl)

    out = re.sub(r"__(?:LINK|TBL|IMG|ATTACH)__\d+__", "", out)
    out = html_lib.unescape(out)
    out = re.sub(r"[ \t]+\n", "\n", out)
    out = re.sub(r"\n{3,}", "\n\n", out)
    out = re.sub(r"[ \t]{2,}", " ", out)

    # 段落去重
    seen, final = set(), []
    for p in out.split("\n\n"):
        key = re.sub(r"\s+", "", re.sub(r"<[^>]+>", "", p))
        if not key or key in seen:
            continue
        seen.add(key)
        final.append(p)
    out = "\n\n".join(final).strip()

    # 附件去重（同 URL 只留一次）
    dedup, seen_u = [], set()
    for u, t in attachments:
        if u in seen_u:
            continue
        seen_u.add(u)
        dedup.append((u, t))

    return out, has_table, dedup


def fetch_list(page):
    import requests
    params = dict(JQ_PARAMS)
    params["page"] = str(page)
    r = requests.get(PROXY_URL, params=params, headers=HEADERS, timeout=30, verify=False)
    r.encoding = ENC
    return r.text


def parse_list(xml_text):
    """<record><![CDATA[<li><a href="/art/..." title="完整标题">…</a><span>YYYY-MM-DD</span></li>]]></record>"""
    items = []
    for m in re.finditer(r"<!\[CDATA\[(.*?)\]\]>", xml_text, re.S):
        frag = m.group(1)
        dm = re.search(r"<span[^>]*>\s*(\d{4}-\d{1,2}-\d{1,2})\s*</span>", frag)
        date = ""
        if dm:
            y, mo, d = re.split(r"-", dm.group(1))
            date = "%s-%s-%s" % (y, mo.zfill(2), d.zfill(2))
        for lm in re.finditer(r'<a[^>]*href="([^"]+)"[^>]*>(.*?)</a>', frag, re.S | re.I):
            href = html_lib.unescape(lm.group(1)).strip()
            tag = lm.group(0)
            tm = re.search(r'title="([^"]*)"', tag)
            txt = clean_title(tm.group(1)) if tm else ""
            if not txt:
                txt = clean_title(re.sub(r"<[^>]+>", "", lm.group(2)))
            # ⚠️ 阈值 2 非 4：「声 明」这类 3 字符短标题是真实公告，用 4 会静默漏抓
            if not txt or len(txt) < 2:
                continue
            if not re.search(r"/art/\d{4}/\d+/\d+/art_\d+_\d+\.html", href):
                continue
            if href.startswith("/"):
                href = BASE_URL + href
            elif not href.startswith("http"):
                href = urllib.parse.urljoin(BASE_URL + "/", href)
            if not date:
                ym = re.search(r"/art/(\d{4})/(\d+)/(\d+)/", href)
                if ym:
                    date = "%s-%s-%s" % (ym.group(1), ym.group(2).zfill(2), ym.group(3).zfill(2))
            items.append((href, txt, date))
    return items


def parse_detail(html_text, page_url):
    title = date = content_html = ""
    soup = None
    if BeautifulSoup:
        try:
            soup = BeautifulSoup(html_text, "html.parser")
            mt = soup.find("meta", attrs={"name": re.compile(r"ArticleTitle", re.I)})
            if mt and mt.get("content"):
                title = clean_title(mt["content"])
            md = soup.find("meta", attrs={"name": re.compile(r"PubDate", re.I)})
            if md and md.get("content"):
                dm = re.search(r"(\d{4})-(\d{1,2})-(\d{1,2})", md["content"])
                if dm:
                    date = "%s-%s-%s" % (dm.group(1), dm.group(2).zfill(2), dm.group(3).zfill(2))
            if not title:
                h1 = soup.find("h1")
                if h1:
                    title = clean_title(h1.get_text(strip=True))
        except Exception:
            soup = None
    if not title:
        m = re.search(r"<title>(.*?)</title>", html_text, re.S | re.I)
        if m:
            t = clean_title(m.group(1))
            # "盐城市大丰区人民政府 生态环境 文章标题" → 取最后一段
            parts = [p for p in re.split(r"\s+", t) if p]
            title = parts[-1] if parts else t

    # 正文: ContentStart … ContentEnd 标记区间
    m = re.search(r'<meta[^>]*name=["\']?ContentStart["\']?[^>]*>(.*?)'
                  r'<meta[^>]*name=["\']?ContentEnd["\']?[^>]*>', html_text, re.S | re.I)
    if m:
        content_html = m.group(1)
    else:
        m2 = re.search(r"<!--ZJEG_RSS\.content\.begin-->(.*?)<!--ZJEG_RSS\.content\.end-->",
                       html_text, re.S | re.I)
        if m2:
            content_html = m2.group(1)
        elif soup:
            el = soup.find(id="zoom") or soup.select_one("div.main-txt")
            if el:
                content_html = "".join(str(c) for c in el.contents)

    if not date:
        dm = re.search(r"(\d{4})[-年/](\d{1,2})[-月/](\d{1,2})", html_text)
        if dm:
            date = "%s-%s-%s" % (dm.group(1), dm.group(2).zfill(2), dm.group(3).zfill(2))
    return title, date, content_html


def main():
    import requests
    sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
    from crawler_lib import push_to_searchdb

    print("=== %s (%s) pages=%d cutoff=%s ===" % (SITE_NAME, SCRIPT_NAME, _PAGES, CUTOFF),
          flush=True)

    new_count = skip_count = cutoff_count = 0
    seen_urls = set()
    jsonl_rows = []
    batch = []

    for page in range(1, _PAGES + 1):
        try:
            xml_text = fetch_list(page)
        except Exception as e:
            print("Page %d fetch error: %s" % (page, e), flush=True)
            break
        items = parse_list(xml_text)
        if not items:
            print("Page %d: empty, stop" % page, flush=True)
            break
        print("Page %d: found %d items" % (page, len(items)), flush=True)

        for abs_url, title, date in items:
            if abs_url in seen_urls:
                continue
            seen_urls.add(abs_url)
            if date and date < CUTOFF:
                cutoff_count += 1
                continue
            dhtml = fetch(abs_url)
            if not dhtml:
                skip_count += 1
                continue
            dtitle, ddate, content_html = parse_detail(dhtml, abs_url)
            if not dtitle:
                dtitle = title
            if not ddate:
                ddate = date
            content, has_table, atts = html_to_text(content_html, abs_url)
            plain_len = len(re.sub(r"<[^>]+>", "", content).strip())
            no_para = (not re.search(r"<p[ >]", content, re.I)
                       and not re.search(r"<a[ >]", content, re.I) and not has_table)
            if plain_len < 5 or no_para:
                print("  skip empty/placeholder: %s" % dtitle[:40], flush=True)
                skip_count += 1
                continue
            item = {
                "site_name": SITE_NAME, "source_url": abs_url, "url": abs_url,
                "title": dtitle, "pub_date": ddate, "summary": "",
                "content": content, "category": CATEGORY, "tags": "",
                "group_name": GROUP_NAME,
                "attachments": json.dumps(
                    [{"title": t, "url": u} for u, t in atts], ensure_ascii=False) if atts else "",
            }
            if JSONL_PATH:
                jsonl_rows.append(item)
            else:
                batch.append(item)
            new_count += 1
            print("  + %s (%s, %d字%s)" % (dtitle[:38], ddate, plain_len,
                                           ", 表格" if has_table else ""), flush=True)
            time.sleep(0.2)
        if not JSONL_PATH and batch:
            push_to_searchdb(batch, batch_label=SCRIPT_NAME)
            batch = []
        time.sleep(0.3)

    if JSONL_PATH:
        with open(JSONL_PATH, "w", encoding="utf-8") as f:
            for row in jsonl_rows:
                f.write(json.dumps(row, ensure_ascii=False) + "\n")
        print("JSONL written: %s (%d rows)" % (JSONL_PATH, len(jsonl_rows)), flush=True)
    else:
        if batch:
            push_to_searchdb(batch, batch_label=SCRIPT_NAME)

    print("新增: %d | 跳过: %d | CUTOFF截断: %d" % (new_count, skip_count, cutoff_count), flush=True)
    print("=== %s done ===" % SITE_NAME, flush=True)


if __name__ == "__main__":
    main()