（详情页看不到附件）
short 短正文（<30 字；纯附件公告/产权声明类本身就只有一句话，属正常）
attnomd 正文出现 markdown 链接（[text](http…)）→ 违反用户铁律，应转
用法:
python3 audit_jsonl.py /tmp/c_p1.jsonl
python3 audit_jsonl.py /tmp/c_p1.jsonl --list-total 253 --cutoff 4
python3 audit_jsonl.py /tmp/c_p1/*.jsonl
退出码: 0=通过（无失败级）, 1=有失败级, 2=用法错误
"""
import json, glob, os, re, sys
# 标题污染黑名单：站点导航词 + 未剥离的后缀变体
BAD_WORDS = [
'网站支持IPv6', '无障碍浏览', '当前位置', 'Copyright', '版权所有', '主办单位',
'热点推荐', '404', '出错了', '| 崇义', '| 崇义县人民政府',
'_通知公告', '_药都资讯', '_建设项目', '_公示公告', '_政务公开',
'_主动公开', '_工作动态', '_部门文件', '_双公示', '_人事信息', '_组织机构',
' - 通知', ' - 建设', '-赤壁', '-赤壁市政府网',
]
def _att_urls(raw):
"""attachments 列 → URL 列表。兼容 JSON 数组（[{"title","url"}]）与管道串（url|name; url2|name2）。"""
if not raw:
return []
s = raw.strip()
if s.startswith('['):
try:
out = []
for a in json.loads(s):
if isinstance(a, dict):
out.append(a.get('url') or a.get('href') or '')
elif isinstance(a, str):
out.append(a)
return [u for u in out if u]
except Exception:
return []
return [p.split('|', 1)[0].strip() for p in re.split(r';', s) if p.strip()]
def audit(path, verbose=True, list_total=None, cutoff=0):
lines = [l for l in open(path, encoding='utf-8').read().strip().split('\n') if l.strip()]
n = len(lines)
empty = short = bad = attnomd = nopara = notin = atturl = 0
bad_samples, atturl_samples = [], []
for l in lines:
try:
d = json.loads(l)
except Exception:
continue
t = d.get('title', '')
c = d.get('content', '') or ''
if not c.strip():
empty += 1
elif len(re.sub(r'<[^>]+>', '', c).strip()) < 30:
short += 1
if any(w in t for w in BAD_WORDS):
bad += 1
if len(bad_samples) < 3:
bad_samples.append(t[:70])
# 失败级：非 http 附件（mailto/tel/相对路径）
urls = _att_urls(d.get('attachments'))
non_http = [u for u in urls if not u.startswith('http')]
if non_http:
atturl += 1
if len(atturl_samples) < 3:
atturl_samples.append('%s -> %s' % (t[:34], non_http[:3]))
# 提示级：段落承载 / 附件内嵌 / markdown 残留
if c.strip() and not re.search(r']', c, re.I) and not re.search(r']*href="http[^"]*"[^>]*>[^<]+', c):
notin += 1
if re.search(r'\[[^\]]+\]\(http', c):
attnomd += 1
counted_total = n + cutoff
count_bad = list_total is not None and counted_total != list_total
if list_total is not None and count_bad:
print(' ❌ 条数对账: 入库 %d + CUTOFF %d = %d , 列表 %d , 差 %d'
% (n, cutoff, counted_total, list_total, list_total - counted_total))
fail = (empty or bad or atturl or count_bad)
flag = '❌' if fail else '✅'
print('%s %s: n=%d empty=%d badtitle=%d atturl=%d'
% (flag, os.path.basename(path), n, empty, bad, atturl))
print(' ⚠️ nopara=%d notin=%d markdown_link=%d short=%d'
% (nopara, notin, attnomd, short))
for s in bad_samples:
print(' BAD TITLE: %s' % s)
for s in atturl_samples:
print(' 非 http 附件（mailto/tel 被当附件?）: %s' % s)
if notin:
print(' ⚠️ 有附件但正文未内嵌
附件名
→ 详情页看不到附件')
if nopara:
print(' ⚠️ 正文无 /