背诵版01 扩展段逐条标源 + 防幻觉校验器
- 扩展段 76 条加来源角标:〔张宇〕40 条(tex 有据)、〔补〕36 条(我补的),参考占比 52.6%
- 修正 §7.2 事实错误:e^{1/x} 在 0 处左 0 右 +∞,属第二类无穷间断点,不是跳跃
- 「教材原文 / 教材例」措辞改为「张宇三十讲整理稿 / 张宇例」——该 tex 是张宇三十讲整理稿,非同济教材
- 新增 .workbuddy/tools/check_source_tags.py:扩展条目缺角标即 FAIL,〔张宇〕占比<20% 也 FAIL
- README 增「背诵版三段式标准 v2」硬规则;MEMORY 增防幻觉硬约束五条
This commit is contained in:
@@ -0,0 +1,130 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""背诵版「扩展」段来源角标校验器。
|
||||
|
||||
规则(2026-11-22 用户定,见 知识点构建/README.md §五):
|
||||
每个知识点的 **扩展** 段,每条列表项末尾必须带来源角标:
|
||||
〔张宇〕 = 该条主体能在 参考仓库/Math/advanced-math 原文找到
|
||||
〔补〕 = 参考仓库没有,是我补的(通用知识 / 真题归纳 / 陷阱 / 流程)
|
||||
混血条目一律标〔补〕——〔补〕的含义是「这句要你自己判断对错」。
|
||||
|
||||
用法:
|
||||
python check_source_tags.py # 扫 知识点构建 全目录
|
||||
python check_source_tags.py --file x.md # 只查一个文件
|
||||
python check_source_tags.py --min-ref 20 # 〔张宇〕占比下限,默认 20%
|
||||
|
||||
退出码:0 通过;1 有缺标或占比异常。
|
||||
只检查含「**教材**」标记的文件(即已按三段式重构的);旧结构文件跳过并列出。
|
||||
"""
|
||||
import io
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
|
||||
TAG_ZY = '〔张宇〕'
|
||||
TAG_BU = '〔补〕'
|
||||
BASE = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
||||
DEFAULT_DIR = os.path.join(BASE, '基于考研的学习', '数学一', '知识点构建')
|
||||
|
||||
|
||||
def check_file(path, min_ref=20):
|
||||
txt = io.open(path, encoding='utf-8').read()
|
||||
name = os.path.basename(path)
|
||||
if '**教材**' not in txt:
|
||||
return None # 旧结构,未重构
|
||||
|
||||
lines = txt.split('\n')
|
||||
# 切小节:### 开头到下一个 ### 或 ## 之前
|
||||
secs = []
|
||||
cur = None
|
||||
for i, ln in enumerate(lines, 1):
|
||||
if re.match(r'^#{2,3} ', ln):
|
||||
if cur:
|
||||
secs.append(cur)
|
||||
cur = {'title': ln.strip('# ').strip(), 'line': i, 'body': []} if ln.startswith('### ') else None
|
||||
if cur is None:
|
||||
if secs and False:
|
||||
pass
|
||||
elif cur is not None:
|
||||
cur['body'].append((i, ln))
|
||||
if cur:
|
||||
secs.append(cur)
|
||||
|
||||
problems = []
|
||||
n_zy = n_bu = n_untag = 0
|
||||
for s in secs:
|
||||
# 找扩展块
|
||||
in_ext = False
|
||||
ext_items = []
|
||||
for i, ln in s['body']:
|
||||
if ln.startswith('**扩展**'):
|
||||
in_ext = True
|
||||
continue
|
||||
if ln.startswith('**') and in_ext:
|
||||
in_ext = False
|
||||
if in_ext and ln.startswith('- '):
|
||||
ext_items.append((i, ln))
|
||||
if not ext_items:
|
||||
problems.append((s['line'], s['title'], '缺少「**扩展**」段或段内无条目'))
|
||||
continue
|
||||
for i, ln in ext_items:
|
||||
if TAG_ZY in ln:
|
||||
n_zy += 1
|
||||
elif TAG_BU in ln:
|
||||
n_bu += 1
|
||||
else:
|
||||
n_untag += 1
|
||||
problems.append((i, s['title'], '扩展条目未标来源:' + ln.strip()[:36] + '…'))
|
||||
|
||||
total = n_zy + n_bu + n_untag
|
||||
ref_pct = (100.0 * n_zy / total) if total else 0.0
|
||||
ok = (not problems) and total > 0 and ref_pct >= min_ref
|
||||
return {
|
||||
'file': name, 'total': total, 'zy': n_zy, 'bu': n_bu,
|
||||
'untag': n_untag, 'ref_pct': ref_pct, 'problems': problems, 'ok': ok,
|
||||
}
|
||||
|
||||
|
||||
def main():
|
||||
args = sys.argv[1:]
|
||||
min_ref = 20
|
||||
target = None
|
||||
for a in args:
|
||||
if a.startswith('--min-ref='):
|
||||
min_ref = float(a.split('=', 1)[1])
|
||||
elif a == '--file':
|
||||
pass
|
||||
if '--file' in args:
|
||||
target = args[args.index('--file') + 1]
|
||||
|
||||
files = [target] if target else sorted(
|
||||
os.path.join(DEFAULT_DIR, f) for f in os.listdir(DEFAULT_DIR)
|
||||
if f.endswith('.md') and '背诵版' in f)
|
||||
if target:
|
||||
files = [target]
|
||||
|
||||
done, skipped, bad = 0, [], []
|
||||
for p in files:
|
||||
r = check_file(p, min_ref)
|
||||
if r is None:
|
||||
skipped.append(os.path.basename(p))
|
||||
continue
|
||||
done += 1
|
||||
flag = 'OK ' if r['ok'] else 'FAIL'
|
||||
print('%s %-46s 条目%3d 〔张宇〕%3d 〔补〕%3d 参考占比%5.1f%%'
|
||||
% (flag, r['file'], r['total'], r['zy'], r['bu'], r['ref_pct']))
|
||||
if r['problems']:
|
||||
for line, sec, msg in r['problems']:
|
||||
print(' L%-5d %-22s %s' % (line, sec[:22], msg))
|
||||
if not r['ok']:
|
||||
bad.append(r['file'])
|
||||
|
||||
print('\n已重构 %d 篇;未重构(旧结构)%d 篇:%s'
|
||||
% (done, len(skipped), '、'.join(x[:8] for x in skipped) if skipped else '无'))
|
||||
if bad:
|
||||
print('不合格:%s' % '、'.join(bad))
|
||||
sys.exit(1)
|
||||
print('全部通过。')
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
Reference in New Issue
Block a user