修正: deepseek 笔记 96 处块级公式改为多行写法(适配 SilverBullet 行首锚点), 索引脚本同步 _abstract 命名

大模型/deepseek/_scripts/math_fix.py

@@ -0,0 +1,122 @@
+"""Audit / convert single-line $$...$$ math blocks for SilverBullet compatibility.
+
+SilverBullet's Math library defines block math with the line-anchored marker
+regex ^\\$\\$$ for BOTH start and end, so `$$...$$` written on one line never
+closes and is rendered as raw text.
+
+Usage:
+    python math_fix.py audit   <base_dir>
+    python math_fix.py convert <base_dir> [--apply]
+"""
+import os
+import re
+import sys
+
+CMD = sys.argv[1]
+BASE = sys.argv[2] if len(sys.argv) > 2 else '.'
+APPLY = '--apply' in sys.argv
+
+LINE_RE = re.compile(r'^\$\$(.+)\$\$$')
+
+
+def notes():
+    for root, dirs, files in os.walk(BASE):
+        if any(x in root for x in ['_scripts', '_text', '_digest', '.git']):
+            continue
+        for fn in sorted(files):
+            if fn.endswith('.md') and fn != 'README.md':
+                yield os.path.join(root, fn)
+
+
+def walk_lines(path):
+    """Yield (index, line, in_fence)."""
+    fence = False
+    lines = open(path, encoding='utf-8').read().split('\n')
+    for i, line in enumerate(lines):
+        if line.lstrip().startswith('```'):
+            yield i, line, fence
+            fence = not fence
+            continue
+        yield i, line, fence
+
+
+def audit():
+    total = indented = multi = in_fence = trailing = 0
+    samples = []
+    for p in notes():
+        for i, line, fence in walk_lines(p):
+            s = line.strip()
+            if not s.startswith('$$') or s.count('$$') < 2:
+                continue
+            total += 1
+            if fence:
+                in_fence += 1
+                samples.append(('FENCE', p, i, s[:90]))
+            if line != line.lstrip():
+                indented += 1
+                samples.append(('INDENT', p, i, s[:90]))
+            if s.count('$$') > 2:
+                multi += 1
+                samples.append(('MULTI', p, i, s[:90]))
+            if not s.endswith('$$'):
+                trailing += 1
+                samples.append(('TAILTEXT', p, i, s[-80:]))
+    print('单行 $$ 块总数        :', total)
+    print('  其中在代码块内      :', in_fence)
+    print('  其中带前导缩进      :', indented)
+    print('  其中一行含 >2 个 $$ :', multi)
+    print('  其中不以 $$ 结尾    :', trailing)
+    for kind, p, i, s in samples[:20]:
+        print('  [%s] %s:%d  %s' % (kind, os.path.relpath(p, BASE), i + 1, s))
+
+
+def convert():
+    files_changed = 0
+    blocks_changed = 0
+    skipped = []
+    for p in notes():
+        lines = open(p, encoding='utf-8').read().split('\n')
+        out = []
+        fence = False
+        n = 0
+        for line in lines:
+            if line.lstrip().startswith('```'):
+                fence = not fence
+                out.append(line)
+                continue
+            m = LINE_RE.match(line)
+            if m and not fence:
+                body = m.group(1).strip()
+                if '$$' in body:
+                    skipped.append((os.path.relpath(p, BASE), line[:90]))
+                    out.append(line)
+                    continue
+                out.append('$$')
+                out.append(body)
+                out.append('$$')
+                n += 1
+                continue
+            out.append(line)
+        if n:
+            files_changed += 1
+            blocks_changed += n
+            print('%3d 处  %s' % (n, os.path.relpath(p, BASE)))
+            if APPLY:
+                with open(p, 'w', encoding='utf-8', newline='\n') as f:
+                    f.write('\n'.join(out))
+    print()
+    print('文件数: %d   转换块数: %d' % (files_changed, blocks_changed))
+    if skipped:
+        print('跳过（正文含 $$，需人工确认）:')
+        for s in skipped:
+            print('   ', s)
+    print('模式:', '已写入' if APPLY else '预演（未写入，加 --apply 才落盘）')
+
+
+if __name__ == '__main__':
+    if CMD == 'audit':
+        audit()
+    elif CMD == 'convert':
+        convert()
+    else:
+        raise SystemExit('unknown command: ' + CMD)