如何使Pygments尊重Catcode的更改

Question

我正在尝试从默认词法分析器开始为LaTeX编写自定义Pygments词法分析器。到目前为止，我有

from pygments.lexer import RegexLexer, include
from pygments.token import *

# We rip off the default styles.
Normaltext = Generic.Inserted
Mathtext = Generic.Deleted
Controlsequence = Keyword
Controlword = Controlsequence
Controlsymbol = Keyword.Pseudo
Specialchar = Name.Builtin

class MyTexLexer(RegexLexer):
    """
    Custom lexer for the TeX and LaTeX typesetting languages.
    """

    tokens = {
        'general': [
            (r'%.*?\n', Comment),
            (r'\\[a-zA-Z]+', Controlword),
            (r'\\.', Controlsymbol),
            (r'\\$', Controlsymbol),
            (r'[&_^{}]', Specialchar),
        ],
        'root': [
            (r'\$\$', Specialchar, 'displaymath'),
            (r'\\\(', Controlsymbol, 'inlinemath'),
            (r'\$', Specialchar, 'inlinemath'),
            (r'\\\[', Controlsymbol, 'displaymath'),
            include('general'),
            (r'[^\\$%&_^{}]+', Normaltext),
        ],
        'math': [
            include('general'),
            (r'[^\\$%&_^{}]+', Mathtext),
        ],
        'inlinemath': [
            (r'\\\)', Controlsymbol, '#pop'),
            (r'\$', Specialchar, '#pop'),
            include('math'),
        ],
        'displaymath': [
            (r'\\\]', Controlsymbol, '#pop'),
            (r'\$\$', Specialchar, '#pop'),
            include('math'),
        ],
    }

现在，TeX在类别代码方面有些特殊，因此可以在编译期间更改语言。我不需要需要所有这些支持。

我经常要支持的一件事是使用\makeatletter和\makeatother，它们基本上允许或禁止在控制字名称中使用@。我想将控制字的正则表达式有效地更改为r'\\[a-zA-Z@]+'，只要找到\makeatletter，找到r'\\[a-zA-Z]+'时就回到\makeatother。

有没有办法以这种方式动态改变类的行为？

---

我不确定是否需要一个TeX示例，但是可以肯定的是，这应该阐明所需的行为：

% Here, \relax is a macro name and @or@do@something is just text:
\relax@or@do@something
\makeatletter
% Here, \relax@or@do@something is the macro name:
\relax@or@do@something
\makeatother
% Here, it's like in the first case:
\relax@or@do@something