These changes aren't related to the title of this PR. Please either split it out into a separate PR, or retitle this PR.

Oh, yes. They are from other issue.

This seems like a better approach, but if you're modernizing this then I have a few other suggestions.

Perhaps this pattern ought to be put at module scope, just above the class definition? The verbose representation might be more readable without all that leading whitespace.

Too bad we can't convert this to a .format() string without breaking backward compatibility.

We can. We can even convert it to f-string. I tried, but returned to the old format.

The only problem with .format() and f-string is that we nee to duplicate literal braces. And the template/f-string will look strange, because braces have special meaning in regexpes and in new format.

With f-string it would look:

    def __init_subclass__(cls):
        super().__init_subclass__()
        if 'pattern' in cls.__dict__:
            pattern = cls.pattern
        else:
            delim = _re.escape(cls.delimiter)
            id = cls.idpattern
            bid = cls.braceidpattern or cls.idpattern
            pattern = fr"""
            {delim}(?:
              (?P<escaped>{delim})  |   # Escape sequence of two delimiters
              (?P<named>{id})       |   # delimiter and a Python identifier
              {{(?P<braced>{bid})}} |   # delimiter and a braced identifier
              (?P<invalid>)             # Other ill-formed delimiter exprs
            )
            """
        cls.pattern = _re.compile(pattern, cls.flags | _re.VERBOSE)

This is a bit of a head scratcher, but I think the comment could be improved to make it more clear. What's really going on is that this is required to initialize pattern when no subclass is created. Can you improve the clarity of the comment (and maybe move it above this line, and make it a full sentence)? If you need a suggestion, let me know.

I would appreciate your help with a comment.