I just gave it a try:

Before:

class StatResult(tuple):
    """
    Support class resembling a tuple like that returned from
    `os.(l)stat`.
    """
    ...

After:

StatResultBase = collections.namedtuple(
                   "StatResult", ["st_mode", "st_ino", "st_dev", "st_nlink",
                                  "st_uid", "st_gid", "st_size", "st_atime",
                                  "st_mtime", "st_ctime"])


class StatResult(StatResultBase):
    """
    Support class resembling a tuple like that returned from
    `os.(l)stat`.
    """
    ...

The problem is that before the StatResult constructor took one argument (the tuple with the items to use) and with the change the StatResult constructor takes 11 arguments (the tuple items).

Since StatResult is a public class for implementation of custom parsers, I don't want to change the StatResult constructor.

As you suggest, I also think the idea behind using a namedtuple isn't the class itself, but the representation of its instances. Therefore I added a repr to StatResult so that the representation looks like that of a namedtuple [236781615dd35adca1dbed54f2393196fbd14523](https://git.sr.ht/~sschwarzer/ftputil/commit/236781615dd35adca1dbed54f2393196fbd14523 "Add __repr__ so that a StatResult looks like a named tuple (#91) ...").