Larry Hastings added the comment:
In case a little background would help: while developing support for
'__text_signature__' I had to move the test and the from_builtin() call to the
top. It used to be more in the middle-ish. I don't have notes specifically on
why I moved it, but I dimly recall that if I didn't try from_builtin first,
there were other approaches that would succeed on the callable, that approach
would fail on a builtin, so it'd throw an exception, and from_builtin wouldn't
get a chance to try.
Also, I assumed that anything that had a __text_signature__ wasn't going to
have any other kind of signature information. Therefore, if the object had a
__text_signature__ attribute, then I already knew that's the best approach and
it should definitely be first.
Also also, I remember specifically that the isinstance(type) code would fail
builtin classes. But there's no generic way (that I know of) to tell whether a
class is a user class or a builtin class. So I wanted from_builtin to try
handling a class first before the isinstance(type) class.
Are there callables in CPython that have *both* a __text_signature__ *and* have
signature information derivable from another source, where you *don't* want to
use __text_signature__?
Stefan: yes, you can write any garbage you want after "sig=(" and the C
function that detects signatures will still assume you have a text signature.
That reminds me of a good joke.
Patient: Doctor, it hurts whenever I move my arm in this funny way.
*demonstrates*
Doctor: Well, then, don't do that.
I would be very interested if you knew of docstrings in the wild that
innocently start with "sig=(" and yet aren't intended to be text signatures
compatible with inspect.Signature.
----------
_______________________________________
Python tracker <[email protected]>
<http://bugs.python.org/issue17159>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
