sys.implementation test for being a tuple fails.

[robert-hh]

The value of sys.implentation cannot be tested as being a tuple. That used to work, but does not any more. That may have been discussed already, bur I cannot recall that. Sample sequence:

>>> import sys
>>> s=sys.implementation
>>> type(s)
<class 'tuple'>
>>> type(s) is tuple
False

but:

>>> t=(1,2,3)
>>> type(t)
<class 'tuple'>
>>> type(t) is tuple
True

[dlech]

Since sys.implementation is required to have named attributes, it doesn't seem like it should be exactly a tuple. Does isinstance(t, tuple) work?

[robert-hh]

Nope:

>>> import sys
>>> s=sys.implementation
>>> isinstance(s, tuple)
False
>>> 

but:

>>> t=(1,2,3)
>>> isinstance(t, tuple)
True
>>> 

[dlech]

It looks like the type depends on MICROPY_PY_ATTRTUPLE, it could be mp_type_tuple or mp_type_attrtuple. But in any case it seems like, type(t).__name__ == "tuple" should work.

Why do you need to test the type?

[robert-hh]

MICROPY_PY_ATTRTUPLE is enabled. type(t).__name__ == "tuple" works.
Actually, I do not need it. I was just puzzled because is used to work a while ago and now it does not any more. I came along it when I tried to run Python code which I did not use since 2020.

[peterhinch]

I had assumed it was a namedtuple but my tests suggest it isn't (in MP or in CPython).

[jimmo]

Yeah, it's an attrtuple which cannot be created or accessed from Python, but (confusingly) shares the same name QSTR as regular tuple.

With attrtuple support disabled, then:

>>> type(sys.implementation) is tuple
True