IntoPy not implemented for Bound

[wyfo]

I've just updated some old code to 0.21, and I had the following code:

let list: Py<PyList> = Python::with_gil(|py| PyList::new(py, vec).into_py(py));

PyList::new is deprecated, so I've changed to

let list: Py<PyList> = Python::with_gil(|py| PyList::new_bound(py, vec).into_py(py));

and I got the following error expected Py, but found PyObject`.

Why is there no impl IntoPy<Py<PyList>> for Bound<'_, PyList>, while an impl IntoPy<Py<PyList>> for &PyList exists?

[adamreichold]

Not really an answer as to why the trait impl does not exist, but I think you want Bound::unbind. We also recently extended the guide on this topic, c.f. #4031.

(I think in this particular case, unbind is generally preferable because it is the more narrow operation and hence better self-documenting. Still for generic context, a blanket impl like impl<T> IntoPy<Py<T>> for Bound<'_, T> might be nice.)

[adamreichold]

I think it would be preferable to engage in #4041 hashing out how to simplify the overall conversion trait approach.

[davidhewitt]

I think it's the second optional item in #3960

If I recall, the problem is we already have IntoPy<Py<PyAny>> for Bound<T>, and this conflicts with a generic IntoPy<Py<T>> for Bound<T>.

Probably there's a solution like what we did for the Deref implementations. It would probably break type inference to ship a solution in a patch release but we could include it in 0.22

This feels also very related to what we're discussing in #4041 and generic traits.

[davidhewitt]

(oops, @adamreichold said the same thing at exactly the same time 😂)

[wyfo]

If I recall, the problem is we already have IntoPy<Py> for Bound, and this conflicts with a generic IntoPy<Py> for Bound.

I think that's why impl IntoPy<Py<$name>> for &$name is in pyobject_native_type_named. Maybe we could add impl IntoPy<Py<$name>> for Bound<'_, $name> and impl IntoPy<Py<$name>> for &Bound<'_, $name> in pyobject_native_type_named as well.

[davidhewitt]

Ah yes, that would probably work. I also need to have a good read of #4060 which hopefully gives us a way out of the current bundle of traits!