Add dir() to ObjectProtocol.

[m-ou-se]

No description provided.

[kngwyu]

Thank you!
Looks the error in travis is because of the version mismatch between cargo and rustc? 🤔 (weird, though)

[kngwyu]

@m-ou-se
Could you please rebase on the master?
#888 would resolve the travis failure.
In addition, any test for this API is appreciated.

[davidhewitt]

A general comment about these kinds of functions - dir, repr etc.

At the moment they're trait methods, but I personally would prefer them to be free functions like this:

fn dir<'py, T>(obj: &T) -> &PyList
where &'py T: AsRef<PyAny>
{ ... }

// Use later as 
pyo3::dir(obj);

It's not a strong feeling, but I think it is more consistent with how these work in Python, where they're also used as dir(obj), repr(obj) etc.

Wonder how you feel about this?

(If we wanted compatibility we could leave them on ObjectProtocol but mark as deprecated?)

[kngwyu]

A general comment about these kinds of functions - dir, repr etc.

A bit off-topic and should be discussed in another thread?

[davidhewitt]

A bit off-topic and should be discussed in another thread?

I thought it was relevant given we're adding dir now, but sure I'll move this to another issue.

[davidhewitt]

Many thanks @m-ou-se for updating this PR. As I guess you have a need for dir(), I'd be interested to hear what you think of the idea proposed in #892 . (We could have pyo3::dir() and o.__dir__() in the way suggested by that PR.)

[m-ou-se]

@davidhewitt Ah, didn't see that one yet. Looks good.

[davidhewitt]

@davidhewitt Ah, didn't see that one yet. Looks good.

I think I pushed that PR about 20 seconds after you updated this branch! 😄 I'll wait for some more feedback to see what others think of that PR, but I will definitely make sure to add dir() .

[davidhewitt]

FWIW please do merge this PR without waiting for #892 - we might wait a little while to decide the right design there.

[kngwyu]

Thank you!