what's the best way to detect a python type in rust?

[shi-yan]

I'm trying to serialize a python dict into a json in rust.

what's the best way to detect python types?

currently I'm doing python_var.get_type().name().unwrap(), with a string comparison. it seems to be inefficient?

is there a better way doing this? Thanks!

Another question, how to initialize python's NoneType? there is no document covering that?

And how can I convert from PyDict to PyAny?

Thanks.

[birkenfeld]

You should be able to directly compare the result of get_type()? (better not use unwrap unless forced) for equality. However, obj.is_instance is also available which might be a better fit.

What do you mean by "initialize NoneType"?

&PyDict should automatically deref to &PyAny, so for specific problems please post a bit of code.

[shi-yan]

Thank you.

I'm creating a binding for a no-sql database. I need to convert what the db returns into a python object.

The database uses json-like format. an entry can be a dict-like {'test':32} or it can be an array like [23,23,23...]

therefore, I want my function return PyAny.

fn fetch_from_db(...) -> PyResult<PyAny> {

if (db.result.is_array) {
create a PyList and return
}
else is_dict {
create a PyDict and return
}
else is_null {
create a python equivalent of None in rust and return.
}
}

so NoneType is python's equivalent of None , because null is valid for json.

&PyAny can't be returned.

[mejrs]

You need to return the owned version of Py<PyAny> (that is what the PyObject alias is), if lifetimes forbid you from returning the borrowed variant. See also https://pyo3.rs/main/doc/pyo3/conversion/trait.IntoPy.html#dynamic-conversion-into-python-objects

[birkenfeld]

You should still be able to return &PyAny if you associate it with an incoming 'py lifetime, i.e.

fn fetch<'py>(py: Python<'py>, ...) -> PyResult<&'py PyAny> { ... }

But Py<PyAny> is just as fine, don't worry too much about that. The most important thing is that you can convert between most of the different object types in some way or other, which is described in detail here.

[davidhewitt]

By the way, from what you're doing sounds like https://github.com/davidhewitt/pythonize might be useful?