TypeError: argument 'input': 'ndarray' object is not an instance of 'ndarray'

[mikelync]

When trying to capture an Array from Python using the following input scheme:

fn _set_node_vector<'py>(
    &mut self, 
    py: Python<'py>, 
    input: Bound<'py, numpy::PyArray1<Py<PyAny>>>,
) -> PyResult<()>

I have three cases:
a) Valid input: the array is correctly specified: v1 = np.array([1.0, 2.0], dtype=object)
b) Invalid i1 = np.array([1.0, 2.0]) (possibly because the dtype is float)
c) Invalid: i2 = np.array([[1.0, 2.0]], dtype=object) (shape is wrong)

In both invalid cases the error returned is:

TypeError: argument 'input': 'ndarray' object is not an instance of 'ndarray'

Bit more steering can be helpful...

[davidhewitt]

Thanks for the report. Agreed this message should be better, but I'm not sure there's an easy way to fix this.

The blanket FromPyObject implementation for Bound uses .cast() internally, which will always produce an error of this form. I guess if PyTypeInfo / PyTypeCheck implementations were removed from PyArray1, the blanket would not apply and a custom FromPyObject implementation could be used.