The following example ```python from nutils.function import Array import numpy Array.cast(numpy.array(['a', 'b'], dtype=object)) ``` gives the following error message ``` Traceback (most recent call last): File "<stdin>", line 3, in <module> Array.cast(numpy.array(['a', 'b'], dtype=object)) File "<stdin>", line 144, in cast value = _Constant(value) File "<stdin>", line 752, in __init__ self._value = types.arraydata(value) File "<stdin>", line 573, in __call__ return args[0].__new__(*args, **kwargs) File "<stdin>", line 747, in __new__ dtype = dict(b=bool, u=int, i=int, f=float, c=complex)[array.dtype.kind] KeyError: 'O' ``` This is not really helpful. Consider adding a proper error message informing the caller that object arrays are not supported.