I was just trying to generate some numbers, stick them into a new Float64Array(N), and return them to Python.

To my surprise, the returned object wasn't even a list. I can't index into it, result[0] just fails:

ArgumentError: Python argument types in
    JSObject.__getitem__(JSObject, int)
did not match C++ signature:
    __getitem__(class CJavascriptObject {lvalue}, class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> >)

But result["0"] does work... That expression on a normal javascript list fails, rightfully, with TypeError: list indices must be integers.

dir(result) showed me [0, 1, 2, ..., N-1], i.e. the dir() returned a list of numbers, which is extremely odd because dir() on any other python object always returns a bunch of strings, which represent the methods of that object.

Calling np.asarray() on a regular (javascript) list object actually gives me something sensible, but calling it on this object just results in a numpy array of dtype=object, which is not useful. It does not appear to support iter() and next(). The regular javascript list does support those.

I think it'd be great to be able to use typed arrays.

If complete numpy adaptation is too complicated, I'd be happy to get something that presents buffer access at least. The less data to copy, the better.