Type signature of numpy::innner does not cover all cases

[lmk0724]

when I use numpy::inner(), I face the following problem,
float64' object cannot be converted to 'PyArray<T, D>

here is my code
let array = numpy::pyarray![py, 1.0, 2.0, 3.0]; let inner: &numpy::PyArray0::<_> = numpy::inner(array, array).unwrap(); println!("{}",inner.item());

[adamreichold]

I can confirm that this issue happens with arrays of element type f64, but not with arrays of element type i32, i.e. NumPy's effective type signature is more dynamic than our

fn inner<'py, T, DIN1, DIN2, DOUT>(
    array1: &'py PyArray<T, DIN1>,
    array2: &'py PyArray<T, DIN2>,
) -> PyResult<&'py PyArray<T, DOUT>>;

and we need to figure out how to extend this to cover this case. (While I understand that it is not a problem for Python itself, this inconsistency seems somewhat unfortunate.)

@pickanameforlove As a workaround, I think you should be able to use ndarray's ArrayBase::dot to compute the inner product of two vectors as long as they are real-valued. (I would always recommend to use ndarray functions if they get the job done.)

[lmk0724]

Ok, Thank you

[adamreichold]

This will require a breaking change in our API and therefore will not be part of a 0.16.x bugfix release. Sorry for that.