How to convert PyReadonlyArray2 to Vec<Vec<_>>?

[tuoeg]

PyReadonlyArray2 how to convert to vec<vec>? thanks

[adamreichold]

I think the easiest way to achieve this is to use ndarray's rows or Columns iterators and collect those, e.g.

fn array2_to_vec2(array: PyReadonlyArray2<f64>) -> Vec<Vec<f64>> {
    array
        .as_array()
        .rows()
        .into_iter()
        .map(|row| row.iter().copied().collect::<Vec<_>>())
        .collect::<Vec<_>>()
}

But that said, this is an usual conversion as the resulting memory layout is relatively inefficient compared to storing the two-dimensional array in a single allocation. Maybe you could expand a little on why you need Vec<Vec<_>> and we can find a way to reduce the number of separate allocations?

[tuoeg]

thanks, I need to send image array from python to rust

[tuoeg]

can you give me some suggestion to study Some underlying principles about python call rust's dll, I first contact rust, I think you are very powerful,thanks

[adamreichold]

I need to send image array from python to rust

If you image data is a NumPy array on the Python side, then PyReadonlyArray2<_> sounds like a good choice. I think my recommendation would then rather be to use the ArrayView2<_> returned by as_array directly to operate on the image data, without copying it into multiple Vec.

This means you can use all of the facilities of ndarray to implement the desired operations. https://docs.rs/ndarray/latest/ndarray/doc/ndarray_for_numpy_users/index.html possibly helpful if you have implemented this using NumPy before, otherwise https://docs.rs/ndarray/latest/ndarray/ is probably a good place to start.,

[tuoeg]

I will consider your suggestion ,I need to explore that whether ArrayView2<_> can convert to opencv function params directly