Interpolation matrix multiply (testing...)

[ecomodeller]

Vector multiplication in loop replaced with matrix multiplication. (See #413)

Still left to do: create weights matrix without loop.

[jsmariegaard]

I think it's better just to replace the whole thing with:

def _interp_itemstep(data, elem_ids, weights=None):
    if weights is None:
        return data[elem_ids]
    else:
        return np.sum(data[elem_ids]*weights, axis=1)

This is equivalent to taking the dot product of each row but much much faster.

[jsmariegaard]

Closer investigation shows that this is needed:

  def _interp_itemstep(data, elem_ids, weights=None):
    if weights is None:
      return data[elem_ids]
    else:
      idat = data[elem_ids] * weights
      return np.sum(idat, axis=1) if weights.ndim == 2 else idat

To also handle the cases when n_nearest=1...

[jsmariegaard]

You closed this? But wouldn't it be good to implement my suggestion?

[ecomodeller]

Sure, I'll create a new branch with a fresh start