Possible wrong calculation in spatialcor.__dhist_angle

[kay-ro]

Hi, I think there might be an error in the calculation of the diff of the function spatialcor.__dhist_angle

I think for periodic boundary conditions the diff needs to calculated with the commented-out code, respecting the periodic images.

...
...
...
    sl = slice(chunk, chunk+chunksize)
    
    # compute the 2D histogram (r,theta)
    hist = np.zeros((len(dbins)-1,len(abins)-1), dtype=float)

    for n in np.arange(len(other_coords))[sl]:
        # calculate distance vectors
        if same:
            # diff = pbc_diff(
            #     other_coords[n],
            #     coords[np.arange(len(coords)) != n], # exclude the particle itself
            #     box_boundary,
            #     pbc=pbc
            # )
            # exclude the particle itself
            diff = other_coords[n] - coords[np.arange(len(coords)) != n]
        else:
            # diff = pbc_diff(other_coords[n], coords, box_boundary, pbc=pbc)
            diff = other_coords[n] - coords
        
        # get distances
        dist = (diff**2).sum(axis=1)**0.5

        # calculate angles
        theta = np.arccos(np.dot(e, diff.T)/dist)
        theta[diff[:,1] < 0.0] = 2*np.pi - theta[diff[:,1] < 0.0] # use angles between 0 and 2\pi
        
        # orient x-axis along mean sample orientation
        # The rotation is applied to the difference vectors instead of the
        # particle positions, since the distances can only calculated 
        # correctly if the particle coordinates and the simulation box
        # fit to each other (which is no longer the case when coordinates
        # are rotated)!
        if angle != 0.0:
            theta = theta - angle
            theta[theta<0.0] = theta[theta<0.0] + 2*np.pi
            theta[theta>2*np.pi] = theta[theta>2*np.pi] - 2*np.pi

        # calculate 2D histogram
        hist += np.histogram2d(dist, theta, [dbins,abins])[0]

Could anyone confirm or correct my understanding here?

[kay-ro]

I think this might not even produce incorrect results but would make the histograms with a lower intensity close to distance=0.
but that's just a quick thought.

[kay-ro]

an on-the-fly chat reinforced my suspicion.
Creating issue for now and linking a branch with a fix.

[kay-ro]

cf. #143