suspected bug in Projection Plots for SPH datasets about yt HOT 2 OPEN

I think basically the same issue also comes up in slices through SPH datasets:
pixelization_routines.pyx, line 1465 onwards: pixelize_sph_kernel_slice. The general approach here seems to be to evaluate the SPH field $A$ at the center $\vec{x}$ of each slice pixel:

where $A_j$ is the value of that field for particle $j$, $m_j$ is its mass, $\rho_j$ is its density, $\vec{x_j}$ is its position, and $h_j$ is its smoothing length. For small particles with smoothing lengths $\lesssim$ the pixel size, the smoothing lengths are artificially stretched, so that the normalized kernel and prefactor parts of the contribution are calculated inconsistently, and small SPH particles add to the total even when they do not intersect the pixel center.

Finally, I would suggest that we really need the z values for this as well. Now, this method calculates the kernel contributions for all particles as if the particle center lies exactly in the plane of the slice, but really, the kernel should probably be evaluated at the 3D-distance to the pixel center, to be consistent with how SPH interpolation is done in a simulation.

from yt.

The 3D version of this grid deposition pixelization_routines.pyx, line 1607 onwards: pixelize_sph_kernel_arbitrary_grid seems to handle the 3D part just find, but again sets minimum smoothing lengths (for the kernel evaluation, but not the pre-factor). This is inconsistent with an evaluation of the SPH field at a voxel center.

from yt.