NULL checks here seem a bit redundant caller can check if state is sane.

Think this could be made into a generic API function & moved to material.c

BKE_material_id_remap() or BKE_material_id_permutate()

It can take an array of shorts and just do bounds check, then, efa->mat_nr = remap[efa->mat_nr];

This means we can more easily sort materials/reverse... etc.

However implementing this, we would also want to have a similar function for object mode, handling ob->matbits, this complicates a little more, but think its still worth doing.

Once this is done, most of this logic should be in BKE, and the operators can just create & free the remap arrays.

think it would be better to get the array from give_matarar, then this can be done at the start along with matbits.

all of these mat_nr accesses need bounds checks, its possible they are larger then the array.

dont need to calloc if its filled right after