Thanks for reporting, I'll get to this when I can!

Coming back to this 18months later, I see that this fix is no longer enough as
lattice = Lattice.from_lengths_and_angles(*lengths_and_angles)
has now been removed.

Replacing this call with

 lattice = Lattice.from_parameters(lengths_and_angles[0][0],
     lengths_and_angles[0][1],
     lengths_and_angles[0][2],
     lengths_and_angles[1][0],
     lengths_and_angles[1][1],
     lengths_and_angles[1][2])

seems to work. I could make a pull request for this - however, I think there is probably a more elegant python fix, so would prefer to defer to a more competent python coder!

Thanks for the investigation and reminder Jonathan, I'll take a look.

@utf I've had a look at this and run into a problem: Lattice.from_parameters does not generate the cell in lower-triangular form as CASTEP does. This would give an inconsistent structure if the atom positions are given in absolute coordinates.

Do you know of any functionality in Pymatgen to standardise the cell to lower-triangular? It's easy in ASE...

@jryates Looking at subroutine cell_abc_to_cart_lattice(a,b,c,alpha,beta,gamma,mycell) in the CASTEP source, it is somewhat complex and uses the spacegroup. Perhaps we should refuse to read .cell files that use lattice_abc with positions_abs, rather than risk getting it wrong.