From: Adam T. <ate...@gm...> - 2012-04-23 15:35:41
|
> Adam: > Yes, the relaxed scan case in not handled by this code. This is due to issues implementing the extraction of geometries. I think the best way to do this would be to look for "optimisation complete" messages and then add the last parsed geometry to a list of scan geometries. However currently the gaussian parser stops searching after a "optimisation complete" message; changing this would affect the exposed list of geometries and potentially break scripts based on cclib. If you can think of a better way of implementing this I would be happy to provide a patch. I agree that this change would break scripts if people were only parsing the first optimization of a scan calculation, but shouldn't most users (at least of released versions) only be using cclib for optimizations at this point, and not scan calculations? Otherwise, they aren't getting the majority of the structures/energies from their results. Isn't this why these patches were created? Or am I making a bad assumption about cclib use? > 2) > This seems sensible Regarding separating the scan and thermochemistry data: I agree this is a good idea going forward, focusing on a independent results separately. As I already made the changes to the Gaussian parser, I'm inclined to leave the two major changes as a single commit instead of going back to a previous revision to separate the two into new separate commits. > 3) > I agree this seems sensible for the rigid scan case. However it brings us back to the issue of the relaxed scan again; these are a series of optimisations so reusing scfenergies could cause confusion as to the exact meaning of the values and it would also mean that data on the individual optimisation steps is lost. It would seem a choice between simplicity of the interface and completeness of the data collected needs to be made. This is perhaps something we should discuss. I think that scfenergies (atomcoords, etc). should hold the energies of every completed SCF calculation, whether or not its an optimization or scan calculation. In the case of a scan (rigid or relaxed) calculation, I think special attributes for the final optimized step should be created. Adam P.S. Ed, did you attach the examples? Ethanol and ethanol-z were in a previous email... |