Turbomole support

[eljost]

Hi,
a request for a TURBOMOLE interface came up in this geomeTRIC issue.
As I already developed a TM-wrapper for my own code I'd be willing to work on an implementation in QCEngine.
Right now the way my wrapper works is quite incompatible to QCEngine I imagine... In my code you have to prepare a control-file beforehand (through the TM-utility define) that gets passed to the wrapper class. All contents from this directory get copied to a temporary directory, from where the calculation is actually run.

What is needed for QCEngine is probably a wrapper for define?!
What would you consider to the minimal feature-set that should be implemented?

Best regards
Johannes

[dgasmith]

TURBOMOLE integration would be fantastic and something that has been asked after before.

A minimal viable block for a compute is a SCF energy/gradient return. This would satisfy the geomeTRIC requirements as well. A lot of the technology behind your wrapper can be reused I believe. The first step would be to join Slack (on the README of this repo) so that we can chat a bit about possible solutions and integrations. If TURBOMOLE does not support some sort of binary output it would like be the case that the wrapper would look similar to NWChem.

Overall QCEngine would like to support projects like pysisyphus so that everyone does not have to rewrite these wrappers. We look forward to working with you on this.

[mkrompiec]

@eljost: great that you are happy to do this. I think a wrapper for define would be a good idea, as long as examples of "typical" use cases are given. One more thing: pysiphus is released under GPL, would you be happy to release the code under a more permissive license?

[dgasmith]

@mkrompiec I believe the Turbomole wrapper is now in here under BSD-3C.

Thanks again @eljost for this! We look forward to running Turbomole :)