Lester Hedges

Thanks, it would be great to have a design discussion about this. I totally agree that the new approach makes sense, and is what you would need with multi molecule...

@jmichel80: Is it possible to merge my [fix](https://github.com/michellab/Sire/tree/fix-somd-2023) branch into `devel` so we have some working SOMD FEP functionality in 2023. I've checked the resulting selection for a few example...

Ah brilliant. Shall I do the PR, or is this something that you wanted to do?

Hi there, we still only have write support for GROMACS alchemical topologies (at least ones set up within BioSimSpace), so this is likely the issue. I'll try to debug this...

Hmmm, I can't even read the files, which isn't surprising if the topology corresponds to an alchemical system: ```python import BioSimSpace.Sandpit.Exscientia as BSS BSS.setVerbose(True) system = BSS.IO.readMolecules(["gromacs.top", "gromacs_out.gro"]) ``` gives:...

Okay, I now see the same error. I'll try to work out what term isn't fitting within the AMBER formatting restrictions following conversion.

It happens if you just use the `gro` file, i.e. you don't even load the topology. I'm guessing that one of the velocities is overflowing the formatting.

Yes, that's it: ```python import BioSimSpace as BSS s = BSS.IO.readMolecules("gromacs_out.gro") m = s[0] # This fails. BSS.IO.saveMolecules("test", m, "rst7") OSError: SireError::io_error: Cannot write the (perhaps some of the )...

Also, for a quick hack, you could just use the `property_map` to make Sire search for an invalid (missing) velocity property key, in which case the velocities would be ignored....

Yes, that's a good point. At present this is triggered by the `writeFloatData` function from `amberformat.cpp`, so the function doesn't know what property the data corresponds to. This could easily...