Virtual sites in alchemical factory and add ions in WaterBox

[andrrizzi]

Ready for review!

• Fix AbsoluteAlchemicalFactory does not support virtual sites #258 : add support for virtual sites in AbsoluteAlchemicalFactory.
• Add positive_ion, negative_ion and ionic_strength constructor arguments to WaterBox. This was necessary to create a TIP4P test with something that could be alchemically modified (the ions).

[jchodera]

Awesome! This is much more future-proof! A quick read through looks good to me, but might be good to have @Lnaden read through in detail!

[Lnaden]

I'll take a dive through it once I am done writing up the part of the website update I'm finishing off

[Lnaden]

Looks good, logic on the adding and removing forces should be fine since the _alchemically_modify_X functions only create copies.

My only question though, what was wrong about using the getVirtualSite method of the System object to copy over the virtual site?

[andrrizzi]

what was wrong about using the getVirtualSite method of the System object to copy over the virtual site?

I tried that first. The VirtualSite SWIG object you obtain by reference_system.getVirtualSite() belongs to the reference System and can't be assigned to others. You'd need to copy it before calling alchemical_system.setVirtualSite(), but OpenMM failed when I tried to deepcopy it. I don't think it's implemented.

[Lnaden]

Ah, that makes sense then. Its probably coded similar to the integrator then once it gets assigned a context, its consumed for the purposes of copying to use elsewhere. Looks good otherwise