objects with units :-)

…ng them to the

documentation and options objects. Just running some simulations now to make sure
they are working as expected. Debugging showed the right bonds and angles were
not being constrained.

I've updated the defaults so that ``h-bonds-not-perturbed`` etc are the
default constraints if none are set. I've also added a discussion of PME
versus RF to the tutorial, and updated the example FEP script to
use ``h-bonds-not-perturbed`` and ``RF`` by default.

On my computer, the entire methane/ethanol FEP simulation now takes
about 3 minutes :-)

…, as these

are now available

adding tests to check that the constraints are correct

WIP

…applied

different schedules for different molecules (eventually will be parts
of perturbable molecules)

This will let us use a single lambda value to control perturbations
of multiple molecules (or parts of molecules) that perturb their
parameters at different rates in different forces (or even go in
different directions).

This will let us do absolute binding calculations where we only turn
off the intermolecular energies (switch off ghost/non-ghost but keep
on ghost/ghost), or to do QM/MM end state corrections
(switch off ghost/ghost while keeping ghost/non-ghost the same)

I need to add tests and documentation

neopentane-methane, comparing to somd1

…l correct

if we link to the perturbed properties and/or swap the end states

The test should check bonds-not-perturbed

…hat was

breaking wrapping of trajectories

…harge correction

in the ghost forces. This will be used to control that correction, e.g. so that
we can separate the ghost/ghost coulomb force from the ghost/non-ghost, e.g.
in absolute binding calculations where we don't want to decouple the ligand
from itself.

Fixed a bug where the h-bonds constraint would look at the highest mass
of perturbed atoms, when it needs to look at the lightest mass

…X), where

X can be the number of steps or the time between updates.

Also fixed tests_openmm_constraints.py

same energies. Also fixed a small bug where setting the dynamics
schedule didn't trigger a change in energy if lambda doesn't change

…dynamic_constraints"

option to mol.dynamics()

…eir r0 lengths

…ions to be

counted twice. They were added to the Custom force used for the perturbable torsions,
but then also added to the PeriodicTorsionForce for non-perturbable torsions.

This is because the incorrect mirrored form of the DihedralID was added to the

parameters which would be changed

…aframe

with energies in reduced units. Also, the sr.morph.to_alchemlyb function
will preserve attributes for the combined dataframe

…lus pert file

…re_force_lever

… into feature_force_lever

… into feature_force_lever

… hand-written ones

This was more painful than I expected... But all working now

…ile and the PerturbableOpenMMMolecule reporting

functions working. Sorting out the function call layout for a tutorial, and also adding unit tests

WIP

…n in the OpenMM context

You can get these via

```
p = mol.perturbation().to_openmm(constraint="...", etc)

print(p.get_changed_atoms())
print(p.get_changed_bonds())
print(p.get_changed_angles())
print(p.get_changed_torsions())
print(p.get_changed_exceptions())
print(p.get_changed_constraints())
```

I have also added a function to zero torsions in the end state where
any atom is a ghost. This is `sire.morph.zero_dummy_torsions`.

…re_force_lever

…use the

new functionality. Now working this up, updating the functionality as needed
to make a nice tutorial.

There are some changes to the code which I've described in the changelog updates.

In particular, the behaviour of ``mol.perturbation().link_to_reference()``
has changed, meaning that you don't need to add a ``.commit()`` on the
end (indeed, it will now give an error).

The right fix for this is to use the new, higher level functions to
link to the reference or perturbed states, e.g.

```
mols = sr.morph.link_to_reference(mols)
```

…onstraints "X",

but will not constrain perturbing bonds *unless* they involve a hydrogen in
any end state.

I have switched this to be the default constraint type, as it is the closest match
to somd1. The real match would be
`constraint="h-bonds", perturbable_constraint="bonds-not-heavy-perturbed"`

I have also improved the code that detects "light" atoms (i.e. hydrogens).

This now uses the maximum mass in any end state being below 2.5, or if
an element in any end state is a H, or if the ambertype in any end state
matches "H*" (case-insensitive).

These three ways of matching are all switched on by default. They can
be disabled via options described in the OpenMM cheatsheet.

…ng (I think) the logic

used in somd1

…nd also

how to set it for specific perturbable molecules.

Also added a record in the schedule for the name of all of the forces,
so these are easier to query.

a previous commit to this feature branch

I've also exposed an option to control the MC barostat frequency

…softening potential.

Set the default to 1 A, so that this matches the somd1 value (and interpretation).

The previous value was equivalent to 10 A.

…gies for

solvated neopentane-methane system

Also updated the somd coulomb constant to be more precise, matching that
used in sire/openmm. This doesn't change any energies, but I am happy
having a good level of consistency.

Also BREAKING CHANGE in the LambdaSchedule API, as described in the changelog

…re_force_lever