GitLab

Apparently it is the slightly more correct way of doing thigns, as explained in
https://stackoverflow.com/a/30676267/538973

* Use list-comprehension in Python code

A minor change, but slighly easier to understand the initialization of
`parent_exclude_list` in my opinion.

* Implement __ne__ in Python classes that has __eq__

In Python 3, `__ne__` is automatically implemented as `not __eq__`.

However, in Python 2 it seems to be implemented as `not is` (so based on object
identity).

Based on setup.py [0] which says that "OpenMM requires Python 2.7 or better", it
should be useful to have better support for Python 2 :)

This was already done in 4 of the 12 classes that implements `__eq__`

```
>>> class WildCard(object):
...     def __eq__(self, other): return True

>>> w = WildCard()

>>> w == 42
True

>>> w != 42
True

>>> w != w
False
```

[0]: https://github.com/openmm/openmm/blob/5cef29ce8d4e17b6d0f9fd8a3c6cc1669ca0ba2b/wrappers/python/setup.py#L237

* Use umambiguous floor division for index calculations in Python

This makes the code work as intended if run as Python 3

```
$ python2 -c 'print(3/2, 3//2)'
(1, 1)
$ python3 -c 'print(3/2, 3//2)'
1.5 1
```

* Use `with` for file handling in Python

* Use `is None` instead of `== None` in Python

This is recommended in PEP8:

> Comparisons to singletons like None should always be done with is or is not, never the equality operators.
> - https://www.python.org/dev/peps/pep-0008/#programming-recommendations

Raise exception if large differences are observed so simtk.testInstallation can be used in CI testing reliably

Add new 'smirnoff' mode to ForceField improper torsion handling behaviors to allow SMIRNOFF trefoil torsions to be added correctly

The following lipid were added to the addMembrane() method: DLPC,
DLPE, DMPC, DOPC, DPPC.

Original pdb files were taken from
http://www.fos.su.se/~sasha/SLipids/Downloads.html .

Original files were obtained at temperatures between 293K and 303K.
NPT MD simulations were run for all the systems at 298K for ~10 ns
and the last simulation  step was used to create the pdb files.

Within Modeller.addExtraParticles(), add the ability to pass down
an option to modify compiled.matchResidueToTemplate() 's
ignoreExternalBonds parameter.

https://simtk.org/plugins/phpBB/viewtopicPhpbb.php?f=161&t=11264&p=0&start=0&view=&sid=e384beeb08f3bfd767dcfb546006ad63

Fixes writing invalid PDBx/mmCIF files when residue insertion codes are blank (not empty) in Topology

From @peastman:

> Test cases are failing. I believe it's because the parent for a ScaledUnit needs to be a Unit, not another ScaledUnit. So in the line
>
> ```
>  hartree_base_unit = ScaledUnit(4.3597447222071e-18, joule_base_unit, "hartree", "Ha")
> ```
>
> replace joule_base_unit with joule.

This is of immediate need to QCEngine, in particular MolSSI/QCEngine#151