Add WaterCluster testsystem

[maxentile]

Creates a few waters in a harmonic restraining potential.

Implementation draws on WaterBox and LennardJonesCluster.

This was suggested by @jchodera as a testsystem that would allow a smaller number of water molecules than possible in WaterBox.

[jchodera]

Looks like there are some travis failures:

======================================================================
ERROR: Failure: NameError (global name 'm' is not defined)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/miniconda/envs/test/lib/python2.7/site-packages/nose/loader.py", line 251, in generate
    for test in g():
  File "/home/travis/miniconda/envs/test/lib/python2.7/site-packages/openmmtools-0.13.5-py2.7.egg/openmmtools/tests/test_integrators_and_testsystems.py", line 88, in test_integrators_and_testsystems
    testsystem = getattr(testsystems, testsystem_name)()
  File "/home/travis/miniconda/envs/test/lib/python2.7/site-packages/openmmtools-0.13.5-py2.7.egg/openmmtools/testsystems.py", line 1778, in __init__
    new_pos = m.getPositions()
NameError: global name 'm' is not defined

[jchodera]

I've checked in a fix for the typo.

[jchodera]

For some reason, the test for is_quantity_close() is failing:

======================================================================
FAIL: Test is_quantity_close method.
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/miniconda/envs/test/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/miniconda/envs/test/lib/python2.7/site-packages/openmmtools-0.13.5-py2.7.egg/openmmtools/tests/test_utils.py", line 82, in test_is_quantity_close
    assert is_quantity_close(quantity1, quantity2) is test_result
AssertionError

[maxentile]

Ahh, thanks!

Looking at the is_quantity_close test...

[maxentile]

On all 3 Travis builds, it looks like the test that checks whether 300.0 K is close to 300.000000004 K is failing.

However, running nosetests test_utils.py locally doesn't reproduce this problem -- I wonder if the behavior of numpy.isclose changed recently...

[maxentile]

I wonder if the behavior of numpy.isclose changed recently...

I think that may be the culprit.

Conda-upgrading numpy: 1.12.1-py36_blas_openblas_200 conda-forge [blas_openblas] --> 1.12.1-py36_blas_openblas_201 conda-forge [blas_openblas]

Changes the behavior of

import numpy as np
value1 = 300
value2 = 300.00000004
np.isclose(value1, value2, rtol=1e-10, atol=0.0) # True before, False now
np.isclose(value2, value1, rtol=1e-10, atol=0.0) # True before, False now

Propose we loosen the test

[maxentile]

Okay that didn't fix it-- the test is still failing on travis but passing locally...

Reverting previous commit...

[jchodera]

NICE CATCH! That was driving me crazy.

[maxentile]

That was @andrrizzi ! He identified this issue in 00d2df1 and pointed me to the fix.

[maxentile]

Tests pass, should be good to merge.

[jchodera]

Thanks! I'll add to the docs and changelog and then merge.

[maxentile]

Oh! I forgot to do that -- no need for you to do that, I'll update the docs and changelog

[jchodera]

We should really add pull request checklists. I forget all the time too.

I'd add the water box here and create a new changelog entry above here called "development snapshot".

[jchodera]

Thanks!