travis error about torsionfit HOT 8 CLOSED

Do you get this error after normal installation with setup.py?

from torsionfit.

The issue seems to be with this line:

  File "/Users/choderaj/miniconda/envs/_test/lib/python2.7/site-packages/torsionfit/TorsionScanSet.py", line 252, in compute_energy
    self.mm_energy -= self.mm_energy.min() + Quantity(value=float(offset.value), unit=kilojoules_per_mole)

It could be that there is a weird numpy/simtk.unit.Quantity issue with the .min() operation. Is self.mm_energy a Quantity wrapping numpy array?

from torsionfit.

I did conda build devtools/conda-recipe to replicate what is done on travis.

from torsionfit.

It could be that there is a weird numpy/simtk.unit.Quantity issue with the .min() operation. Is self.mm_energy a Quantity wrapping numpy array?

That can be. Maybe we can change this too self.mm_energy._value for the numpy array without the Quantity wrapper?

from torsionfit.

That can be. Maybe we can change this too self.mm_energy._value for the numpy array without the Quantity wrapper?

It's best to never access the value without converting to units first. It's much better to pick a standard unit and explicitly convert to that:

energy_unit = unit.kilocalories_per_mole
self.mm_energy -= unit.Quantity((self.mm_energy / energy_unit).min() + (offset/energy_unit), energy_unit)

but we really shouldn't need to be doing all of those weird conversions. It should "just work" unless there is a bug in simtk.unit.

from torsionfit.

but we really shouldn't need to be doing all of those weird conversions. It should "just work" unless there is a bug in simtk.unit.

Yeah. The version of OpenMM that I'm using is 6.3 and I don't get this failure. What's the version of OpenMM you got with conda build devtools/conda-recipe

from torsionfit.

You can see my changes to the conda dev recipe here: #28

from torsionfit.

Closing. Resolved with #29

from torsionfit.