openff-units

A common units module for the OpenFF software stack

Please note that there may still be some changes to the API prior to a stable 1.0.0 release.

This package provides a common unit registry for all OpenFF packages to use in order to ensure consistent unit definitions across the software ecosystem.

The unit definitions are currently sourced from the NIST 2018 CODATA, but may be updated in future versions as new CODATA updates are made.

While this repository is based on Pint, the main classes (Unit, Quantity, and Measurement) have been slightly modified in order to provide non-dynamic, more readily serialisable representations.

Installation

Install via conda or a replacement:

conda install openff-units -c conda-forge

Getting Started

Below shows how to tag a number with a unit (generating a Quantity object), get its magnitude with and without units, convert to another unit, and also get its magnitude after converting to another unit.

>>> from openff.units import unit
>>> bond_length = 1.4 * unit.angstrom
>>> bond_length
<Quantity(1.4, 'angstrom')>
>>> bond_length.magnitude
1.4
>>> bond_length.to(unit.nanometer)
<Quantity(0.14, 'nanometer')>
>>> bond_length.magnitude_as(unit.nanometer)
0.14

One could also do the Pint tutorial using the unit object above as a drop-in replacement for ureg in the tutorial.

Serialization

Scalar quantities can be serialized to strings using the built-in str() function and deserialized using the unit.Quantity constructor.

>>> k = 10 * unit.kilocalorie / unit.mol / unit.nanometer**2
>>> k
<Quantity(10.0, 'kilocalorie / mole / nanometer ** 2')>
>>> str(k)
'10.0 kcal / mol / nm ** 2'
>>> unit.Quantity(str(k))
<Quantity(10.0, 'kilocalorie / mole / nanometer ** 2')>

OpenMM Interoperability

For compatibility with OpenMM units, a submodule (openff.units.openmm) with conversion functions (to_openmm, from_openmm) is also provided.

>>> from openff.units import unit
>>> from openff.units.openmm import to_openmm, from_openmm
>>> distance = 24.0 * unit.meter
>>> converted = to_openmm(distance)
>>> converted
Quantity(value=24.0, unit=meter)
>>> type(converted)
<class 'openmm.unit.quantity.Quantity'>
>>> roundtripped = from_openmm(converted)
>>> roundtripped
<Quantity(24.0, 'meter')>
>>> type(roundtripped)
<class 'openff.units.units.Quantity'>

An effort is made to convert from OpenMM constructs, such as when OpenMM provides array-like data as a list of Vec3 objects:into Pint's wrapped NumPy arrays:

>>> from openmm import app
>>> positions = app.PDBFile("top.pdb").getPositions()
>>> positions
Quantity(value=[Vec3(x=-0.07890000000000001, y=-0.0198, z=-0.0), Vec3(x=-0.0006000000000000001, y=0.039200000000000006, z=-0.0), Vec3(x=0.07950000000000002, y=-0.0194, z=0.0), Vec3(x=0.9211, y=0.9802, z=1.0), Vec3(x=0.9994000000000001, y=1.0392, z=1.0), Vec3(x=1.0795000000000001, y=0.9805999999999999, z=1.0)], unit=nanometer)
>>> type(positions)
<class 'openmm.unit.quantity.Quantity'>
>>> type(positions._value)
<class 'list'>
>>> type(positions._value[0])
<class 'openmm.vec3.Vec3'>
>>> converted = from_openmm(positions)
>>> converted
<Quantity([[-7.8900e-02 -1.9800e-02 -0.0000e+00]
 [-6.0000e-04  3.9200e-02 -0.0000e+00]
 [ 7.9500e-02 -1.9400e-02  0.0000e+00]
 [ 9.2110e-01  9.8020e-01  1.0000e+00]
 [ 9.9940e-01  1.0392e+00  1.0000e+00]
 [ 1.0795e+00  9.8060e-01  1.0000e+00]], 'nanometer')>
>>> converted.m
array([[-7.8900e-02, -1.9800e-02, -0.0000e+00],
       [-6.0000e-04,  3.9200e-02, -0.0000e+00],
       [ 7.9500e-02, -1.9400e-02,  0.0000e+00],
       [ 9.2110e-01,  9.8020e-01,  1.0000e+00],
       [ 9.9940e-01,  1.0392e+00,  1.0000e+00],
       [ 1.0795e+00,  9.8060e-01,  1.0000e+00]])
>>> type(converted)
<class 'openff.units.units.Quantity'>
>>> type(converted.m)
<class 'numpy.ndarray'>

Dealing with multiple unit packages

You may find yourself needing to normalize a quantity to a particular unit package, while accepting inputs from either openff.units or openmm.unit. The ensure_quantity function simplifies this. It takes as arguments a quantity object from either unit solution and a string ("openff" or "openmm") indicating the desired unit type, and returns a quantity from that package. If the quantity argument is already the requested type, the function short-circuits, so it should not introduce substantial overhead compared to simply requiring the target quantity type.

>>> from openff.units import unit, ensure_quantity
>>> ensure_quantity(unit.Quantity(4.0, unit.angstrom), "openff")
<Quantity(4.0, 'angstrom')>  # OpenFF
>>> ensure_quantity(unit.Quantity(4.0, unit.angstrom), "openmm")
Quantity(value=4.0, unit=angstrom)  # OpenMM
>>>
>>> import openmm.unit
>>> ensure_quantity(openmm.unit.Quantity(4.0, openmm.unit.angstrom), "openmm")
Quantity(value=4.0, unit=angstrom)  # OpenMM
>>> ensure_quantity(openmm.unit.Quantity(4.0, openmm.unit.angstrom), "openff")
<Quantity(4.0, 'angstrom')>  # OpenFF

Copyright

Acknowledgements

Project based on the Computational Molecular Science Python Cookiecutter version 1.5.

openforcefield / openff-units

readme