NOTES on installing OPENMM in a *virtualenv* instead of a Conda env

OpenMM project URL: https://openmm.org/
What is it?
OpenMM is a nice python toolkit for molecular dynamics. They provide installation via Conda.
But regular old virtualenvs usually "just work" and have many advantages.

Advantages of virtualenv: it's not Conda!

Hence these rough notes. They should provide a high-level guide to getting openmm in a virtualenv.

we'll have to compile from source
http://docs.openmm.org/latest/userguide/library/02_compiling.html#compiling-openmm-from-source-code

but there are some modifications to make this work with a pip virtualenv rather than conda
don't worry, it's not too hard.

* FIRST: make sure to activate the environment before doing anything else! (> source bin/activate)
* Make a src/ folder for the environment, in your virtualenv's root dir.
* Download source for openmm from https://github.com/openmm/openmm/releases
    * extract into src/ folder
* will need cmake and ccmake, these can be installed through apt if don't already have
* Download doxygen https://www.doxygen.nl/download.html
    * extract into src/ folder
    * find the already compiled doxygen executable in bin/, run pwd to take note of the path to it
    * something like: /home/me/env_root/src/doxygen-1.10.0/bin/doxygen
* you'll need swig. fortunately, it can be obtained via pypi package
    * https://pypi.org/project/swig/
    * > pip install swig
    * run and take note for future use: > which swig
* you'll also need cython
    * > pip install Cython --install-option="--no-cython-compile"
* We'll have to pick a destination for openmm. Start from virtualenv root dir
    * > mkdir builds
    * > cd builds
    * > mkdir openmm
    * > cd openmm
    * run pwd, and take note of the result, this is our destination path
    * something like: /home/me/env_root/builds/openmm
* okay, now you can (from the downloaded openmm folder) run the given instructions for build from source
    * > mkdir build
    * > cd build
    * > ccmake ..
    * Great! But we'll need to fix some ccmake variables first!
        * first we've got to pick the install location
        * set CMAKE_INSTALL_PREFIX to the destination path created above
    * set DOXYGEN_EXECUTABLE to the location we found above
    * set SWIG_EXECUTABLE to $(which swig)
    * if you can't see any ccmake variables?
        * first try toggling to advanced mode by pressing [t]
        * otherwise, seems like maybe [c] and having errors produced and then retrying works???
    * if we want CUDA, it's probably good to check that the appropriate variables are set (eg. path)
* now we can proceed with the rest of the installation: press [g] to generate
* a makefile should have been generated, run it0: > make
* then install:                                   > make install
* then install python bindings:                   > make PythonInstall