Reviewer 2 request 1 changes for improved instructions on installing GOMC package

[CalCraven]

Add installation instructions for GOMC in README and docs, as well as… fix tests for improved error messages and automated checking of PATH for GOMC installation

[bc118]

I took a quick look at this and here is my recommendations, All tested in RHEL:

@CalCraven Another thought and I actually like this better . We may want to avoid building it in "opt" as many not experience users will not be able to get rid of it. Why not build it in the actual conda env directly and symlink it to the conda bin directory.

This should also eliminate having to modify their bash.rc (meaning not messing it up and likely not differences between Mac and Ubuntu and RHEL).

Note the subprocess command. Despite the GOMC binary (GOMC_CPU_NVT) being available in my terminal through the auto-path setup when activating themosdef_dihedral_fitenv, it still required python to run this subprocess command to be able to access the path auto-setup conda env path. This may be the reason for the Mac vs. Ubuntu error, but not 100% sure.

I would add these instructions directly to the mosdef_dihedral_fit docs, and then in the error if GOMC binary (GOMC_CPU_NVT) not found, reference back to these docs. We should forward them to our docs which shows the simple and minimal install. Referencing GOMC docs is fine, but we should try not to get them to build it via the GOMC docs, as it is complicated, more unneeded disk space, and likely will result in more errors from building with GPUs.

This will need changed in a few files with checks/errors if

Code and notes :

Need to add to all effected python files:
import subprocess
**In the dihedral fit code, if “gomc_binary_directory”=None (if not default, make it so). We may want to check if conda exists first or print an error if this command fails maybe try/except if statement to see if it is in the path or the local path (=“”) to get those who have it there (this will eliminate issue with the Docker build(s) also)
gomc_binary_directory=
if gomc_binary_directory=None:
gomc_binary_directory = subprocess.check_output("echo ${CONDA_PREFIX}/bin", shell=True, text=True).strip()

Setup the GOMC binary (GOMC_CPU_NVT) in the correct:

conda activate mosdef_dihedral_fit
or (from the activated env)
cd $CONDA_PREFIX (env needs to be activated so will only work when in that env, which is likely what we want)

git clone https://github.com/GOMC-WSU/GOMC.git --branch v2.75a
cd GOMC
chmod u+x metamake.sh

These options turn off all but the GOMC_CPU_NVT ensemble, which is all that is really needed and used for mosdef_dihedral_fit. Specify setup is with anaconda, which the only way "mosdef_dihedral_fit" it is currently distributed so we should be fine
./metamake.sh NVT # this only builds the 1 we need, so it will avoid excess space and GPU build errors.

Note: was tested with copy not ls , so may need to change a few Python checks or just use copy (‘cp’ instead of ‘ln -s’)
ln -s $CONDA_PREFIX/GOMC/bin/GOMC_CPU_NVT $CONDA_PREFIX/bin # create a symlink to standard/auto-sourced conda/envs/<active_env_path>

[bc118]

LGTM, just the 1 comment

[bc118]

Should we mention can do 'conda install' but 'mamba' recommended like the docs?

This is the only suggestion I have, otherwise looks good! Good work!

[bc118]

I missed the conda mention. above. Approved