Example of reading a "settings" python module

mysettings.py

"""A settings module based on global variables"""

name = "Custom Settings"
x = 4
y = 0

read_settings_py.py

"""main module

usage: python read_settings_py.py mysettings

This will import settings from mysettings.py
"""
import argparse
import importlib


def main(args=None):
    parser = argparse.ArgumentParser(description="")
    parser.add_argument("settings", help="settings file")
    args = parser.parse_args(args)

    # Dynamic import based on the command line argument
    settings = importlib.import_module(args.settings)

    print("Name: %s" % settings.name)
    print("Coords: %d,%d" % (settings.x, settings.y))


if __name__ == "__main__":
    main()

submit_py.sh

#!/bin/bash
# Slurm submission script
# usage: sbatch submit_py.sh mysettings

#SBATCH -p hourly
#SBATCH -n 1
#SBATCH --account=merlin

module purge
module load anaconda
conda activate dev27

# pass sbatch arguments to python
python read_settings_py.py "$@"

Here is an example of submitting a slurm job with an argument. It uses a bare module to store settings, similar to the current implementation.

• Call sbatch submit_py.sh mysettings
• Inside submit_py.sh, mysettings is stored as $@ (standard bash argument handling)
• On the compute node, submit_py.sh calls python read_settings_py.py mysettings
• read_settings_py.py gets "mysettings" passed as sys.argv[1]. This is parsed by the cli library (here the built-in argparse, but click is nicer) and stored as args.settings
• The module gets dynamically imported with importlib.import_module. This line would be equivalent to import mysettings. It would search sys.path and find ./mysettings.py

Using python modules for storing settings is not recommended. The official way of storing settings is in ini files. Here's a better implementation: https://gist.github.com/sbliven/10c3ddabcb32301b9fb475748ed75e35