Create a python script produces a range of VR input files out of the SA file and a template input

[dalonsoa]

This script, building on the minimal one defined in #240 , will load the SA file, a template config file for VR and create a collection of also valid configuration files differing from one another in the value of just one (or more) parameters. Specific tasks that this script will perform are:

• Read both files (validating the SA one; see Define the format and contents of the sensitivity analysis (SA) input file #240)
• Verify that all parameters to vary defined in the SA file are indeed in the template config file
  • NOTE: As there might not be a single configuration file, but a collection of them, it might be necessary to run the configuration part of VR first in order to get a single config file to use as template.
• Create the set of new parameters following the procedure described in Define the format and contents of the sensitivity analysis (SA) input file #240
• For each set, create a new config file where the relevant parameters are replaced by the new values
• An auxiliary CSV file will be created summarising the parameter sweep, eg.:

Config file	parameter_1	parameter_2
madagascar01.toml	15	20
madagascar02.toml	15	30
madagascar03.toml	15	40
madagascar04.toml	20	20
madagascar05.toml	20	30
madagascar06.toml	20	40
madagascar07.toml	25	20
madagascar08.toml	25	30
madagascar09.toml	25	40

The naming convention of the new config files will be such that they can be easily ingested by the submission script for the Imperial's HPC (see #242 ).

Question:

• How would be best to handle the case of configuration spread across multiple config files?

[jacobcook1995]

I think initialising a Config and then using the methods defined there might be the simplest way of combining the config files (including the validation step validate_config).

I would mainly recommend this as the configuration schema define default values for a lot of parameters, which get added into the config when the validation gets carried out. This means we usually won't be using config files that define every parameter as in a lot of cases we will just want to use the default values. I can't see an issue in us generating full config files for the purpose of parameter sensitivity analysis though, if that turns out to be simpler.

[alexdewar]

Given that I have already effectively tackled this using Snakemake, I think this is now obsolete.