Upon meeting with Matthew where we presented the latest state of the mvesuvio repository, it became clear that the command line interface is not suitable for his work.
Matthew feels this complicates the workflow of creating directories rather than simplifying it.
Matthew expressed that using the command line to set up experiments (i.e. 4/5 experiments during beam time) is not suitable for him. He mentioned that not only the manual effort and time is at stake, but also the mental effort he puts in ensuring he remembers the commands on the command line.
The improvement I have in mind is to make it as simple as manually copying and pasting the inputs script to create new results directories.
Now that there is a cache system that keeps track of the input paths required, the improvement should be as simple as making sure that when a input script is run, its path is stored on the cache system.
All input and output directory paths will be based on the cache system, rather than on relative paths. This should solve all of the issues Matthew is currently experiencing with output paths being created on the wrong directories.
Key non-negotiables
Creation of directories should be no more effortful than is was in the previous repository (a single copy-and-paste of the inputs script)
Running of the routine should be triggered from within the input scripts
Code-base should only be kept in a single place (inside the conda environment)
Output directories should be easily accessible from the directory where the corresponding input script was run
Upon meeting with Matthew where we presented the latest state of the mvesuvio repository, it became clear that the command line interface is not suitable for his work.
Matthew feels this complicates the workflow of creating directories rather than simplifying it.
Matthew expressed that using the command line to set up experiments (i.e. 4/5 experiments during beam time) is not suitable for him. He mentioned that not only the manual effort and time is at stake, but also the mental effort he puts in ensuring he remembers the commands on the command line.
The improvement I have in mind is to make it as simple as manually copying and pasting the inputs script to create new results directories. Now that there is a cache system that keeps track of the input paths required, the improvement should be as simple as making sure that when a input script is run, its path is stored on the cache system. All input and output directory paths will be based on the cache system, rather than on relative paths. This should solve all of the issues Matthew is currently experiencing with output paths being created on the wrong directories.
Key non-negotiables