Repository Structure

The project is organised with the following components,

.
├── conda                                 # conda environment
│   ├── environment.yml
│   └── meta.yaml
├── COPYRIGHT.txt
├── examples
│   └── Experiment_generator_example.yaml # An example YAML
├── LICENSE
├── pyproject.toml                        # package metadata, deps, console_scripts
├── README.md
├── setup.py                              # conda recipe versioning
├── src                                 
│   └── experiment_generator              # source code
│       ├── base_experiment.py            # Common path+repo config helpers
│       ├── config_updater.py             # Payu configuration updater
│       ├── experiment_generator.py       # Orchestrator (end-to-end flow)
│       ├── f90nml_updater.py             # Fortran namelist configuration updater
│       ├── main.py                       # CLI entrypoint - loads YAML, runs Orchestrator
│       ├── mom6_input_updater.py         # MOM6 configuration updater
│       ├── nuopc_runconfig_updater.py    # nuopc.runconfig configuration updater
│       ├── nuopc_runseq_updater.py       # nuopc.runseq configuration updater
│       ├── perturbation_experiment.py    # Houses control/perturbation setups
│       ├── tmp_parser/                   # temporary configuration parsers, this will be replaced by https://github.com/ACCESS-NRI/access-parsers
│       └── utils.py                      # Helpers
├── .github/workflows/                    # CI/CD
└── tests                                 # Unittests suite

6 directories, 35 files

Some major components are explained a bit as follows,

• base_experiment.py: A helper base class that parses general settings from the input dictionary and provides common attributes (such as paths and repository info) used by both control and perturbation workflows.
• experiment_generator.py: The high-level orchestrator that sets up the experiments. It inherits from BaseExperiment and coordinates the overall workflow.
• perturbation_experiment.py: Manages the creation of the control experiment and all perturbation experiments. This class extends BaseExperiment, so it has access to the same configuration attributes. It provides methods like manage_control_expt() for applying parameter updates to the control branch, and manage_perturb_expt() for looping through each defined perturbation, creating branches and applying changes.

  • Inside it, there is a simple data structure (ExperimentDefinition) that represents each perturbation experiment definition. Each instance holds,

    1. block name (a label/group from the YAML, e.g. Parameter_block1),
    2. target branch name for that experiment,
    3. a dict of file parameters to change.

  • The generator flattens the nested YAML definitions into a list of ExperimentDefinition objects, each corresponding to one experiment case, before actually creating branches and applying changes.