override_dict accepts a JSON/YAML string with possible unexpected results

[brynpickering]

What happened?

It's technically possible to provide calliope with a JSON/YAML string as the input to override_dict, instead of a dictionary. AttrDict handles this fine, even if it is unexpected. However, it sees it as a pseudo-file and resolves any "imports". This bug corpped up, because the order in which it resolves those imports relative to loading the base file doesn't work as expected (in the case I've seen it on, setting an override on a timeseries filepath at the tech level doesn't filter through).

m = calliope.Model("path/to/model.yaml", override_dict='{"import": ["path/to/relative/import.yaml"]}', scenario="scenario_loaded_from_imported_file")

It's quite niche, you need to only have the model/run config in model.yaml and then everything else loaded as an import via the override dict YAML string as:

def _model_imports():
    return {
        "import" [
            "model_config/locations.yaml",
            "model_config/techs.yaml",
            "scenarios.yaml"
        ],
        "model.timeseries_data_path": f"timeseries_data",
    }
override_dict = json.dumps(_model_imports())
m = calliope.Model("path/to/model.yaml", override_dict=override_dict, scenario="scenario_loaded_from_imported_file")

Which operating systems have you used?

• macOS
• Windows
• Linux

Version

v0.6.10

Relevant log output

No response