synthetic contirbutions

[younesStrittmatter]

Should we use the following format for synthetic runners that are not pulled into autora-synthetic:

autora[experiment-runner-synthetic-...]

or

autora[synthetic-...]

?
The first one aligns with the import but is longer. For example, the import for the abstract equation package looks like this:

from autora.experiment_runner.synthetic.abstract.equation import equation_experiment

[younesStrittmatter]

Related:
we have optional dependencies for experiment-runners, should we add the synthetic ones as optional dependencies for the runners, or for a new category (synthetic) or for a subcategory of experiment-runners? Or are there other ideas?

[chadcwilliams]

I prefer autora[synthetic-...] however I think sticking to our formatting is the better idea here: autora[experiment-runner-synthetic-...] despite it leading to larger import statements.

One way to make the import statements slightly shorter, is to allow for the dropping of the synthetic category on (e.g., abstract, psychology):
from autora.experiment_runner.synthetic.abstract.equation import equation_experiment
becomes
from autora.experiment_runner.synthetic import equation_experiment

That's not to say that the categories do not exist, but rather you can import them using either method above. So, we'd need an __init__ file in the synthetic folder that accumulates all runners across all subfolders as a shortcut.

[younesStrittmatter]

not sure how the dropping of the category would work for optional dependencies though. I honestly think it wouldn't (or at least not straightforward)

[chadcwilliams]

I just mean for importing, not installing.

[younesStrittmatter]

Decision:
Full path in both the import and installation, make convenience optional dependencies:

• experiment-runners-all
• experiment-runners-synthetic-all