It feels weird to have a mix of object oriented design with the current way to attach runs to studies (or tasks to suites).
Instead of:
study.attach_to_study(study_id, runs_ids)
I'd expect to be able to do:
OpenMLStudyObject.attach_runs(run_ids)
However, at the same time as far as I can currently remember all methods on objects are local (except for publish) and this would go against that. This is natural as all other entities are primarily immutable (edits are also done through external functions), whereas for studies (and to some extent suites) updates seem natural. E.g.: (pre-release) a suite to verify everything works as intended (see automl benchmark), or add runs to a study as runs are completed on a cluster.
It feels weird to have a mix of object oriented design with the current way to attach runs to studies (or tasks to suites).
Instead of:
I'd expect to be able to do:
However, at the same time as far as I can currently remember all methods on objects are local (except for
publish) and this would go against that. This is natural as all other entities are primarily immutable (edits are also done through external functions), whereas for studies (and to some extent suites) updates seem natural. E.g.: (pre-release) a suite to verify everything works as intended (see automl benchmark), or add runs to a study as runs are completed on a cluster.