Enhancement
When outputs.yml contains a bfabric_dataset spec, register_outputs does not check whether a dataset with the same name already exists on the workunit. If the output registration is re-run (e.g. after a retry or resume), it creates a duplicate dataset instead of updating the existing one.
Expected
SaveDatasetSpec should support an update_existing field (like CopyResourceSpec and SaveLinkSpec already do) and check for an existing dataset before creating a new one, applying the UpdateExisting policy accordingly.
Current behavior
CopyResourceSpec — checks for existing resource, respects update_existingSaveLinkSpec — checks for existing link, respects update_existingSaveDatasetSpec — always creates a new dataset, no duplicate check
Enhancement
When
outputs.ymlcontains abfabric_datasetspec,register_outputsdoes not check whether a dataset with the same name already exists on the workunit. If the output registration is re-run (e.g. after a retry or resume), it creates a duplicate dataset instead of updating the existing one.Expected
SaveDatasetSpecshould support anupdate_existingfield (likeCopyResourceSpecandSaveLinkSpecalready do) and check for an existing dataset before creating a new one, applying theUpdateExistingpolicy accordingly.Current behavior
CopyResourceSpec— checks for existing resource, respectsupdate_existingSaveLinkSpec— checks for existing link, respectsupdate_existingSaveDatasetSpec— always creates a new dataset, no duplicate check