Consistent Primary/Secondary Assignment Behavior

[bencap]

For certain resources like Experiments and Score Sets, items like publication identifiers may be assigned as either primary or secondary. In the UI, when a user provides a single publication it is automatically assigned as the primary publication for that resource. In the UI, there is no separate property surfaced to the user for primary or secondary publications. In the API, no such automatic coercion of the primary status exists, as the list of primary and secondary publications are provided separately.

We should consider either API constraints which guarantee the consistency of the behavior enforced by the UI, or make the UI choices clear to the user. We might either:

• Add a constraint to the Score Set and Experiment models which enforces the primary publication list is non-empty when at least one publication is provided.
• Display the primary publication as being selected and editable to the user in the UI once a single publication is submitted, so they may change the attribute if desired.

My view is that given each MaveDB record should at minimum cite its relevant published or pre-printed work, we should enforce that each record which has a publication identifier must also have an associated primary publication and make it clear in the documentation this is intended to be the contribution to the scientific literature to which this Experiment/Score Set belongs. Ie: The only way for a publication to not have an associated primary publication is for that publication to not have any associated publications.

If we decide against such a constraint, we should at least make the UI compatible with allowing a single publication to be associated with a record and not have it be the primary publication.

[afrubin]

We do have some datasets that don't have any associated publication (and may never have one), so we do need to allow the case where there are no publications as described.

I agree that the hypothetical case where there's no primary publication describing the data but the user wants to put in secondary publications is a rare edge case that is not worth retaining. I don't think there are any such datasets currently, and if it becomes an issue we can revisit this.

Arguably allowing a secondary publication for a dataset with no associated literature is a UI issue itself, since it will be less clear to the user that the data is unpublished, which is just another reason to adopt the above constraint.