Unpublish Dataset: Create an API endpoint to unpublish a dataset

[kcondon]

We need an API endpoint to unpublish datasets in some cases where a user may mistakenly publish their dataset in root when it needs to be reviewed first in by a journal dataverse.

Here are the essential steps by hand that hopefully will become the endpoint:

Tables involved:
Dataset
Datasetlock
Datasetversion
Dvobject

To Unpublish a Dataset:

1. Lock dataset
   insert into datasetlock (reason, starttime, dataset_id, user_id) values ('InReview', now(), 3092, 7);
2. (N/A) Reset DOIs to reserved, both dataset and files
   log on to ezid console, manage dois, set dois from public to unavailable
   Not really needed: would resolve to need perms link and works again once published.
3. Change datasetversion table: releasetime, versionstate, versionnumber, minorversionnumber
   update datasetversion set versionstate='DRAFT', releasetime=NULL, versionnumber=NULL, minorversionnumber=NULL where dataset_id=3092 and id=98;
4. Change dvobject table for dataset: publicationdate, globalidcreatetime
5. (N/A) Change dvobject table for datafiles: same as for dataset
   update dvobject set publicationdate=NULL, globalidcreatetime=NULL where id in (3093, 3094, 3095);
   Not really needed, no functional impact and ezid has original create time stored.
6. Change dataset table: lastexporttime
   update dataset set lastexporttime=NULL where id=3092;

[kmika11]

Would this include the ability to "unpublish" datasets that are deaccessioned?

[donsizemore]

This would be a nice feature to use against suspected spam or pseudo-science datasets; this, and the ability to prevent future publishing, which could be overridden by a superuser?

[pdurbin]

Related:

• https://groups.google.com/g/dataverse-community/c/C6hl_mTx1yM/m/m8sxdnQ6AQAJ

[DonRichards]

Just wondering, but is the best alternative to deaccessioned the dataset and publish a new version later? Suggestion from this discussion. https://groups.google.com/g/dataverse-community/c/C6hl_mTx1yM/m/XM75Qy5GAQAJ

[pdurbin]

@DonRichards yes, that's a completely valid and 100% supported approach.

[cmbz]

2024/09/16: ABR review, there is a manual process, but worth investigating for API. @sbarbosadataverse please weigh in.

[landreev]

We absolutely need to prioritize this.
Yes, there is a process - but it is a direct database access hack; both inherently unsafe and requiring somebody with ssh access to have to do this all the time.
Whatever the initial rationale was for not providing a supported mechanism for this - I'm assuming it was simply for the sake of being able to tell users "once published, it cannot be undone" on the dataset page, without being liars - it is time to accept the reality of the situation. It keeps happening - seemingly more and more often - and it's not sustainable.
If we have an admin-only "destroy" API, there's no good reason not to have an admin-only "unpublish".