validate API endpoint doesn't enforce DataCite requirements

[donsizemore]

Possibly related to #6380 though the Dataverse instance in question is still on 4.16. It is possible to create a dataset with metadata which passes validation via endpoint:

$ curl -H "X-Dataverse-key:" "http://localhost:8080/api/admin/validate/dataset/556?variables=true"
{"status":"OK","data":{"message":"valid"}}

but lacks certain fields required by DataCite:

Response code: 422, [facet 'minLength'] The value has a length of '0'; this underruns the allowed minimum length of '1'. at line 31, column 0]]

In our case this turned out to be a producer URL not being accompanied by a producer name, but with only the above to go on from DataCite, this took a while to find.

If catching such a condition is outside the scope of the validate endpoint and/or I'm comparing tangerines to clementines, feel free to close this issue.

[jggautier]

Hi Don. Your mention of producer URL makes me think you're referring to the Producer Name and URL fields in the Citation metadata block:

Is that right?

[pdurbin]

@jggautier sounds right but let's let @donsizemore confirm. 😄

@donsizemore I assume that minLength error is on publish?

[donsizemore]

@jggautier @pdurbin correct on both points.

[jggautier]

Thanks. This sounds like the problem described in this GitHub issue. It still happens in Dataverse 4.19, but I'm guessing it hasn't come up often because people haven't been trying to publish datasets where the Producer URL field has a value but the Producer Name field does not.

[qqmyers]

FWIW - I think I played around at one point with the idea of being able to make a metadatablock child entry 'required' with the parent 'optional', so one could potentially require a producer name and url if one is entered (if I understand the structure of producer). If that's useful, let me know. (I think I had code - it may even be in qdr's fork.)

[landreev]

I agree it would be useful to be able to validate for things like this - but should it be a different API? This "admin/validate/dataset" API was only meant for database constraints; all it does it tries to instantiate entities and checks for ConstraintViolationExceptions.

[djbrooke]

Closing in favor of #7551