Add ability to delete a published dataset ("destroy")

[pdurbin]

http://guides.dataverse.org/en/4.15/api/native-api.html#delete-published-dataset describes a "destroy" API that allows superusers to delete datasets even after they are published. I can think of a couple use cases for this.

• Harvard Dataverse, for example, allows anyone to create an account and publish datasets immediately. This means that spam datasets get created from time to time and must be deleted.
• While I iterate on https://github.com/IQSS/dataverse-sample-data I'd like to be able to clear out a test server completely from all sample data I created and published. Yesterday I started working on a script for this at https://github.com/IQSS/dataverse-sample-data/commit/0630d67 but it currently requires that all dataset are unpublished.

Here are the curl command examples from the API Guide link above:

Destroy by Persistent ID (PID):

curl -H "X-Dataverse-key:$API_TOKEN" -X DELETE http://$SERVER/api/datasets/:persistentId/destroy/?persistentId=doi:10.5072/FK2/AAA000

Destroy by dataset ID:

curl -H "X-Dataverse-key:$API_TOKEN" -X DELETE http://$SERVER/api/datasets/999/destroy

I'm happy to make a pull request if you'd like. Please let me know.

[skasberger]

This would be an easy first PR to pyDataverse, what do you think @pdurbin ?

Just take the upload_file() function in api.py, copy it and change the query_str and related stuff. :)

[pdurbin]

@skasberger sure, I'll try to make a pull request. Please go ahead an assign this issue to me if you like.

Instead of "upload file" I planned on using "delete dataset" as a starting point:

https://github.com/AUSSDA/pyDataverse/blob/v0.2.1/src/pyDataverse/api.py#L744

[pdurbin]

@skasberger huh. I'm surprised by the Dataverse API behavior. I tried using the pyDataverse "delete dataset" method with a superuser API token on a published dataset and to my surprise the dataset was destroyed!

Here's the code in question: https://github.com/IQSS/dataverse/blob/v4.15/src/main/java/edu/harvard/iq/dataverse/api/Datasets.java#L296

I guess this means that I don't really need a "destroy dataset" method in pyDataverse. I can just use the regular "delete dataset" method as a superuser.

@sekmiller seems to agree with me that the behavior I'm seeing is probably a bug in the Dataverse API. When I use "delete dataset" as a superuser on a published dataset, we think the API should return an error like "Are you sure you want to destroy a published dataset? If so, use the dedicated 'destroy' API endpoint for this!"

The bottom line is that I don't plan to make a pull request for this issue after all. Not anytime soon anyway. Again, I don't need "destroy dataset", surprisingly. 😄

[skasberger]

So, i can close the issue, right? Will there be any updates on the API side, who will fix this in the nearer future?

[pdurbin]

I'll just close this issue. Again, I don't need any changes in pyDataverse. Most people will probably just try "delete dataset" rather than hunting around for "destroy dataset" like I did.

I'm reluctant to change any APIs in Dataverse itself. Maybe this is because I watched "Spec-ulation" by Rich Hickey: https://www.youtube.com/watch?v=oyLBGkS5ICk . Transcript: https://github.com/matthiasn/talk-transcripts/blob/master/Hickey_Rich/Spec_ulation.md

"Now what happens if you just put foo2 next to foo? You can still tell people. They can say, "I am in Bermuda this week, but next week I will try foo2. That sounds awesome." But right now, my web service is going to keep working, because it calls foo, and you did not take it away from me while I was on vacation."