Huh, I didn't realize this was a feature of DVN 3.x but sure enough "Link Datasets/Studies" appears on the "3.6 & 4.0 Feature Comparison" spreadsheet at https://docs.google.com/spreadsheets/d/1ftHM6E4b9Dft_AvA6KTzeklJyLguItz2_0hNAFL8-VM/edit?usp=sharing

Please note that even superusers cannot delete links. That's what #1364 is about.

Today in Slack @thegaryking brought up this feature, requesting:

trying to add an existing data set to my dataverse... the options here are new dataverse and new data set. can we add ‘existing dataverse’ and ‘existing dataset’?

Also, related to the feature's name, "link" he said:

‘add’ sounds more coherent to the user i think. i’m just adding to my dataverse

All good feedback to consider when adding this feature for all dataverse admin's. Bonus feature request, also from Gary, to have the system suggest datasets to dataverse admin's:

“we found this data set that’s like the ones in your dv, would you like to add it?“

Some related issues to also consider:

• Linked/Saved Search: Review Functionality + Usability Test Linked/Saved Search: Review Functionality + Usability Test #1840
• Saved Search - Improve Performance & User Experience Saved Search - Improve Performance & User Experience #1939
• Linked Search: Creating linked searches fails in production with blank popup due to timeout. Works on smaller db's like demo. Linked Search: Creating linked searches fails in production with blank popup due to timeout. Works on smaller db's like demo. #2167

After much deliberation, some updates to this issue:

• This issue will cover dataset linking. Spike: Dataverse Linking - as a dataverse owner, I want to link a related dataverse to my dataverse to help relevant researchers find it. #4936 covers dataverse linking.

• The main impact of this issue is to open up dataset linking to all logged in users, not just superusers.

• Users who create or are otherwise in control of a dataverse will be able to link to any dataset.

• Specifically, to create a link, a user's account must have an assigned role that includes the "Publish Dataset"(SEK 9/7/18) permission on the dataverse that is DOING the linking.

• A user creating a new link will not need any special permissions on the dataset BEING linked.

• Any user can link to any published dataset.

• For now, there will not be a way to remove links in the UI, but there is a currently existing way to do it via API.

• The success message when you have successfully created a new link should happen in the same place as now, but it should read:

Success! - [linked dataset] has been successfully linked to [linking dataverse]. If you need to remove this link in the future, please contact Support at [support email].

• This feature will use the existing linking UI, with one exception: On the "Link Dataset" modal, we will replace the "Your Dataverses" drop-down list with a type-ahead box.

For future reference, this is our document where the design team deliberated on this feature. Keep in mind that it is not 100% up to date, especially regarding permissions: https://docs.google.com/document/d/14J8uBV4kNPHDfbiumNNKb01tdEksPpnfbRklrY1KhAc/edit

Discussed with UX team, will revise dataset.message.linkSuccess message to not include the support email. Instead will move the "to remove this link..." msg to a help msg that is displayed in the Link Dataverse/Dataset popup, and will have it open the support popup like we do for other msgs.

Just adding here for posterity the decisions made about how the type-ahead field works for finding dataverses to link to: Dataverse allows users to create dataverses with names that have 1 or more characters. But when a user starts typing in the type-ahead field to look for a dataverse to add a dataset link into, the field will start displaying matching dataverse names after 3 characters have been entered into the field. (The component used for the type-ahead field has a default character limit of 2 characters.) Which means that the user won't be able to find dataverses that have names that are fewer than 3 characters. This is fine for Harvard Dataverse because all published dataverses have names (display names, not aliases) that are more than 3 characters (and the unpublished dataverses that have names with fewer characters are obviously test dataverses).

Open questions we've decided to address later:

• Is it reasonable that dataverse names in other languages could be fewer than 3 characters?
• Should Dataverse enforce the number of characters that users can enter for dataverse names (e.g. through field validation), and should setting this character limit be an easily configurable setting for each Dataverse installation?

Added configurable support team name and text link to contact popup to dataset linking popup help msg, removed it from success msg.

Sending back to @sekmiller for a bug fix.

• When using capitalization in the autocomplete input, it can not find the same results that using lower case returns.

@dmurphy @scolapasta @sekmiller @djbrooke
Noticed a couple things that may or may not require a change but does need a ruling so I'm passing back to Dev for comment:

1. Note above says UI linking a published dataset only but API allows unpublished dataset linking (the original implementation). Might be useful in the prepublication author/project dataverses scenario.
   Update: there is a bug here in that it doesn't check perms on src dataset only target dv so can link an unpublished dataset you don't own to your dataverse but can't see it.
2. Dataset Unlinking API exists, why isn't it available to the end user like create link API is? Seems like it would not take much effort. Plus, it would enable users to remove their own links rather than having Support do it.

Just talked about that first point with @djbrooke - let's have both the UI and API require that a dataset is published in order to link it.

We talked about the second point too. For now let's leave the unlinking API as it is, and if we end up getting a ton of support tickets we can revisit.