@green-football-43791 I threw together this PR: https://github.com/linkedin/datahub/pull/2320 Let me know if i’m missing something here.

For context, the problem I was seeing is an error like this when trying to create a new tag:

java.lang.IllegalArgumentException: com.linkedin.common.urn.TagUrn is not a supported URN type

Also, in the current implementation, ingesting tags on datasets seems to overwrite any tags set through the UI. Am I missing something?

regarding overwriting, you're correct- on the entity level at the moment we only have 1 tag aspect. given the current state, the approach I'd recommend is fetching the tags of the entity from datahub on writes, appending your new tag to the existing array, and writing back the newly constructed array. However, there are some solutions: 1. create a second tag aspect "EditableGlobalTags". Edits from the UI will update this aspect while edits from ingestion will update GlobalTags. This is the same pattern we used for adding tags to schemas so tags wouldn't be overwritten any time the schema was re-ingested. 2. Develop a more general solution to the problem of multiple writers. As datahub develops, I foresee an increase in writes coming from the UI. Ideally we wouldn't need to create an

EditableXYZ

clone of every aspect. Perhaps the solution here is to add a new flag to our metadata objects of whether they are the ui-editable version or not. In addition, if we could come up with a smart, more generic way of merging the editable and non-editable aspects, this would save us even more headache.

3. A third option could be exposing some APIs akin to

addTag(urn)

and

removeTag(urn)

- this could free any given writer from needing to understand the bigger picture about an entity