Hi all, unfortunately I am not able to add a gloss...
# advice-metadata-modeling
c
Hi all, unfortunately I am not able to add a glossary term to a schema field of a Tableau resource, e.g. a chart or dashboard, via the UI. It's only possible to assign them to the entire entity. From what I understood, https://github.com/datahub-project/datahub/pull/6119 addresses this issue, but has not made it into DataHub. Can anyone give me a status on this or have an idea for a workaround?
b
hey Tim! so the main reason we ended up not merging that PR is because we wanted to avoid confusion around where the source of truth is for those schemas. For Charts and Dashboards, they don't actually own their own schemas, they're just copies from some upstream dataset's schema. so if you started adding descriptions, tags, terms on the Chart, it might be confusing since you're not updating where those columns actually come from.
can you tell me some more about why you want this change and your particular use cases? it would help figure out if there's a workaround or if we should file a feature request to gather more feedback from the community and rethink whether or not this makes sense to finish up
c
Well, that's a solid reasoning. Let me briefly paraphrase our use case: Currently, we only have metadata from Athena and Tableau in DataHub. We already have plans to ingrate other data sources, but have decided to focus on the two existing ones first and provide as many features as possible for them. One of our first steps is to migrate/establish our KPI glossary to/in DataHub. We don't want to link glossary terms to an entire Tableau entity because it typically contains multiple KPIs, such as reports modeled as charts in DataHub. Therefore, we rely on the ability to add glossary terms to individual schema fields of these types of entities. The above reasoning is fine as long as there is a way to automatically transfer glossary terms from upstream datasets. If that were the case, there would be no reason for us to complain 🙂 There is already a feature request for that, right?
Another aspect here is (maybe its's tied to Tableau) that the delivered schema descriptions are not ideal. Often they are rather longish and complex looking formulas in Tableau's internal syntax. So it might be reasonable to change them. Therefore, I would suggest to consider to somehow allow for changing those schemas. Might be something that you need to enable/configure explicitly.
b
okay gotcha! thanks so much Tim, this really helps a lot. I could definitely see other features solving this problem, like you said propagating metadata like glossary terms downstream to BI dashboard and chart schemas which is definitely a feature request. However I think it makes sense to fill out a feature request just to get this documented and see how other people feel. Would you mind creating one at this link? https://feature-requests.datahubproject.io/ honestly you could basically just copy what you have above as reasoning as I think that provides some nice context!
c
Alright, I'll do so Thanks for your support 🙂
b
of course! thanks for all your great feedback!
c
I would have one additional question: In case we would like to have a short term solution and modify DataHub's behavior in our fork, could we just apply the changes from the PR above or is it outdated in the mean time? 🙂
b
yeah I think that would be fine! there might some conflicts with more recent changes that have gone out but overall I think you should be able to take the diff from the PR and apply it to your fork.
c
Great, thanks a lot 🙂
b
awesome thank you!