Metadata types

I can’t find the details of this in the documentation. Not sure if I need to be pointed in the right direction, or if it’s something that needs doing.

I’m editing the metadata of my database. There is now a long list of metadata types including stuff like ‘Owner’, ‘Creation Time Stamp’, ‘Profit’.
Is there any documentation describing the impact of these? For example, how is it different to define a column as ‘Owner’ rather than ‘Category’?

1 Like


I haven’t seen anything in the documentation either. But I also noticed some to me: “new”? types popped up recently. I guess best “documentation” currently is what’s in the source code with those few comments the types.clj file has. Here’s a link to the “blame” view of that file on the master branch:

A git blame view has the added benefit that you can see who changed what when - and on GitHub it’s easy to navigate into the code-changes, and the issues/features that got resolved. I guess most recent changes to the type system has to do with the X-ray features. I’d gather the more relevant names/types you tag stuff with the better the “automagic” X-ray explorations can categorize and name stuff. (In the end there is no magic - just clever stuff).

Here are the X-ray rules: - I guess what what you’re specifically asking about is in the the /fields folder there.

Maybe someone on the Metabase core team can enlighten us a bit on where this came from, why/how it changed recently and where they see it going?

1 Like

It’s only Wednesday and you’ve already spoilt my week. :sob::sob:

1 Like

Wednesday? Pretty sure it’s Thursday here. Are you not fully awake yet or already into the stronger beverages? :rofl:

Lol! My wife normally works Mon-Weds. This week it’s Tues-Thurs, hence I 've been a day out all week.

1 Like

We are slowly adding to our type system so that we can capture the semantics of your data better in turn allowing us to be smarter with defaults and doing better automatic analysis via x-rays. The impact is in flux (as are x-rays in general) hence no documentation (and especially no commitment). On a conceptual level we’d like to:

  • surface the most important fields in your data and based on that the most important dashboards for you
  • automatically generate complex metrics such as DAU, MRR, churn, …
  • recommend segmentations
  • understand the topology of your data warehouse and semantics of your business (eg. this looks to be a subscription business, therefore best-practice dashboards & metrics are such and such …)
  • create insightful “exploration paths” via related sidebar

All of these are in various stages of being worked on, but it’s still early days. There will be at least one major reworking of how x-rays are defined as the current format is too limiting and too repetitive.