This might seem a little circular in nature, but ..
We are doing some architectural prototyping with Databricks, and we have been able to pull DB-hosted content into Stardog as virtual graphs.
Then, we tried to set up a Foreign Catalog connection from Databricks to Stardog BI. The idea behind this actually seems pretty elegant:
- DB basically federates content into a nice, cataloged pile of resources
- Stardog is able to
- ingest selected bits of that content as raw material for VKGs
- And is able to host various models that act as lenses on the ingested content
- And then is able to expose a relational model on it all, using Stardog BI
- And we then add this Stardog BI content as a Foreign Catalog
This seems like it fits well with a Data Product philosophy. If Databricks is essentially "the mesh", then Stardog is basically a transformation engine that:
- takes "raw material" in as VKGs
- allows synthesis of various models using OWL, saved queries etc.,
- and then exposes those "semantified results" in relational format using Stardog BI
So, if we could include the Stardog BI stuff in the catalog, it would basically be an "indexed, refined data product", nestled in there with the raw material that it served to refine.
Unfortunately, while we can include Stardog BI as a source, it doesn't seem to completely work ..
For example, in the attached DBeaver screen-shot, I can't see the tables under the "stardog_bi_foreign_catalog" virtual database.
