About PG-Schema

[roccho-dev]

kuzudb's schema constraints are already very good.

Meanwhile, the current transition has seen the GQL standard established
PG-Schema Will this also become standard?

I'd love to hear what you all think and discuss how kuzudb's constraint definitions can be followed.

[semihsalihoglu-uw]

Hi @PorcoRosso85, I don't know a lot about the standardization efforts going on in the graph community. I met a few people at Connected Data London who are actively working on it. The schema that we want to support for directed graphs is simple: a set of node tables with predefined and typed attributes, and a set of relationship tables that can be from a set of node tables to another set of node tables (previously, our schema was stricter and you could only have a 1 from node table and 1 to node table).

One feature I would be interested to learn more about is whether the people working schema standards are supporting undirected graphs and if so, how? We have discussed from time to time about natively supporting undirected graphs, and before we do that, I would want to make sure we study what the standards are doing. They spend more deeply about these topics than we do.

Other than that, a simple relational-style node/rel table definitions is what we have implemented and are interested in supporting.

[roccho-dev]

@semihsalihoglu-uw
Very nice to hear the impression you are getting from your environment.

Perhaps kuzudb will adopt GQL too in the future?
Do you plan to proceed with your own format for the schema?

[prrao87]

Hi @PorcoRosso85, just wanted to chime in: Kuzu aims to follow the developments in the ecosystem and how other graph databases are doing it, and of course track the larger adoption of GQL. Eventually, Cypher can be "absorbed" into GQL, but that process can take time and it definitely won't happen overnight. Based on internal discussions we've had in the team, the structured schema of Kuzu aligns well with GQL's requirements and we foresee no issues in gradually adding support for it over the coming years. But again, this is a slow process and we would only do it based on how the larger ecosystem evolves.

[roccho-dev]

@prrao87
Thanks.

Based on internal discussions we've had in the team, the structured schema of Kuzu aligns well with GQL's requirements

It's the answer for me.
I want to depend on kuzudb deeply, so I have to consider how hard to follow the standard.