Seems harmless but a little unusual. Do you have a project that would use this?

Thank you for your contribution. I'm also curious (if you can share) what you're using feldera for?

Hello! I'm using DBSP as an incremental computation backend for a Datalog language I've built to experiment with running CRDTs (convergent replicated data types) on.

Hello! I'm using DBSP as an incremental computation backend for a Datalog language I've built to experiment with running CRDTs (convergent replicated data types) on.

That's awesome. I assume you know about Differential Datalog?

Yes, I've encountered it while researching approaches in the beginning. If I'm not mistaken, it has a compilation step (using rustc), right?

I ended up writing my own interpreter-based approach because:

1. I wanted to "own" my tech stack, at least down to the computation backend. Admittedly, DBSP's internals are still mostly a black box to me.
2. For the environments I'm interested in, I don't want to bundle a full compiler.
3. I'm also interested in comparing different incremental computation approaches, that is, differential dataflow and DBSP (do you know about other similarly capable systems?). Although I'm not quite there yet, I want to implement a differential dataflow backend, compare runtimes and understand differences (or limitations) in their expressive power. Until today, I haven't found a definitive analysis on what exactly sets the two apart in terms of their expressive power, underlying time assumptions, and performance.
4. Also I want to understand the gap between incremental and non-incremental approaches for CRDT use cases (frequent and small updates to a view defining the state of the CRDT). For that I plan to wire up a DuckDB backend to my Datolog language to also have a representative of a non-incremental query engine.

Yes, I've encountered it while researching approaches in the beginning. If I'm not mistaken, it has a compilation step (using rustc), right?

Yes.

...

3. I'm also interested in comparing different incremental computation approaches, that is, differential dataflow and DBSP (do you know about other similarly capable systems?).

There are, at least, Python and Java implementations of the DBSP approach. I am not sure whether they are production systems or just demos or toys.

It sounds like you've got some fun goals. Good luck!

Clearly, there's also Materialize.com, which is on top of differential dataflow, but you probably know that.