Bernhard Scholz

Sorry for the late reply. This is so that facts are not parallelised.

I am happy to elaborate more. Send me an email ([email protected]) /let's zoom - also, I would like to talk about something else.

Note that the choice-domain is not deterministic. That can be an artefact of choice (i.e., the first tuple that is functionally independent stays in the relation and subsequent ones are...

Does this work if you run it without the `brie` qualifier? Most likely, it is an unfixed brie problem as reported in previous issues.

By removing the brie qualifiers, the first program produces an extra tuple `(-1, -1)`. This seems to be a previously reported brie issue that hasn't been fixed yet. I would...

That is an interesting question. Yes, the positive description (what it can be) can be expressed in Souffle. However, this comes at the cost of enumerating all permitted branches. We...

Note this might not be the most efficient way, and extending the Souffle language to deal with the use case you describe might be better (i.e. let the Souffle language...

This is an excellent way of doing it!

Yes - indeed. The interpreter is the main stumble block. The idea here is to de-specialize data structures as much as possible. For tuples with arbitrary element lengths, we would...

The problem is that the interpreter uses a bitwise comparison, and the compiler uses a proper floating-point comparison. A fix is possible but requires dealing with corner cases, i.e. weakening/strengthening...