Why the non-standard syntax?

[gisborne]

There must be good reasons, but why doesn't Cozo use the standard syntax of Datalog — a sub-language of Prolog. As I understand it, Datalog programs ought to even run on a Prolog engine.

This also includes me wondering why we need to differentiate := from ~>

[xpe]

I also have similar questions. I've used various logic programming languages (Prolog, Datomic, and others). I am surprised at how different Cozo's datalog syntax is compared to others.

[xpe]

According to https://docs.cozodb.org/en/latest/queries.html :

The constant rules which use <- to connect the head and the body are syntax sugar. For example:

const_rule[a, b, c] <- [[1, 2, 3], [4, 5, 6]]

is identical to:

const_rule[a, b, c] <~ Constant(data: [[1, 2, 3], [4, 5, 6]])

[aramallo]

I also would like a standard syntax.

As I understand it, Datalog programs ought to even run on a Prolog engine.

@gisborne That is not correct :-) Datalog programs do not run on a Prolog engine. For starters Prolog operates one tuple at a time, whereas Datalog operates with sets. Standard Datalog guarantees termination as it is designed mainly for query solving, whereas Prolog can't as it is a general purpose language. Datalog uses query optimizations techniques, in Cozo's case the evaluation is bottom-up and thus requires the Magic Sets transformation, something Prolog doesn't offer, etc.

[gisborne]

On Sep 3, 2024, at 11:25, Alejandro M. Ramallo ***@***.***> wrote: I also would like a standard syntax. As I understand it, Datalog programs ought to even run on a Prolog engine. @gisborne <https://github.com/gisborne> That is not correct :-) Datalog programs do not run on a Prolog engine. For starters Prolog operates one tuple at a time, whereas Datalog operates with sets. Standard Datalog guarantees termination as it is designed mainly for query solving, whereas Prolog can't as it is a general purpose language. Datalog uses query optimizations techniques, in Cozo's case the evaluation is bottom-up and thus requires the Magic Sets transformation, something Prolog doesn't offer, etc.

Yeah, but for Turing Completeness in Prolog, you need cut and/or functions, don’t you? And a Datalog query won’t have either. I agree the Datalog program would be more efficient because of query optimisation.

[boticello]

My memory is that the syntax was designed to be compatible with the Python syntax highlighter in Jupyter, but I can't find a reference to it.