Comment by ulrikrasmussen

Comment by ulrikrasmussen 3 days ago

Thank you! I have been searching for something like this but for some reason couldn't find your work.

I am currently implementing a Datalog to PostgreSQL query engine at work as we want to experiment with modeling authorization rules in Datalog and then run authorization queries directly in the database. As I want to minimize the round trips to the database I use a different approach than yours and translate Datalog programs to recursive CTEs. These are a bit limited, so we have to restrict ourselves to linearly recursive Datalog programs, but for the purpose of modeling authorization rules that seems to be enough (e.g. you can still model things such as "permissions propagate from groups to group members").

philzook 3 days ago

My suspicion is that if you can get away with it that recursive CTEs would be more performant than doing the datalog iteration query by query. AFAIK for general rules the latter is the only option though. I had a scam in that post to do seminaive using timestamps but it was ugly. I recently came across this new feature in duckdb and was wondering if there is some way to use it to make a nice datalog https://duckdb.org/2025/05/23/using-key.html . Anything that adds expressiveness to recursive CTEs is a possibility in that direction

Reply View 1 reply

ulrikrasmussen a day ago

Yes, I think so too, if not just for the reduced number of round-trips to the database. The evaluation strategy that PostgreSQL employs for CTEs is essentially a degenerate form of seminaive evaluation where it only has to consider the delta for one subgoal in each rule due to being restricted to linearly recursive programs with no mutual recursion. The restriction to not have mutual recursion means that PostgreSQL can just stratify the CTE and compute a fixed point using the degenerate seminaive algorithm for each strata. Once a stratum is done, it can consider its idb as ground facts in the next, and so on.
I am really unhappy about the design of CTEs, and frankly I think it is a clunky hack designed by someone who was not aware of all the research on Datalog evaluation, which Ullman and others had already written excellent textbooks about at the time. CTEs are simultaneously too powerful - Turing-completeness in a declarative query language is a sure way to tell that the designers screwed up - and too restricted because being stuck in the Turing tarpit rules out so many techniques for optimization.
I didn't know about DuckDB's approach, but it seems to be a way to mark your relations as moded and exploit that during the evaluation. It appears to improve performance, but unfortunately the design it builds on is inherently flawed. I wish SQL would get a proper pure Datalog fragment in the future.

Reply View | 0 replies

whitten 3 days ago

What does CTE stand for, and how do I research it ?

Reply View 1 reply

burakemir 3 days ago

Common Table Expression, a SQL concept that enables more expressive programming with SQL queries. They are introduced using WITH ...

Reply View | 0 replies