I just read Better Together: Unifying Datalog and Equality Saturation.
A lot of good ideas in this paper. I'm interested in trying to use it for some of our analyses, or to implement a simple type system. However, like most logic languages I wonder how hard it is to for example read and parse a file, or if you need to parse a program in a more traditional language and then generate the egglog code for analyzing the program after the fact.
Anyways, the full version of the paper is on arxiv and includes an appendix with practical examples. The repo says that the python binding has a bit more documentation, but the link is bad. The rust crate has some docs but mostly for the api bindings and not the language itself. So I guess I'll be looking into the appendix of their extended paper if I want to learn more.
From the same conference (PLDI) I watched the presentation for this paper, as well as another talk from that section about Indexed Streams which was also interesting, but not really related.