I made a TypeScript library for querying in-memory JS data. Originally, the goal was to have a more modular and lightweight alternative to AlaSQL, but then I was asked to include support for SQL/XML, which I personally find way too verbose. So I got to thinking, remembered Virtuoso's combined SQL and SPARQL queries and extrapolated from there.
The resulting engine consists of a shared core that provides the optimizer, query infrastructure etc., and separate languages that can be added as plugins. The way this is implemented is that the languages are translated into a common algebra. Thanks to that, you can mix multiple languages in a single query - imagine you need to process some tabular data, but for each row you want to extract information from a deep tree-like structure. Well, you can write an outer SQL query and put in an XQuery subquery and it just works. The algebra itself is designed to be extensible - I don't presume to have covered everything any query language might need (but there is a language implemented for every major data model - SQL, XQuery and a Cypher-like language (licensing)). For example, XQuery includes two special algebra operators of its own, without needing to modify the core.
There is a link to a demo on the GitHub README, where you can play with query plans, or execute queries on some sample data. There is also a link to my diploma thesis about this project, if you're interested in that. The whole thing was presented as a demo at VLDB last year.
This is a result of more than 2 years of work, and I'm at the point where I think other people might find it useful too - mainly because I was able to make docs I am happy with. The docs were created with the help of a LLM, but I spent weeks on that, so hopefully they read good. The code itself was written by hand, with the exception of the initial batch of regression tests.