For a well-constrained Postgres schema, a deterministic SQL compiler can be built (with plenty of logic programming/constraint-solving/hypergraph-analysis) that can integrate arbitrary external systems data.
While this is NP-hard, with some clever use of embeddings and ontologies, and use of every single constraint AST within the DB to reduce the search space, this becomes a feasible problem.
At the moment I'm saving up for the next development cycle, so not only PG->PG schema mappings can be solved for (JSON schema is next!). Hope this sounds interesting :)
NixOS/GNU Guix is uniquely positioned in this area, as it tracks all dependencies explicitly, with exact versions. If there is a paradigm where this can be achieved on the OS level, they are the closest to it today.
That's what I've been trying to do with: https://github.com/schemamap/schemamap
For a well-constrained Postgres schema, a deterministic SQL compiler can be built (with plenty of logic programming/constraint-solving/hypergraph-analysis) that can integrate arbitrary external systems data.
While this is NP-hard, with some clever use of embeddings and ontologies, and use of every single constraint AST within the DB to reduce the search space, this becomes a feasible problem.
For any Clojurists interested, I've packaged `pg_query`, so you can use it in your applications: https://github.com/schemamap/pg-query-clj
At the moment I'm saving up for the next development cycle, so not only PG->PG schema mappings can be solved for (JSON schema is next!). Hope this sounds interesting :)