Yeah it gave me this hunch too.

I could imagine a language /framework where we declare the schema of data, indexes on this schema, and either let a query planner optimise data access or explicitly specify which index to use