AFICT, nothing in the ecosystem today (postgrex nor extensions/peers) supports postgres xml types, functions, or queries.
For types, you can use custom Ecto.Types to handle serialization between PG and Elixir. For DDLs, I’d just write my migrations in straight SQL strings, executed by Repo.query! rather than a DSL.
For functions and queries, you’re going to want to use fragments. Essentially, escaping the pure data-modelling of Ecto to insert raw SQL into an otherwise safely Ecto-built query.
This is similar to what I do in my python projects that use XML columns, and it works fairly well. Our use-case is write-once query-occasionally so it’s fairly trivial: we rarely need to read the full document into memory once inserted, and never need to update it.
More integrated functionality would require more work. Perhaps a library calling out to be made?
