I actually wrapped partition creation in a PL/PgSQL function and since I no longer have CREATE TABLE in the schema, sqlc doesn't create the useless structs. But I guess if sqlc did database introspection instead of parsing the schema file, the problem might return.