First of all, a commit message does not need to contain SQL examples of what it does. We should provide human-readable explanations and that's it.
Next, about changes to src/test/regress/sql/copy2.sql. I find the sql you used to test really unintuitive. How about CREATE TABLE ... PARTITION OF syntax? It is also one command instead of two (create + alter). It is also hard to say what partition structure is, because column names on different partition levels are the same, just order is switched. Let's change it to something more intuitive too?
-- Best regards, Kirill Reshke
Maybe we can tab-complete here if prefix matches pg_% ? Does that makes good use case?
