A curious "anomoly" with the NATURAL JOIN compared to the JOIN USING

Question

The following query returns what I expect from the HR schema [106 rows]:
select last_name, department_name from employees JOIN departments USING (department_id);
However, when I do the full natural join with this, I only get 32 rows.
select last_name, department_name from employees  NATURAL JOIN departments;
Is the full NATURAL JOIN not using DEPARTMENT_ID to join on? But if not what else would it use? I just ran into this and am mostly curious as I do not use the natural join syntax in production but instead prefer the JOIN ON.
Thanks.

Answer 1

Just expanding on user2361373's point a bit, the EMPLOYEES and DEPARTMENTS table both have a DEPARTMENT_ID column and both have a MANAGER_ID column. So the natural join will join on both columns.
This is why the natural join is evil and ought never, ever, ever, ever, ever to be used in real production code. It violates fundamental tenants of building maintainable applications because it creates the potential that adding columns to an existing table will cause existing code to return incorrect results. If you decide that you need to add a CREATE_DATE column to a few tables, for example, to track when a row was inserted, suddenly, all your natural joins may decide that they need to add the CREATE_DATE to the list of columns to join on and your app starts behaving incorrectly. Or your initial DEPARTMENTS table doesn't have a MANAGER_ID column and you decide to add that later on-- all the existing natural joins between EMPLOYEE and DEPARTMENT start returning incorrect results.
Justin