confused by the break command of sqlplus - oracle

Question

From documentation of sqlplus for break
When specify multiple on clause, sqlplus would search from the first on clause to the last on clause for a break. When it finds a break it will perform the actions in reverse order, from the last to first.
I connect to the scott schema and have some test:
SQL> break on deptno skip 1 on job skip 2
SQL> break
break on deptno skip 1 nodup
on job skip 2 nodup
SQL> select deptno, job from emp order by deptno, job;
DEPTNO JOB
---------- ---------
10 CLERK
MANAGER
PRESIDENT
20 ANALYST
DEPTNO JOB
---------- ---------
20 ANALYST
CLERK
MANAGER
DEPTNO JOB
---------- ---------
30 CLERK
MANAGER
SALESMAN
14 rows selected.
SQL>
After print the first row(10, CLERK), there is a change in job, it just skips 2 lines, why not 3 lines ? In the second page, there are tow places sqlplus skip 3 lines. At the end of the report is skips 5 lines. Just don't know how it decides how many lines to skip when specify multiple on clause.

Answer 1

After print the first row(10, CLERK), there is a change in job, it just skips 2 lines, why not 3 lines ?
Why would it skip 3 lines when you have mentioned on job skip 2. SQL*Plus did exactly what you asked it to do, to skip 2 lines.
In the second page, there are tow places sqlplus skip 3 lines.
That is because you also mentioned break on deptno skip 1 So, for a new department, there is one more line skipped. So, 2 lines skipped for job and 1 line skipped for department, thus a total of 3 lines skipped.
Also, for department 20, there are 2 ANALYST and 2 CLERK, so it skips 2 lines for each job.
SQL> break on job skip 2
SQL> select deptno, job from emp order by deptno, job;
DEPTNO JOB
---------- ---------
10 CLERK
10 MANAGER
10 PRESIDENT
20 ANALYST
20
20 CLERK
20
20 MANAGER
30 CLERK
30 MANAGER
30 SALESMAN
30
30
30
14 rows selected.
SQL>
At the end of the report is skips 5 lines.
Because, in the end for department 30, there are 4 SALESMAN, and there will always be an extra line before the total number of rows message is displayed.
SQL> set pagesize 50
SQL> select deptno, job from emp order by deptno, job;
DEPTNO JOB
---------- ---------
10 CLERK
10 MANAGER
10 PRESIDENT
20 ANALYST
20 ANALYST
20 CLERK
20 CLERK
20 MANAGER
30 CLERK
30 MANAGER
30 SALESMAN
30 SALESMAN
30 SALESMAN
30 SALESMAN
14 rows selected.
SQL>
The best way to understand this is, add another column without any skip for example add empno and see the lines which are actually skipped.
SQL> break on deptno skip 1 on job skip 2
SQL> select empno, deptno, job from emp order by deptno, job;
EMPNO DEPTNO JOB
---------- ---------- ---------
7934 10 CLERK
7782 MANAGER
7839 PRESIDENT
7788 20 ANALYST
7902
7876 CLERK
7369
7566 MANAGER
7900 30 CLERK
7698 MANAGER
7654 SALESMAN
7521
7499
7844
14 rows selected.
SQL>

Answer 2

You aren't seeing quite what you think. You're only querying the department and job columns, which can have duplicates - as more than one employee can be in each job. The nodup part of the break setting means duplicates of those two columns will not be shown; which means you are seeing a mix of rows with have both columns blanked as duplicates, and rows that are skipped.
I have a different table but for the same effect:
break on department_id skip 1 on job_id skip 2
select department_id, job_id
from employees
where rownum < 10;
DEPARTMENT_ID JOB_ID
------------- ----------
90 AD_PRES
AD_VP
60 IT_PROG
100 FI_MGR
9 rows selected.
If I add another non-null column, and/or the rownum pseudocolumn, you can see what is really happening:
ROWNUM DEPARTMENT_ID JOB_ID EMPLOYEE_ID
---------- ------------- ---------- -----------
1 90 AD_PRES 100
2 AD_VP 101
3 102
4 60 IT_PROG 103
5 104
6 105
7 106
8 107
9 100 FI_MGR 108
9 rows selected.
Now you can see it skips two lines if the job changes within the department, and three when the department changes. Which is what you asked it to do.
You probably don't want to, but you could also just get the distinct values:
select distinct department_id, job_id ...
DEPARTMENT_ID JOB_ID
90 AD_VP
AD_PRES
60 IT_PROG
100 FI_MGR
4 rows selected.
... and again you'll see two skipped lines between jobs within a department, and three between departments.
You might also be confusing the description of the behaviour in the documentation; you said "When it finds a break" but it can find multiple breaks for one row in the results; the docs you linked to actually say:
SQL*Plus searches each row of output for the specified breaks, starting with the outermost break and proceeding—in the order you enter the clauses—to the innermost.
With your query it will either match department, job, or both.
Next, SQL*Plus executes actions beginning with the action specified for the innermost break and proceeding in reverse order toward the outermost break ... SQL*Plus executes each action up to and including the action specified for the first break encountered in the initial search.
The innermost break in your case is on job, the outermost on department. If the department didn't change but the job did then it will only find a matching break on job, and only the innermost skip (of two rows) will be applied. If the department changed - whether or not the job changed too - it will find a match on that, but will still apply all of the on clause from department onwards, in reverse order; so it will apply the skip 1 from job and the skip 2 from department; hence skipping three lines in total.
Yes I am using rownum without an order by, it happens to work better with this query and my data; don't try this at home