Maximum Constraint on One to Many Relationship - Oracle SQL

This cannot be done with a check constraint. It should be possible to create a materialized view that count the number of occurences of each manager, with a check constraint on the count, and that refreshes on commit on the original table. The same can be implemented with a compound trigger, as demonstrated by Littlefoot. But this is not very scalable, since the whole table needs be scanned to refresh the materialized view after each commit.

One alternative solution would be to:

• create a new table that keeps track of the number of occurences of each manager, say employee_manager_cnt

• set up a trigger on the employee table to keep table employee_manager_cnt up to date (no need to scan the whole table, just reflect the changes based on the old and new value of manager_id)

• add a check constraint to the employee_manager_cnt that forbids values above the target count

Here is a step by step demo, which is inspired by the answer by nop77svk on this SO question

Original table:

create table employees (
    employee_id number primary key,
    manager_id number
);

Insert a few records:

begin
    insert into employees values(1, null);
    insert into employees values(2, 1);
    insert into employees values(3, 1);
    insert into employees values(4, 1);    -- manager 1 has 3 employees
    insert into employees values(5, null);
    insert into employees values(6, 5);    -- manager 5 has just 1 employee
end;
/

Create the new table:

create table employee_manager_cnt (
    manager_id          number not null primary key,
    cnt                 number(1, 0) not null check (cnt <= 3)
);

Populate it:

insert into employee_manager_cnt(manager_id, cnt)
select manager_id, count(*) 
from employees 
where manager_id is not null
group by manager_id

Check the results:

MANAGER_ID  CNT
1           3
5           1

Now, create the trigger:

create or replace trigger trg_employee_manager_cnt
    after insert or delete or update of manager_id
    on employees
    for each row
begin

    -- decrease the counter when an employee changes manager or is removed
    if updating or deleting then
        merge into employee_manager_cnt t
        using dual
        on ( t.manager_id = :old.manager_id )
        when matched then
            update set t.cnt = t.cnt - 1
            delete where t.cnt = 0
        ;
    end if;

    -- increase the counter when a employee changes manager or is added
    if inserting or updating then
        merge into employee_manager_cnt T
        using dual
        on ( t.manager_id = :new.manager_id )
        when matched then
            update set t.cnt = t.cnt + 1
        when not matched then
            insert (manager_id, cnt) values (:new.manager_id, 1)
        ;
    end if;
end;
/

Now try to add a new record that references manager 1 (who already has 3 employees)

insert into employees values(4, 1);
-- error: ORA-00001: unique constraint (FIDDLE_QOWWVSAIOXRDGYREFVKM.SYS_C00276396) violated

Meanwhile it is still possible to affect a new employee to manager 5 (he just has one employee):

insert into employees values(10, 5);
-- 1 rows affected

In order to find number of employees per manager in the table, you have to count them, right? But, if you do that, you'll hit the mutating table error as you can't select from a table which is currently being changed.

Nowadays, we fix that using the compound trigger. Here's an example:

Sample table:

SQL> create table temp
  2    (empid number primary key,
  3     name  varchar2(20),
  4     mgrid number references temp (empid)
  5    );

Table created.

Compound trigger:

SQL> create or replace trigger trg_3emp
  2    for update or insert on temp
  3    compound trigger
  4
  5    type emprec is record (mgrid temp.mgrid%type);
  6    type row_t  is table of emprec index by pls_integer;
  7    g_row_t     row_t;
  8
  9  after each row is
 10  begin
 11    g_row_t (g_row_t.count + 1).mgrid := :new.mgrid;
 12  end after each row;
 13
 14  after statement is
 15    l_cnt number;
 16  begin
 17    for i in 1 .. g_row_t.count loop
 18      select count(*)
 19      into l_cnt
 20      from temp
 21      where mgrid = g_row_t(i).mgrid;
 22
 23      if l_cnt = 4 then
 24         raise_application_error(-20000, 'No more than 3 employees per manager');
 25      end if;
 26    end loop;
 27  end after statement;
 28  end;
 29  /

Trigger created.

SQL>

Testing:

SQL> -- This will be the manager
SQL> insert into temp values (1, 'Little', null);

1 row created.

SQL> -- Next 3 rows will be OK
SQL> insert into temp values (2, 'Foot'  , 1);

1 row created.

SQL> insert into temp values (3, 'Scott' , 1);

1 row created.

SQL> insert into temp values (4, 'Tiger' , 1);

1 row created.

SQL> -- The 4th employee for the same manager should fail
SQL> insert into temp values (5, 'Mike'  , 1);
insert into temp values (5, 'Mike'  , 1)
            *
ERROR at line 1:
ORA-20000: No more than 3 employees per manager
ORA-06512: at "SCOTT.TRG_3EMP", line 22
ORA-04088: error during execution of trigger 'SCOTT.TRG_3EMP'


SQL> -- Someone else can be Mike's manager
SQL> insert into temp values (5, 'Mike', 2);

1 row created.

SQL>