How to handle the trigger sequence

Question

I really do not have any idea for this question.
the trigger need to implement:      
An employee can not work on more than 2 projects supervised by any given department.
DEPARTMENT (dname, dnumber, mgrssn, mgrstartdate) KEY: dnumber.
PROJECT (pname, pnumber, plocation, dnum)      KEY: pnumber.
CREATE TABLE WORKS_ON
( ESSN VARCHAR2(9) NOT NULL,
  PNO NUMBER NOT NULL,
  HOURS NUMBER(3,1) NOT NULL,
PRIMARY KEY (ESSN, PNO));
DESC WORKS_ON;
INSERT INTO WORKS_ON VALUES
('123456789',1,32.5);
INSERT INTO WORKS_ON VALUES
('123456789',2 ,7.5);
INSERT INTO WORKS_ON VALUES
('666884444',3 ,40.0);
INSERT INTO WORKS_ON VALUES
('453453453',1 ,20.0);
INSERT INTO WORKS_ON VALUES
('453453453',2,20.0);
CREATE TABLE PROJECT
(PNAME VARCHAR2(15) NOT NULL,
PNUMBER NUMBER NOT NULL,
PLOCATION VARCHAR2(15),
DNUM NUMBER NOT NULL,
PRIMARY KEY (PNUMBER), UNIQUE(PNAME));
DESC PROJECT;
INSERT INTO PROJECT VALUES(
'ProjectX',1, 'Bellaire',5);
INSERT INTO PROJECT VALUES(
'ProjectY',2, 'Sugarland',5);
INSERT INTO PROJECT VALUES(
'ProjectZ',3,'Houston',5);
INSERT INTO PROJECT VALUES(
'Computerization',10,'Stafford',4);i create package and two trigger like following in order to avoid mutating table.
PACKAGE                     "HW4AIDB" AS
TYPE EmploadY IS RECORD(
     Empessn varchar2(9),
     Empprojectcount number,
     Empdepartcount number
TYPE EmploadarrayY is table of EmploadY
                      index by BINARY_INTEGER;
Empload EmploadarrayY;
Emploadsize number;
END;
create or replace
TRIGGER TPROJECTA AFTER INSERT OR UPDATE OR DELETE ON WORKS_ON
DECLARE
i number:=0;
BEGIN
  HW4AIDB.Emploadsize :=0;
  for m in (select w.essn empessn,
  count(w.pno) empprojectcount, count(distinct p.dnum) empdepartcount
  from works_on w, project p
  where w.pno=p.pnumber
  group by w.essn)loop
  i :=i+1;
  HW4AIDB.Empload(i).Empessn := m.empessn;
  HW4AIDB.Empload(i).Empprojectcount := m.empprojectcount;
  HW4AIDB.Empload(i).Empdepartcount := m.empdepartcount;
  dbms_output.PUT_LINE('TPROJECTA '||' '||m.empessn||' '||m.empprojectcount||
                       ' '||m.empdepartcount);
  end loop;
  HW4AIDB.Emploadsize :=i;
  dbms_output.PUT_LINE('==================================');
END;TRIGGER TSALARYB BEFORE INSERT OR UPDATE OR DELETE ON EMPLOYEE
FOR EACH ROW
DECLARE
w HW4AID.EmploadX;
i number;
mypos number;
BEGIN
mypos :=-1;
for i in 1..HW4AID.Emploadsize loop
if(:new.ssn = HW4AID.Empload(i).Empssn) then
w.Empmgrsalary := HW4AID.Empload(i).Empmgrsalary;
w.Empsalary := HW4AID.Empload(i).Empsalary;
mypos :=i;
exit;
end if;
end loop;
i :=mypos;
dbms_output.put_line('NEW '||:new.ssn||'-'||:new.salary);
if updating then
dbms_output.put_line('updating......');
if(:new.salary > w.Empmgrsalary) then
raise_application_error(-20005,'the salary is higher than that of department manager');
end if;
dbms_output.PUT_LINE('update done');
end if;
END;
i defined two trigger for AFTER, trigger run orders are acorrding to. before statement level, before row-level, after row level and after statement level.
so my trigger actually can not implement requirement. how to handle it. now i try to use this rule: select w.essn, w.pno,p.pnumber, p.dnum from works_on w, project p where w.pno=p.pnumber; in statement level trigger. howerver in row-level trigger i can not get value i need(p.dnum i can not get its count). if you guy have any other ideas, pls give me some advise. my brain is really run out of.
Thanks for your help!!!

Answer 1

I would recommend creating a package/procedure that is the only way to insert a user into the works_on table. Then inside this procedure you need to serialize access to the works_on table by using a select for update. Then you can count the number of projects and ensure that the new works on is valid.
If you try to use triggers then due to Oracle's mutli-versioning, if two users insert the same data at the same time, the data integrity will be violated.
You will obviously have an employee table. I have created one here as follows:
create table employee (essn varchar2(9) primary key);
insert into employee values ('999999999');
Then you can use this table to serialize access to the works_on table by issuing a select for update against it. If you don't do this then two users executing the procedure at the same time will violate the constraint. It is important that you understand this and try it out with multiple sessions so that you understand it.
create or replace procedure insert_works_on
(p_essn varchar2, p_pno number, p_hours number)
as
l_essn varchar2(9);
l_count number;
begin
select essn
into l_essn
from employee
where essn = p_essn
for update;
select count(*)
into l_count
from works_on w
join project p on p.pnumber = w.pno
where w.essn = p_essn
and p.dnum = (select dnum
from project
where pnumber = p_pno);
if l_count < 2
then
insert into works_on values (p_essn, p_pno, p_hours);
else
dbms_output.put_line('Employee works on more than 1 project for dept');
end if;
end;
Then you can try it out:
begin
insert_works_on('999999999',1, 10);
end;
commit;
begin
insert_works_on('999999999',2, 20);
end;
commit;
begin
insert_works_on('999999999',3, 30);
end;
commit;
The first two will succeed and the last will fail. Try it out in multiple session as well to ensure it works for a multi-user scenario.