Oracle SQL Developer突变表和触发器

Question

I am making a small project for my university subject and I have a problem with designing a trigger in Oracle SQL Developer for my small database.

I have the following tables:

create table ASSIGNED_TO(
  ASSIGNMENT_ID int not null primary key,
  ROOM_ID int not null,
  foreign key (ROOM_ID) references ROOM(ROOM_ID),
  NURSE_ID int not null,
  foreign key (NURSE_ID) references NURSE(PERSON_ID)
);

create table MESSAGE_LOG(
  ID int not null,
  MESSAGE varchar2(100) null
);

Each time a row from table ASSIGNED_TO is changed by INSERT or UPDATE operation and there are 2 rows with the same ROOM_ID I want the trigger to write a simple message into the MESSAGE_LOG table.

What causes me problems is the 'mutating' of table ASSIGNED_TO . My current solution is:

create or replace TRIGGER TRIG3
  AFTER INSERT OR UPDATE
  ON ASSIGNED_TO
  FOR EACH ROW
DECLARE
  nr_nurses INTEGER;  
BEGIN
      SELECT COUNT(*) INTO nr_nurses
        FROM ASSIGNED_TO
      WHERE ROOM_ID = :NEW.ROOM_ID;

      IF nr_nurses >= 2 THEN
        INSERT INTO MESSAGE_LOG values (3, '2 or more nurses per room detected');
        DBMS_OUTPUT.PUT_LINE('3| 2 or more nurses per room detected');
      END IF;
END;

This trigger gives me 'Mutating Table' error and so far I have been unable to fix it. Using :OLD.ROOM_ID instead of ROOM_ID fails as :OLD.ROOM_ID does not exist in case of UPDATE operation. Any advice on making this work is appreciated.

Answer 1

As @MichaelBroughton has pointed out, you can't use a row trigger here as you'll get the dreaded "mutating table" exception. There are various ways to work around this issue; perhaps the simplest is to use a statement trigger to accomplish the same goal. Rework your trigger as:

CREATE OR REPLACE TRIGGER TRIG3
  AFTER INSERT OR UPDATE
  ON ASSIGNED_TO
BEGIN
  -- The following cursor will only find rows for
  -- which multiple nurses have been assigned.

  FOR aRow IN (SELECT ROOM_ID, COUNT(*) AS NURSE_COUNT
                 FROM ASSIGNED_TO
                 GROUP BY ROOM_ID
                 HAVING COUNT(*) >= 2)
  LOOP
    INSERT INTO MESSAGE_LOG
      (ID, MESSAGE)
    VALUES
      (3, aRow.NURSE_COUNT || 'ASSIGNED TO ROOM ' || aRow.ROOM_ID);

    DBMS_OUTPUT.PUT_LINE('3| ' || aRow.NURSE_COUNT || 'ASSIGNED TO ROOM ' ||
                         aRow.ROOM_ID);
  END LOOP;
END TRIG3;

Note: this is a statement trigger because it doesn't include the line FOR EACH ROW .

This trigger will be invoked once for each INSERT or UPDATE statement which affects the ASSIGNED_TO table instead of once for each row altered, but done this way the trigger will provide the same functionality as the original row trigger and will not suffer from the "mutating table" problem.

Share and enjoy.

Answer 2

Referencing the table that you are inserting into causes this issue because you are asking about the contents of the table right as it is changing, but before that change is committed.

The answer is to not do this check through a trigger. I prefer pl/sql business-level interfaces to be accessed rather than pure table interaction from the UI. For example, code an insert_room_assigment(nurse_id, room_id) procedure that does the insert, the check, and the logs the problem if neccessary, or could check first and not allow the insert in the first place.

But however you solve it - the trigger is NOT going to work.