Xrhrft

I have two INT integer types in my table that set as UNIQUE KEY to avoid duplication. I wanted to rollback any duplicate occur in my stored procedure using EXIT HANDLER.

CREATE TABLE curriculum

(

  id INT PRIMARY KEY AUTO_INCREMENT,

  gradelevel_id INT,

  schoolyear_id INT,

  CONSTRAINT uc_curriculum UNIQUE (gradelevel_id, schoolyear_id)

)



DELIMITER //

CREATE PROCEDURE insertCurriculum(

   IN pIN_gradelevelId INT, 

   IN pIN_schoolyearId INT

)

BEGIN



  DECLARE EXIT HANDLER FOR SQLEXCEPTION

  BEGIN

     ROLLBACK;

        SELECT "Duplicate keys found";

     RESIGNAL;

  END;



  INSERT INTO curriculum (gradelevel_id, schoolyear_id) VALUES (pIN_gradelevelId, pIN_schoolyearId);



END //

DELIMITER ;

I inserted manually in my curriculum table for the test.

INSERT INTO curriculum (gradelevel_id, schoolyear_id) VALUES (1, 1);



SELECT * FROM curriculum



`id  |  gradelevel_id | schoolyear_id

 1           1              1`

FIRST TRY:

Calling my first stored procedure that has duplicate it automatically rollback.

call insertCurriculum(1, 1);

call insertCurriculum(2, 2);



SELECT * FROM curriculum



`id  |  gradelevel_id | schoolyear_id

 1           1              1`

SECOND TRY:

Calling my first stored procedure that has no duplicate it allows the first stored procedure to be executed and insert into my table. Even the second stored procedure has duplicate. I wanted to rollback this. How can I achieve this? Any tips will greatly appreciated!

call insertCurriculum(2, 2);

call insertCurriculum(1, 1);



SELECT * FROM curriculum



`id  |  gradelevel_id | schoolyear_id

 1           1              1

 2           2              2`

Your question is very unclear, but maybe I can still answer it.

Why is it unclear?

What do you mean with duplicates exactly? In your examples there are none. A duplicate relates to rows. You would have a duplicate in your table, if you tried to insert (1, 1) two times.

If you meant with duplicates, that it shouldn't be allowed to have the same value in both columns, then you will have to catch that with a trigger. Create a before insert trigger and have the code something along the line if (new.gradelevel_id = new.schoolyear_id) then signal...; end if;

Anyway...(some more explaining)

In your procedure there won't be any rollback, cause there's nothing to roll back. A rollback statement only makes sense, when your transaction consists of multiple statements. To have multiple statements in a transaction you either have the auto_commit variable set to 0 or you explicitly start a transaction with START TRANSACTION; When you didn't do either, a single statement is also a transaction. When a statement fails, it is undone, you won't see it partially applied on the data. And like I said, the ROLLBACK is meant for the other statements in the transaction that have already executed.

Finally, a hint:

You might also want to have a look at INSERT IGNORE (you have already fulfilled the prerequisite of having a unique index). This way there won't be any errors when duplicates occur and your code keeps running without interruption.