20

I would like to have a trigger to perform following operation for inserted records:

 # pseudocode
 if new.group_id is null
    set new.group_id = new.id
 else
    # don't touch it
 end

More clearly: say I have one table with three columns: id primary key, group_id int, value varchar.

When I insert with group_id like that:

INSERT INTO table(value, group_id) VALUES ('a', 10)

I'd like to have:

id | group_id | value
---+----------+------
 1 |       10 | a

but when I omit group_id:

INSERT INTO table(value) VALUES ('b')

it should be automatically set to the id of this record:

id | group_id | value
---+----------+------
 2 |        2 | b

Is it possible with a trigger? (I know I can update the record after inserting but having the trigger would be nicer.)

szeryf
  • 3,197
  • 3
  • 27
  • 28

7 Answers7

31

I don't know of any way to do this in one statement, even using a trigger.

The trigger solution that @Lucky suggested would look like this in MySQL:

CREATE TRIGGER MyTrigger BEFORE INSERT ON MyTable
FOR EACH ROW BEGIN
  SET NEW.group_id = COALESCE(NEW.group_id, NEW.id);
END

However, there's a problem. In the BEFORE INSERT phase, the auto-generated id value hasn't been generated yet. So if group_id is null, it defaults to NEW.id which is always 0.

But if you change this trigger to fire during the AFTER INSERT phase, so you have access to the generated value of NEW.id, you can't modify column values.

MySQL doesn't support expressions for the DEFAULT of a column, so you can't declare this behavior in the table definition either. *Update: MySQL 8.0.13 supports DEFAULT (<expression>) but the expression still can't depend on an auto-increment value (this is documented).

The only solution is to do the INSERT, and then immediately do an UPDATE to change the group_id if it's not set.

INSERT INTO MyTable (group_id, value) VALUES (NULL, 'a');
UPDATE MyTable SET group_id = COALESCE(group_id, id) WHERE id = LAST_INSERT_ID();
Bill Karwin
  • 538,548
  • 86
  • 673
  • 828
  • 2
    That sucks. Even in 2016 – AA_PV Jan 05 '17 at 08:52
  • @AA_PV, this could be solved by an independent sequence object, like they do in Oracle. This has been an open feature request in MySQL since 2003: https://bugs.mysql.com/bug.php?id=1625 But in Oracle, to use the sequence, you *must* call it either in application code, which means writing code for *every* INSERT to a table you want to have an auto-inc primary key. Or else write a trigger for *every* such table. MySQL users enjoy being able to skip writing so many triggers, and just declare `AUTO_INCREMENT` as a column option instead. – Bill Karwin Jan 05 '17 at 15:31
  • yes but two AUTO_INCREMENT columns are not allowed in MySQL, right? Although I think something similar to http://stackoverflow.com/a/7345183/4411645 can be done. And to avoid the nested query, maybe it can be maintained in the application code as well, though that might be even more unsafe, not sure. – AA_PV Jan 05 '17 at 17:08
  • @AA_PV, correct, two AUTO_INCREMENT columns are not supported. But most users find this is not a problem. – Bill Karwin Jan 05 '17 at 17:45
  • _In an INSERT trigger, only NEW.col_name can be used; there is no old row. In a DELETE trigger, only OLD.col_name can be used; there is no new row. In an UPDATE trigger, you can use OLD.col_name to refer to the columns of a row before it is updated and NEW.col_name to refer to the columns of the row after it is updated._ [Syntax](https://dev.mysql.com/doc/refman/5.7/en/trigger-syntax.html) So you can use `OLD.id` to access the ID. – ravo10 Dec 10 '17 at 01:49
  • @ravo10, Please test that. I don't think `OLD.id` can be referenced in an INSERT trigger. – Bill Karwin Dec 10 '17 at 02:46
  • @BillKarwin Oh sorry. No, that is true! You can only use `OLD.id` for `DELETE` and `UPDATE`, but you need to use `NEW.id` for `INSERT`. I just made a system, and has been working fine until now. – ravo10 Dec 10 '17 at 03:45
3

This works for me

DELIMITER $$
CREATE TRIGGER `myTriggerNameHere`
BEFORE INSERT ON `table` FOR EACH ROW
BEGIN
    SET NEW.group_id = IF(NEW.group_id IS NULL, LAST_INSERT_ID()+1, NEW.group_id);
END;
$$
DELIMITER ;
1

I am answering here, as in the accepted answer Bill Karwin states:

In the BEFORE INSERT phase, the auto-generated id value hasn't been generated yet. So if group_id is null, it defaults to NEW.id which is always 0.


I have an answer for it - for OP (and the visitors to come), here are few points: You cannot update the table from where the trigger gets invoked, for it you'll get Error 1442:

Error Code: 1442
Can't update table 'MyTable' in stored function/trigger because it is already used by statement which invoked this stored function/trigger.

1.to Update the new row Use BEFORE INSERT ON trigger, this way you can update all the fields for the new row, which can be accessible via NEW operator i.e.

set NEW.group_id = NEW.id

2.Get auto_increment value before insert:

SELECT AUTO_INCREMENT FROM information_schema.TABLES WHERE TABLE_SCHEMA=DATABASE() AND TABLE_NAME='MyTable'

To sum up - the trigger SQL for 'd be something as following:

DELIMITER //
DROP TRIGGER IF EXISTS MyTrigger//
CREATE TRIGGER MyTrigger BEFORE INSERT ON MyTable
FOR EACH ROW BEGIN
    IF new.group_id IS NULL
        set @auto_id := (SELECT AUTO_INCREMENT FROM INFORMATION_SCHEMA.TABLES
                         WHERE TABLE_NAME='MyTable' AND TABLE_SCHEMA=DATABASE() ); 
        set NEW.group_id = @auto_id;
    ENF IF;
END;
//
DELIMITER ;
Nabeel Ahmed
  • 18,328
  • 4
  • 58
  • 63
1

An old topic, but I've recently struggled with it. If you still interested, have you tried?:

    CREATE TRIGGER `my_trigger_BEFORE_INSERT` BEFORE INSERT ON `MyTable` FOR EACH ROW BEGIN
        IF NEW.group_id IS NULL OR LENGTH(NEW.group_id) = 0 THEN
            SET NEW.group_id = (SELECT MAX(id) +1 FROM MyTable ) ;
        END IF;
    END
Gin Sheng
  • 73
  • 1
  • 9
0

I believe that this will work for you

I have two tables

test_b: a_id, value
test_c: a_id, value

And here is a trigger on the insert of test b. It checks to see if a_id is null and if it is it inserts 0

CREATE TRIGGER test_b AFTER INSERT ON test_b
  FOR EACH ROW 
    INSERT INTO test_c (a_id, value) VALUES (IFNULL(NEW.a_id, 0),NEW.value)
Justin Giboney
  • 3,271
  • 2
  • 19
  • 18
-3

This trigger should do what you asked.

   CREATE TRIGGER mytrigger BEFORE INSERT ON mytable
          IF new.group_id IS NULL
            SET new.group_id = new.id
          END IF
      END;

It's copied from a very similar example in the MYSQL documentation page.

Lucky
  • 646
  • 4
  • 9
  • Does not work on MySQL: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'WHERE new.group_id IS NULL END' at line 1 – szeryf Sep 25 '09 at 11:12
  • You should be able to get the exact syntax from the examples at http://dev.mysql.com/doc/refman/5.1/en/create-trigger.html - I didn't test. I'll try an edit – Lucky Sep 27 '09 at 08:35
  • The NEW.id is not set and if group_id is unique, this will fail on the second row being inserted. – Richard A Quadling Oct 31 '19 at 18:08
-4

A trigger seems like overkill in this situation. Simply apply a default.

CREATE TABLE `test` (
    `id` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT,
    `value` varchar(100) NOT NULL,
    `group_id` TINYINT(3) UNSIGNED NOT NULL DEFAULT '2'
)
  • I'm sorry but you missed the point. It's not about inserting literal value `2` as `group_id`. I want the `id` of this record as the `group_id` unless I specify `group_id` explicitly. – szeryf Jan 16 '13 at 19:33