I am new to MySQL. Please can you advice on how can i modify below function to ensure it does not throw locking errors when called by multiple users at the same time.
CREATE FUNCTION `get_val`(`P_TABLE` VARCHAR(50)) RETURNS int(11)
DETERMINISTIC
BEGIN
DECLARE pk_value INT DEFAULT 0;
DECLARE pk_found INT DEFAULT 0;
SELECT 1 INTO pk_found FROM pk_keys WHERE TABLE_NAME = P_TABLE;
IF pk_found = 1
THEN
UPDATE pk_keys SET TABLE_VALUE = (TABLE_VALUE + 1 ) WHERE TABLE_NAME = P_TABLE;
ELSE
INSERT INTO pk_keys VALUES ( P_TABLE, 1 );
END IF;
SELECT TABLE_VALUE INTO pk_value FROM pk_keys WHERE TABLE_NAME = P_TABLE;
RETURN pk_value;
END
thanks
CREATE FUNCTION `get_val`(`P_TABLE` VARCHAR(50)) RETURNS int(11)
DETERMINISTIC
MODIFIES SQL DATA
BEGIN
DECLARE pk_value INT DEFAULT 0;
IF EXISTS (SELECT 1 FROM pk_keys WHERE TABLE_NAME = P_TABLE)
THEN
SELECT TABLE_VALUE + 1 INTO pk_value FROM pk_keys WHERE TABLE_NAME = P_TABLE FOR UPDATE;
UPDATE pk_keys SET TABLE_VALUE = pk_value WHERE TABLE_NAME = P_TABLE;
ELSE
SET pk_value = 1;
INSERT INTO pk_keys VALUES ( P_TABLE, pk_value );
END IF;
RETURN pk_value;
END
Have a look at SELECT ... FOR UPDATE and SELECT ... LOCK IN SHARE MODE Locking Reads
Let us look at another example: We have an integer counter field in a
table child_codes that we use to assign a unique identifier to each
child added to table child. It is not a good idea to use either
consistent read or a shared mode read to read the present value of the
counter because two users of the database may then see the same value
for the counter, and a duplicate-key error occurs if two users attempt
to add children with the same identifier to the table.
Here, LOCK IN SHARE MODE is not a good solution because if two users
read the counter at the same time, at least one of them ends up in
deadlock when it attempts to update the counter.
To implement reading and incrementing the counter, first perform a
locking read of the counter using FOR UPDATE, and then increment the
counter. For example:
SELECT counter_field FROM child_codes FOR UPDATE;
UPDATE child_codes SET counter_field = counter_field + 1;
A SELECT ... FOR UPDATE reads the latest available data, setting
exclusive locks on each row it reads. Thus, it sets the same locks a
searched SQL UPDATE would set on the rows.
Also I replaced your if condition. EXISTS stops as soon as a row is found.
Related
I have a BEFORE INSERT trigger on a sameTable,
BEGIN
set new.seq = (select (max(seq) + 1) from sameTable);
END;
Question:
Does it immune to concurrency?
Let's say, for simplification, I need seq to be unique without enforcing unique-index.
Using: MariaDB, InnoDB engine.
EDIT
To be clear that I can't make use of auto_increment, the trigger looks like this:
BEGIN
case
when new.seq is null then
set new.seq = (
select
(ifnull(max(seq),0) + 1)
from sameTable
where
aField = new.aField
);
else
set new.seq = new.seq;
end case;
END;
It's now clear that seq is not behaving like auto_increment since there's aField. Surely aField-seq is composite-unique, but, let's say, again, for simplification, I need aField-seq to be unique without enforcing composite-unique-index.
I do know that using SELECT ... FOR UPDATE within a transaction,
START TRANSACTION
set val = (
select
(ifnull(max(seq),0) + 1)
from sameTable
where
aField = new.aField
for update
);
case
when seq is null then
set seq = val;
else
set seq = seq;
end case;
COMMIT;
will lock sameTable. Although, I don't know if it works within a trigger, but that's another question.
My question is: If I do a single query, would it lock the table or not? Preferably, why?
I have two (2) databases of dissimilar Schematics,
db1 migrated from MSSQL to MYSQL
and
db2 created from Laravel Migration.
Here's the challenge:
The tables of db1 do not have id columns (Primary Key) like is easily found on db2 tables. So I kept getting the warning message:
Current selection does not contain a unique column. Grid edit, checkbox, Edit, Copy and Delete features are not available.
So I had to inject the id columns on the tables in the db1
I need to extract fields [level_name, class_name] from stdlist in db1,
Create levels (id,level_name,X,Y) on db2
classes (id,class_name,level_id) on db2
To throw more light: The level_id should come from the already created levels table
I have already succeeded in extracting the first instance using the following snippet:
First Query to Create Levels
INSERT INTO db2.levels(level_name,X,Y)
SELECT class_name as level_name,1 as X,ClassAdmitted as Y
FROM db1.stdlist
GROUP BY ClassAdmitted;
This was successful.
Now, I need to use the newly created ids in levels table to fill up level_id column in the classes table.
For that to be possible, must I re-run the above selection schematics? Is there no better way to maybe join the table column from db1.levels to db2.stdlist and extract the required fields for the new insert schematics.
I'll appreciate any help. Thanks in advance.
Try adding a column for Processed and then do a while exists loop
INSERT INTO db2.levels(level_name,X,Y)
SELECT class_name as level_name,1 as X,ClassAdmitted as Y, 0 as Processed
FROM db1.stdlist
GROUP BY ClassAdmitted;
WHILE EXISTS(SELECT * FROM db2.levels WHERE Processed = 0)
BEGIN
DECLARE #level_name AS VARCHAR(MAX)
SELECT TOP 1 #level_name=level_name FROM db2.levels WHERE Processed = 0
--YOUR CODE
UPDATE db2.levels SET Processed=1 WHERE level_name=#level_name
END
You may need to dump into a temp table first and then insert into your real table (db2.levels) when you're done processing. Then you wouldn't need the Unnecessary column of processed on the final table.
This is what worked for me eventually:
First, I picked up the levels from the initial database thus:
INSERT INTO db2.levels(`name`,`school_id`,`short_code`)
SELECT name ,school_id,short_code
FROM db1.levels
GROUP BY name
ORDER BY CAST(IF(REPLACE(name,' ','')='','0',REPLACE(name,' ','')) AS UNSIGNED
INTEGER) ASC;
Then I created a PROCEDURE for the classes insertion
CREATE PROCEDURE dowhileClasses()
BEGIN
SET #Level = 1;
SET #Max = SELECT count(`id`) FROM db2.levels;
START TRANSACTION;
WHILE #Level <= #Max DO
BEGIN
DECLARE val1 VARCHAR(255) DEFAULT NULL;
DECLARE val2 VARCHAR(255) DEFAULT NULL;
DECLARE bDone TINYINT DEFAULT 0;
DECLARE curs CURSOR FOR
SELECT trim(`Class1`)
FROM db1.dbo_tblstudent
WHERE CAST(IF(REPLACE(name,' ','')='','0',REPLACE(name,' ','')) AS UNSIGNED INTEGER) =#Level
GROUP BY `Class1`;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET bDone = 1;
OPEN curs;
SET bDone = 0;
REPEAT
FETCH curs INTO val1;
IF bDone = 0 THEN
SET #classname = val1;
SET #levelID = (SELECT id FROM db2.levels WHERE short_code=#Level limit 1);
SET #schoolId = 1;
SET #classId = (SELECT `id` FROM db2.classes where class_name = #classname and level_id= #levelID limit 1);
IF #classId is null and #classname is not null THEN
INSERT INTO db2.classes(class_name,school_id,level_id)
VALUES(#classname,#schoolId,#levelID);
END IF;
END IF;
UNTIL bDone END REPEAT;
CLOSE curs;
END;
SELECT CONCAT('lEVEL: ',#Level,' Done');
SET #Level = #Level + 1;
END WHILE;
END;
//
delimiter ;
CALL dowhileClasses();
With this, I was able to dump The classes profile matching the previously created level_ids.
The whole magic relies on the CURSOR protocol.
For further details here is one of the documentations I used.
I have created the following trigger to populate a field with a unique integer value.
I am using this in an InnoDB table and there is a UNIQUE key constraint on the field.
Is it possible that two concurrent inserts produce the same value and due to the unique constraint one of them fails, or are triggers "atomic"?
Are there any other issues with this code that I may not have thought of?
Is there a better whay to get the behaviour I want? Maby some soft of isomorphism on the auto increment value?
CREATE TRIGGER `generate_customer_id` BEFORE INSERT ON `users`
FOR EACH ROW BEGIN
DECLARE i INT;
DECLARE duplicate INT DEFAULT 1;
DECLARE tries INT DEFAULT 0;
WHILE duplicate > 0 DO
SET tries = tries + 1;
IF tries > 100 THEN
SIGNAL SQLSTATE '45000'
SET MESSAGE_TEXT = 'no customer id found after 100 tries', MYSQL_ERRNO = 1001;
END IF;
SET i = ROUND((RAND() * (999999999-100000000))+100000000);
SET duplicate = (SELECT COUNT(*) FROM users WHERE customer_id = i);
END WHILE;
SET NEW.customer_id = i;
END
A possible alternative to the random number generator is to create a customer_ids table with an auto_increment primary key - create a new row in the table to create a unique customer_id for the users table
i have a trigger like below, the logic is to change FID status after fidRule status changed.
in my app, i update 1 row in each statement,
but i found sometimes(very rare) the trigger not firing.
ALTER TRIGGER [dbo].[triggerSetFIDStatus]
ON [dbo].[FIDRules]
AFTER UPDATE
AS
BEGIN
set nocount on
DECLARE #ruleStatus INT
DECLARE #newRuleStatus INT
DECLARE #FIDAlertStatus INT
DECLARE #FIDId INT
DECLARE #isFIDEnabled int
DECLARE #ruleId INT
SELECT #ruleStatus = deleted.alertStatus,
#FIDId = deleted.FIDID,
#ruleId = deleted.id
from
deleted
SELECT #newRuleStatus = inserted.alertStatus
from
inserted
SELECT #FIDAlertStatus = alertStatus,
#isFIDEnabled= isEnabled
FROM FID
WHERE id = #FIDId
IF #FIDAlertStatus <> #newRuleStatus
BEGIN
-- change FID-status by FIDRule-status
UPDATE [dbo].[FID] SET alertStatus=#newRuleStatus WHERE id=#FIDId
END
IF #newRuleStatus >= 0 AND #newRuleStatus <> #ruleStatus
UPDATE [dbo].[FIDRules] SET isAlertStatChanged=1, AlertStatChangeTime = SYSUTCDATETIME() WHERE id=#ruleId
END
The trigger will not be fired if the UPDATE statement fails or another triggers fails to execute before this trigger is fired.
One comment about your trigger itself:
You are expecting one records from DELETED which is not always correct.
Please make your trigger robust enough in case DELETED contains multiple records
-- What if deleted contains multiple records?
SELECT #ruleStatus = deleted.alertStatus,
#FIDId = deleted.FIDID,
#ruleId = deleted.id
FROM
deleted
You can either use SELECT TOP(1) or make sure your trigger is able to handle multiple records from the DELETED list.
I have a table which contains relative large data,
so that it takes too long for the statements below:
SELECT MIN(column) FROM table WHERE ...
SELECT MAX(column) FROM table WHERE ...
I tried index the column, but the performance still does not suffice my need.
I also thought of caching min and max value in another table by using trigger or event.
But my MySQL version is 5.0.51a which requires SUPER privilege for trigger and does not support event.
It is IMPOSSIBLE for me to have SUPER privilege or to upgrade MySQL.
(If possible, then no need to ask!)
How to solve this problem just inside MySQL?
That is, without the help of OS.
If your column is indexed, you should find min(column) near instantly, because that is the first value MySQL will find.
Same goes for max(column) on an indexed column.
If you cannot add an index for some reason the following triggers will cache the MIN and MAX value in a separate table.
Note that TRUE = 1 and FALSE = 0.
DELIMITER $$
CREATE TRIGGER ai_table1_each AFTER INSERT ON table1 FOR EACH ROW
BEGIN
UPDATE db_info i
SET i.minimum = LEAST(i.minimum, NEW.col)
,i.maximum = GREATEST(i.maximum, NEW.col)
,i.min_count = (i.min_count * (new.col < i.minumum))
+ (i.minimum = new.col) + (i.minimum < new.col)
,i.max_count = (i.max_count * (new.col > i.maximum))
+ (i.maximum = new.col) + (new.col > i.maximum)
WHERE i.tablename = 'table1';
END $$
CREATE TRIGGER ad_table1_each AFTER DELETE ON table1 FOR EACH ROW
BEGIN
DECLARE new_min_count INTEGER;
DECLARE new_max_count INTEGER;
UPDATE db_info i
SET i.min_count = i.min_count - (i.minimum = old.col)
,i.max_count = i.max_count - (i.maximum = old.col)
WHERE i.tablename = 'table1';
SELECT i.min_count INTO new_min_count, i.max_count INTO new_max_count
FROM db_info i
WHERE i.tablename = 'table1';
IF new_max_count = 0 THEN
UPDATE db_info i
CROSS JOIN (SELECT MAX(col) as new_max FROM table1) m
SET i.max_count = 1
,i.maximum = m.new_max;
END IF;
IF new_min_count = 0 THEN
UPDATE db_info i
CROSS JOIN (SELECT MIN(col) as new_min FROM table1) m
SET i.min_count = 1
,i.minimum = m.new_min;
END IF;
END $$
DELIMITER ;
The after update trigger will be some mix of the insert and delete triggers.