I have one table: drupal.comments, with amongst others, the columns:
cid: primary key
uid: foreign key to users table, optional
name: varchar, optional
email: varchar, optional
The description says: UID is optional, if 0, comment made by anonymous; in that case the name/email is set.
I want to split this out into two tables rails.comments and rails.users, where there is always a user:
id: primary key
users_id: foreign key, always set.
So, for each drupal.comment, I need to create either a new user from the drupal.comments.name/drupal.comments.email and a rails.comment where the rails.comment.users_id is the ID of the just created user.
Or, if username/email already exists for a rails.user, I need to fetch that users_id and use that on the new comment record as foreign key.
Or, if drupal.comment.uid is set, I need to use that as users_id.
Is this possible in SQL? Are queries that fetch from one source, but fill multiple tables possible in SQL? Or is there some (My)SQL trick to achieve this? Or should I simply script this in Ruby, PHP or some other language instead?
You could do this with a TRIGGER.
Here's some pseudo-code to illustrate this technique:
DELIMITER $$
DROP TRIGGER IF EXISTS tr_b_ins_comments $$
CREATE TRIGGER tr_b_ins_comments BEFORE INSERT ON comments FOR EACH ROW BEGIN
DECLARE v_uid INT DEFAULT NULL;
/* BEGIN pseudo-code */
IF (new.uid IS NULL)
THEN
-- check for existing user with matching name and email address
select user_id
into v_uid
from your_user_table
where name = new.name
and email = new.email;
-- if no match, create a new user and get the id
IF (v_uid IS NULL)
THEN
-- insert a new user into the user table
insert into your_user_table ...
-- get the new user's id (assuming it's auto-increment)
set v_uid := LAST_INSERT_ID();
END IF;
-- set the uid column
SET new.uid = v_uid;
END IF;
/* END pseudo-code */
END $$
DELIMITER ;
I searched further and found that, apparently, it is not possible to update/insert more then one table in a single query in MySQL.
The solution would, therefore have to be scripted/programmed outside of SQL.
Related
I have a program in Laravel where after users register they need a badge number, I want that number to be generated randomly after they register in the database. I should use triggers but I struggle with syntax.
users table
id bigint(20)
name varchar(255)
surname varchar(255)
nr_legitimatie varchar(255)
I want that 'nr_legitimatie' field to be unique.
This is what I tried but with no success
Trigger
CREATE OR REPLACE TRIGGER numar_leg
AFTER INSERT ON users
FOR EACH ROW
DECLARE
legitimatie VARCHAR(191)
BEGIN
legitimatie =('
SELECT FLOOR(RAND() * 99999) AS random_num
FROM numbers_mst
WHERE "random_num" NOT IN (SELECT my_number FROM numbers_mst)
LIMIT 1' );
set `users`.`nr_legitimatie` = legitimatie;
END;
Here's an example of a MySQL BEFORE INSERT trigger that assigns a value to the nr_legitimatie column.
DELIMITER $$
DROP TRIGGER numar_leg$$
CREATE TRIGGER numar_leg
BEFORE INSERT ON users
FOR EACH ROW
BEGIN
DECLARE li_nrn BIGINT DEFAULT NULL;
DECLARE li_cnt BIGINT DEFAULT 1;
WHILE li_cnt > 0 DO
-- generate a new random number
SELECT FLOOR(RAND()*99999) AS nrn INTO li_nrn;
-- check if the new random number is already used
SELECT COUNT(1) INTO li_cnt FROM users u WHERE u.nr_legitimatie = li_nrn;
END WHILE;
SET NEW.nr_legitimatie := li_nrn;
END$$
DELIMITER ;
Note that this does not guarantee that the value assigned to the nr_legitimatie will be unique, because the code in the trigger is subject to a race condition. There is potential for two (or more) simultaneous sessions to each discover the same random number is not yet "unused", and each session will use it. (The check for an existing value precedes the assignment to the column.)
If we want to guarantee uniqueness, we should add a UNIQUE constraint (UNIQUE KEY) on the column in the users table.
We can also use a separate table to track the numbers that are used, with a UNIQUE constraint on the column, we can attempt inserts, and catch the error when an attempt to insert a duplicate is made.
If we introduce a tracking table, e.g.
CREATE TABLE nrn (nrn BIGINT PRIMARY KEY) ;
Then we can avoid the race condition, making the test for existing duplicate and reservation of the new value at the same time. Something like this:
DELIMITER $$
CREATE TRIGGER numar_leg
BEFORE INSERT ON users
FOR EACH ROW
BEGIN
DECLARE li_nrn BIGINT DEFAULT NULL;
DECLARE li_dup BIGINT DEFAULT 1;
DECLARE CONTINUE HANDLER FOR 1062 SET li_dup := 1;
WHILE li_dup > 0 DO
SELECT FLOOR(RAND()*99999) AS nrn INTO li_nrn;
SET li_dup := 0;
INSERT INTO nrn (nrn) VALUES (li_nrn);
END WHILE;
SET NEW.nr_legitimatie := li_nrn;
END$$
DELIMITER ;
The edge case here is the trigger is executed, a new random number is generated and reserved, but the insert into the users table fails for some reason, and we don't issue a ROLLBACK. If we issue a ROLLBACK, then our new random number reservation will also be rolled back (unless nrn is a MyISAM table).
CREATE PROCEDURE Sp_IU_Group(
GID int,
GroupName nvarchar(200),
UserID int,
Status int
)
BEGIN
IF GID=0 THEN
Insert into tblGroup (GroupName,UserID,Status)
values (GroupName,UserID,Status);
else
update tblGroup set GroupName=GroupName,UserID=UserID,Status=Status WHERE GID=GID;
END IF;
END
This query:
update tblGroup set GroupName=GroupName,UserID=UserID,Status=Status WHERE GID=GID
Will update every record in the table... to itself. This matches every record, because this is always true:
WHERE GID=GID
And this updates a value to itself:
GroupName=GroupName
The problem is that you're using the same names for multiple things. Give things different names. Something as simple as this:
CREATE PROCEDURE Sp_IU_Group(
GIDNew int,
GroupNameNew nvarchar(200),
UserIDNew int,
StatusNew int
)
(Or use any other standard you want to distinguish the variables from the database objects, such as prepending them with a special character like an #.)
Then the query can tell the difference:
update tblGroup set GroupName=GroupNameNew,UserID=UserIDNew,Status=StatusNew WHERE GID=GIDNew
(Modify the rest of the stored procedure for the new variable names accordingly, of course.)
Basically, as a general rule of thumb, never rely on the code to "know what you meant". Always be explicit and unambiguous.
What I'm trying to achieve is, I want to automate the values of the table between the users and folders table. Since it's a many-to-many relationship I created the user_folders table. Currently the server (nodejs) gets the request with userid, clientfolderid and some an array of bookmarks (which are not important now). It checks if the user already has this folder, by selecting from the user_folders table and if it's not existing it inserts a new row into the folder table. Then it has to send another statement to insert into the user_folders table.
So I have to "manually" keep the users_folder table updated.I guess this is a common problem and wanted to know if there is a pattern or a proven solution? The odd thing is that MySQL automatically handles the deletion of rows with an AFTER DELETE trigger but there is no (at least that I know of) automation with an AFTER INSERT trigger.
As I already said an AFTER INSERT trigger could possibly solve it, but I think it's not possible to pass some extra parameters to the AFTER INSERT trigger. This would be the user_id and the folder_client_id in my case.
I was thinking of a solution that I could create another table called tmp_folder which would look like:
tmp_folder
-- id
-- title
-- changed
-- user_id
-- folder_client_id
Then create an AFTER INSERT trigger on this table which inserts into folders and user_folders and then removes the row from tmp_folder again. Would this be the right way or is there a better one?
I would basically do the same with the bookmarks and user_bookmarks table. The best thing would be if it's even possible to insert a folder then the owner into the user_folders table with user_id and folder_client_id and then multiple other users into user_folders with the user_id and an default folder_client_id of -1 or something which will be updated later.
Meanwhile thanks for reading and I hope you can help me :)
PS: Is there a name for the table between 2 other tables in an m-2-m relationship?
I don't see an easy way to do this via triggers, but a stored procedure may suit you:
DELIMITER //
CREATE PROCEDURE
add_user_folder(
IN u_user_id BIGINT UNSIGNED,
IN u_folder_client_id BIGINT UNSIGNED,
IN v_title VARCHAR(255)
)
BEGIN
DECLARE u_found INT UNSIGNED DEFAULT 0;
SELECT
1 INTO u_found
FROM
user_folders
WHERE
user_id = u_user_id AND
folder_client_id = u_folder_client_id;
IF IFNULL(u_found, 0) = 0 THEN
START TRANSACTION;
INSERT INTO
folders
SET
title = v_title,
changed = UNIX_TIMESTAMP();
INSERT INTO
user_folders
SET
user_id = u_user_id,
folder_id = LAST_INSERT_ID(),
folder_client_id = u_folder_client_id;
COMMIT;
END IF;
END;
//
Is it possible to enforce uniqueness across two tables in MySQL?
I have two tables, both describing users. The users in these tables were for two different systems previously, however now we're merging our authentication systems and I need to make sure that there are unique usernames across these two tables. (it's too much work to put them all into one table right now).
You can't declare a UNIQUE constraint across multiple tables. MySQL 8.0 supports CHECK constraints, but those constraints cannot reference other tables. But you can design a trigger to search for the matching value in the other table. Here's a test SQL script:
DROP TABLE IF EXISTS foo;
CREATE TABLE FOO (username VARCHAR(10) NOT NULL);
DROP TABLE IF EXISTS bar;
CREATE TABLE BAR (username VARCHAR(10) NOT NULL);
DROP TRIGGER IF EXISTS unique_foo;
DROP TRIGGER IF EXISTS unique_bar;
DELIMITER //
CREATE TRIGGER unique_foo BEFORE INSERT ON foo
FOR EACH ROW BEGIN
DECLARE c INT;
SELECT COUNT(*) INTO c FROM bar WHERE username = NEW.username;
IF (c > 0) THEN
-- abort insert, because foo.username should be NOT NULL
SET NEW.username = NULL;
END IF;
END//
CREATE TRIGGER unique_bar BEFORE INSERT ON bar
FOR EACH ROW BEGIN
DECLARE c INT;
SELECT COUNT(*) INTO c FROM foo WHERE username = NEW.username;
IF (c > 0) THEN
-- abort insert, because bar.username should be NOT NULL
SET NEW.username = NULL;
END IF;
END//
DELIMITER ;
INSERT INTO foo VALUES ('bill'); -- OK
INSERT INTO bar VALUES ('bill'); -- Column 'username' cannot be null
You also need similar triggers ON UPDATE for each table, but you shouldn't need any triggers ON DELETE.
the best way to do this is to declare another table with the unique columns, and have the multiple tables reference these tables
Maybe not direct answer to your question, but:
You should consider rewriting your code and restructuring your database to unite those two tables into one.
The design you are trying to enforce now will complicate your code and database schema and it will make any further upgrade to other database software or frameworks harder.
You could add an extra table with a single column as a primary key. Then create a trigger on each of your old user tables to insert the id into this extra table.
create table users1 (
user_id integer primary key,
username varchar(8) not null unique
);
create table users2 (
user_id integer primary key,
username varchar(8) not null unique
);
create table all_usernames (
username varchar(8) primary key
);
create trigger users1_insert before insert on users1 for each row
insert into all_usernames values(new.username);
create trigger users2_insert before insert on users2 for each row
insert into all_usernames values(new.username);
create trigger users1_update before update on users1 for each row
update all_usernames set username = new.username
where username = old.username;
create trigger users2_update before update on users2 for each row
update all_usernames set username = new.username
where username = old.username;
create trigger users1_delete before delete on users1 for each row
delete from all_usernames where username = old.username;
create trigger users2_delete before delete on users2 for each row
delete from all_usernames where username = old.username;
You can then populate the table with
insert into all_usernames select username from users1;
insert into all_usernames select username from users2;
Obviously if there are already duplicates in the two tables you will have to solve that problem by hand. Moving forward, you could write a trigger that checks both tables to see if the value already exists, and then apply that to both tables.
Would changing the type of the ID column be affordable? Then you could go for GUIDs which would be unique across as many tables as you want.
I don't know MySQL but this is how you can do it in Oracle and I believe MySQL does support materialized views too.
You create a materialized view on those two tables. And you add a unique constraint on this view.
This view needs to be refreshed every time a change to one of the two base tables is committed.
I have a trigger in which I want to have a variable that holds an INT I get from a SELECT, so I can use it in two IF statements instead of calling the SELECT twice. How do you declare/use variables in MySQL triggers?
You can declare local variables in MySQL triggers, with the DECLARE syntax.
Here's an example:
DROP TABLE IF EXISTS foo;
CREATE TABLE FOO (
i SERIAL PRIMARY KEY
);
DELIMITER //
DROP TRIGGER IF EXISTS bar //
CREATE TRIGGER bar AFTER INSERT ON foo
FOR EACH ROW BEGIN
DECLARE x INT;
SET x = NEW.i;
SET #a = x; -- set user variable outside trigger
END//
DELIMITER ;
SET #a = 0;
SELECT #a; -- returns 0
INSERT INTO foo () VALUES ();
SELECT #a; -- returns 1, the value it got during the trigger
When you assign a value to a variable, you must ensure that the query returns only a single value, not a set of rows or a set of columns. For instance, if your query returns a single value in practice, it's okay but as soon as it returns more than one row, you get "ERROR 1242: Subquery returns more than 1 row".
You can use LIMIT or MAX() to make sure that the local variable is set to a single value.
CREATE TRIGGER bar AFTER INSERT ON foo
FOR EACH ROW BEGIN
DECLARE x INT;
SET x = (SELECT age FROM users WHERE name = 'Bill');
-- ERROR 1242 if more than one row with 'Bill'
END//
CREATE TRIGGER bar AFTER INSERT ON foo
FOR EACH ROW BEGIN
DECLARE x INT;
SET x = (SELECT MAX(age) FROM users WHERE name = 'Bill');
-- OK even when more than one row with 'Bill'
END//
CREATE TRIGGER clearcamcdr AFTER INSERT ON `asteriskcdrdb`.`cdr`
FOR EACH ROW
BEGIN
SET #INC = (SELECT sip_inc FROM trunks LIMIT 1);
IF NEW.billsec >1 AND NEW.channel LIKE #INC
AND NEW.dstchannel NOT LIKE ""
THEN
insert into `asteriskcdrdb`.`filtre` (id_appel,date_appel,source,destinataire,duree,sens,commentaire,suivi)
values (NEW.id,NEW.calldate,NEW.src,NEW.dstchannel,NEW.billsec,"entrant","","");
END IF;
END$$
Dont try this # home
`CREATE TRIGGER `category_before_ins_tr` BEFORE INSERT ON `category`
FOR EACH ROW
BEGIN
**SET #tableId= (SELECT id FROM dummy LIMIT 1);**
END;`;
I'm posting this solution because I had a hard time finding what I needed. This post got me close enough (+1 for that thank you), and here is the final solution for rearranging column data before insert if the data matches a test.
Note: this is from a legacy project I inherited where:
The Unique Key is a composite of rridprefix + rrid
Before I took over there was no constraint preventing duplicate unique keys
We needed to combine two tables (one full of duplicates) into the main table which now has the constraint on the composite key (so merging fails because the gaining table won't allow the duplicates from the unclean table)
on duplicate key is less than ideal because the columns are too numerous and may change
Anyway, here is the trigger that puts any duplicate keys into a legacy column while allowing us to store the legacy, bad data (and not trigger the gaining tables composite, unique key).
BEGIN
-- prevent duplicate composite keys when merging in archive to main
SET #EXIST_COMPOSITE_KEY = (SELECT count(*) FROM patientrecords where rridprefix = NEW.rridprefix and rrid = NEW.rrid);
-- if the composite key to be introduced during merge exists, rearrange the data for insert
IF #EXIST_COMPOSITE_KEY > 0
THEN
-- set the incoming column data this way (if composite key exists)
-- the legacy duplicate rrid field will help us keep the bad data
SET NEW.legacyduperrid = NEW.rrid;
-- allow the following block to set the new rrid appropriately
SET NEW.rrid = null;
END IF;
-- legacy code tried set the rrid (race condition), now the db does it
SET NEW.rrid = (
SELECT if(NEW.rrid is null and NEW.legacyduperrid is null, IFNULL(MAX(rrid), 0) + 1, NEW.rrid)
FROM patientrecords
WHERE rridprefix = NEW.rridprefix
);
END
Or you can just include the SELECT statement in the SQL that's invoking the trigger, so its passed in as one of the columns in the trigger row(s). As long as you're certain it will infallibly return only one row (hence one value). (And, of course, it must not return a value that interacts with the logic in the trigger, but that's true in any case.)
As far I think I understood your question
I believe that u can simply declare your variable inside "DECLARE"
and then after the "begin" u can use 'select into " you variable" ' statement.
the code would look like this:
DECLARE
YourVar varchar(50);
begin
select ID into YourVar from table
where ...