How to remove special characters in column with MySQL?

Question

I participate in a project and met a bad design by others,we have a table called task and each task has many users,the task table is as below:

+----------------+---------------+------+-----+---------+-------+
| Field          | Type          | Null | Key | Default | Extra |
+----------------+---------------+------+-----+---------+-------+
| id             | varchar(40)   | NO   | PRI | NULL    |       |
| name           | varchar(100)  | YES  |     | NULL    |       |
| task_users     | varchar(1000) | YES  |     | NULL    |       |

and the user is stored like aaa,bbb,ccc in task_users column,which means many user id are put into one column,I know this a very bad design but since it's an old project,I can not modify the table design.

Now I have a problem,if the user is deleted,how can I remove it from the task_users column?

the user id is generated by UUID and it's at fixed length with 32 characters,so each user id is unique,such as 40cf5f01eb2f4d2c954412f27b3bf6eb,but the problem is that the user id may appaer in any position of the task_users column,so I do not know how to remove it

aaa,40cf5f01eb2f4d2c954412f27b3bf6eb,bbb -- in center
40cf5f01eb2f4d2c954412f27b3bf6eb,aaa,bbb -- in head
aaa,bbb,40cf5f01eb2f4d2c954412f27b3bf6eb -- in end

when remove the user id,the updated result is like

   aaa,bbb

I want to know can we use one update sql to remove the specified user id and still keep the same data format?

Note:I am doing it in a MySQL stored procedure,additional variable may be helpful,but I still want to just use one sql to do it,the MySQL version is 5.0

Thanks in advance!

Answer 1

I think we can do this with a single query:

UPDATE yourTable
SET task_users = SUBSTRING(
    REPLACE(CONCAT(',', task_users, ','), CONCAT(',', uid, ','), ','),
    2,
    LENGTH(task_users) - LENGTH(uid) - 1)
WHERE task_users REGEXP CONCAT('[[:<:]]', uid, '[[:>:]]');

Here is a link to a demo (to be used only for testing purposes):

Demo

This answer uses a trick, by which we append commas to the start and end of the task_users string. Then, we compare a given user ID by also appending commas to its start and end. If a match is found, we replace with just a single comma. But, this leaves the replacement still with its starting and ending commas, so we remove those with a substring operation.

SQL Olympics aside, hopefully you can see by the complexity of these answers that working with CSV data in a SQL database can be a real headache. Maybe you can even use this page as evidence to your colleagues that the table design needs to change.

Answer 2

Try this CASE expression where uid is the argument to your stored procedure...

UPDATE `task` SET `task_users` = CASE
  -- at the start
  WHEN `task_users` LIKE CONCAT(uid, ',%')
    THEN REPLACE(`task_users`, CONCAT(uid, ','), '')
  -- at the end
  WHEN `task_users` LIKE CONCAT('%,', uid)
    THEN REPLACE(`task_users`, CONCAT(',', uid), '')
  -- in the middle
  WHEN `task_users` LIKE CONCAT('%,', uid, ',%')
    THEN REPLACE(`task_users`, CONCAT(',', uid, ','), ',')
  -- only that user
  ELSE ''
END
WHERE `task_users` LIKE CONCAT('%', uid, '%');

Demo ~ http://sqlfiddle.com/#!9/1d2baa/1

---

Original "four queries" answer below

-- only that user
UPDATE `task`
SET `task_users` = ''
WHERE `task_users` = uid;

-- at start
UPDATE `task`
SET `task_users` = REPLACE(`task_users`, CONCAT(uid, ','), '')
WHERE `task_users` LIKE CONCAT(uid, ',%');

-- at end
UPDATE `task`
SET `task_users` = REPLACE(`task_users`, CONCAT(',', uid), '')
WHERE `task_users` LIKE CONCAT('%,', uid);

-- in the middle
UPDATE `task`
SET `task_users` = REPLACE(`task_users`, CONCAT(',', uid, ','), ',')
WHERE `task_users` LIKE CONCAT('%,', uid, ',%');

Demo ~ http://sqlfiddle.com/#!9/8e9b9bb/1