Question

I have got an older database for which (at some really questionable and obscure reason I do not like to put too much on topic here) I want to randomize or shuffle the primary keys.

I right now have auto-increment fields in the Mysql database tables.

I have not many relations, those that exist are not defined as foreign keys. The relationships do not need to be preserved.

All I’m looking for is to take the current values of the primary keys and replace it with a random value out of those like:

ID := new(ID)

Where the new function returns a value from the set of all OLD ids with a 1:1 match. E.g.

2 := 3
3 := 2

But not

2 := 3
3 := 3

Is there a way to change the data in the database with (ideally) a single SQL query per table?

Edit: I do not have really strict requirements. Consider to have exclusive access to the database if it helps, including changing constraints on the primary key back and forth, e.g. alter the table, do the operation, alter the table to previous schema. It is also possible to add another column for the new (or old) PK value.

Answer 1

Just a scetch of the procedure. Create two temporary tables

CREATE TABLE temp_old
( ai INT NOT NULL AUTO_INCREMENT
, id INT NOT NULL
, PRIMARY KEY (ai)
, INDEX old_idx (id, ai)
) ENGINE = InnoDB ;

CREATE TABLE temp_new
( ai INT NOT NULL AUTO_INCREMENT
, id INT NOT NULL
, PRIMARY KEY (ai)
, INDEX new_idx (id, ai)
) ENGINE = InnoDB ;

Copy the id values in different orders to the two tables (randomly in the 2nd table):

INSERT INTO temp_old
  (id)
SELECT id
FROM tableX 
ORDER BY id ;

INSERT INTO temp_new
  (id)
SELECT id
FROM tableX 
ORDER BY RAND() ;

Then we drop the primary key:

ALTER TABLE tableX
  DROP PRIMARY KEY ; 

to run the actual UPDATE statement:

UPDATE tableX AS t
  JOIN temp_old AS o
    ON o.id = t.id
  JOIN temp_new AS n
    ON n.ai = o.ai
SET t.id = n.id ;

Then recreate the primary key and drop the temp tables:

ALTER TABLE tableX
  ADD PRIMARY KEY (id) ; 

DROP TABLE temp_old ;
DROP TABLE temp_new ;

Tested in SQL-Fiddle