Let’s start with the original simple scenario of checking duplicates in a single column.

Some example data, a Users table:

+----+----------------+
| Id | Email          |
+----+----------------+
|  1 | joe@bloggs.com |
|  2 | joe@bloggs.com |
|  3 | joe@bloggs.com |
|  4 | jane@doe.com   |
|  5 | jane@doe.com   |
|  6 | john@doe.com   |
+----+----------------+

You can see that joe@bloggs.com and jane@doe.com have been duplicated. This could have been prevented by putting a unique index on the Email column.

So to find what emails have duplicates in our table:

SELECT Email, COUNT(Email) AS Duplicates
FROM `Users`
GROUP BY Email
HAVING ( Duplicates > 1 )

Results:

+----------------+------------+
| Email          | Duplicates |
+----------------+------------+
| jane@doe.com   |          2 |
| joe@bloggs.com |          3 |
+----------------+------------+

So, to help us manually correct our data, what are the Ids of the duplicates? In MySQL (4.1+), we can use GROUP_CONCAT (after casting the numerical Id to a character string):

SELECT Email, COUNT(Email) AS Duplicates, GROUP_CONCAT( CAST(Id AS CHAR) ) AS Culprits
FROM `Users`
GROUP BY Email
HAVING ( Duplicates > 1 )

Our results:

+----------------+------------+----------+
| Email          | Duplicates | Culprits |
+----------------+------------+----------+
| jane@doe.com   |          2 | 4,5      |
| joe@bloggs.com |          3 | 1,2,3    |
+----------------+------------+----------+

That’s quite handy, but what about just a list of the duplicates we can go through, instead of these rows of comma-separated Ids?

This fugly query will do that for us: (I’m sure I could do this a better way but I’m tired and this works!)

SELECT Id, Email FROM `Users` WHERE Email IN
(SELECT Email FROM `Users` GROUP BY Email HAVING ( COUNT(Email) > 1 ))
ORDER BY Email

+----+----------------+
| Id | Email          |
+----+----------------+
|  4 | jane@doe.com   |
|  5 | jane@doe.com   |
|  1 | joe@bloggs.com |
|  2 | joe@bloggs.com |
|  3 | joe@bloggs.com |
+----+----------------+

Now you can edit / delete the rows you want to get rid of if you ran the query in something like phpMyAdmin.

And don’t forget, after the clean-up job, add that index to prevent duplicates re-appearing:

ALTER TABLE `Users` ADD UNIQUE (`Email`)

Now, the new scenario. What about duplicates across multiple columns? For example, our Locations table:

+----+-------------+----------+--------+
| Id | CountryCode | AreaCode | Prefix |
+----+-------------+----------+--------+
|  1 | 64          | 9        | 489    |
|  2 | 64          | 9        | 489    |
|  3 | 64          | 9        | 489    |
|  4 | 64          | 3        | 942    |
|  5 | 64          | 3        | 942    |
|  6 | 64          | 9        | 536    |
+----+-------------+----------+--------+

Here, we want to find duplicates that have the same values in the 3 columns. For example, you can see that 64-9-489 is duplicated three times, and 64-3-942 two times.

We can do this without much alteration to our original queries:

SELECT CountryCode, AreaCode, Prefix, COUNT(*) AS Duplicates
FROM `Locations`
GROUP BY CountryCode, AreaCode, Prefix
HAVING ( Duplicates > 1 )

+-------------+----------+--------+------------+
| CountryCode | AreaCode | Prefix | Duplicates |
+-------------+----------+--------+------------+
| 64          | 3        | 942    |          2 |
| 64          | 9        | 489    |          3 |
+-------------+----------+--------+------------+

Then to get the Ids:

SELECT CountryCode, AreaCode, Prefix, COUNT(*) AS Duplicates, GROUP_CONCAT( CAST(Id AS CHAR) ) AS Culprits
FROM `Locations`
GROUP BY CountryCode, AreaCode, Prefix
HAVING ( Duplicates > 1 )

+-------------+----------+--------+------------+----------+
| CountryCode | AreaCode | Prefix | Duplicates | Culprits |
+-------------+----------+--------+------------+----------+
| 64          | 3        | 942    |          2 | 4,5      |
| 64          | 9        | 489    |          3 | 1,2,3    |
+-------------+----------+--------+------------+----------+

I think you’re getting the point. Here’s to get the rows for the culprits:

SELECT Id, CountryCode, AreaCode, Prefix FROM `Locations` WHERE Id NOT IN
(SELECT Id FROM `Locations` GROUP BY CountryCode, AreaCode, Prefix HAVING ( COUNT(CountryCode) = 1 ))
ORDER BY CountryCode, AreaCode, Prefix

+----+-------------+----------+--------+
| Id | CountryCode | AreaCode | Prefix |
+----+-------------+----------+--------+
|  4 | 64          | 3        | 942    |
|  5 | 64          | 3        | 942    |
|  1 | 64          | 9        | 489    |
|  2 | 64          | 9        | 489    |
|  3 | 64          | 9        | 489    |
+----+-------------+----------+--------+

Again I’m sure there’d be an easier way to do that, but hey, it works, and for something that should be a one-off.

So how to prevent the duplicated data in our second scenario? Add a composite unique key on those columns:

ALTER TABLE `locations` ADD UNIQUE (`CountryCode`, `AreaCode`, `Prefix`)