How to delete from select in MySQL?

Question

This code doesn't work for MySQL 5.0, how to re-write it to make it work

DELETE FROM posts where id=(SELECT id FROM posts GROUP BY id  HAVING ( COUNT(id) > 1 ))

I want to delete columns that dont have unique id. I will add that most of the time its only one id(I tried the in syntax and it doesnt work as well).

Gudrunguelderrose

Answer 1

SELECT (sub)queries return result sets. So you need to use IN, not = in your WHERE clause.

Additionally, as shown in this answer you cannot modify the same table from a subquery within the same query. However, you can either SELECT then DELETE in separate queries, or nest another subquery and alias the inner subquery result (looks rather hacky, though):

DELETE FROM posts WHERE id IN (
    SELECT * FROM (
        SELECT id FROM posts GROUP BY id HAVING ( COUNT(id) > 1 )
    ) AS p
)

Or use joins as suggested by Mchl.

Digiovanni

I had a table with 150 duplicate keys. I executed the above query and it said "144 rows affected", but there where still duplicate keys. So I executed the query again and it says 5 rows affected, again: 1 row affected. Then all the duplicate keys where gone. Why is this? – Pronto

This is happening, because your are only deleting 1 entry out of each set of duplicates: SELECT id FROM posts GROUP BY id HAVING ( COUNT(id) > 1 ) – Georgetown

#1248 - Every derived table must have its own alias – Highwrought

@thang: Which is why I said to alias the inner subquery. – Digiovanni

No, still doesn't work. If you don't alias the inner subquery, it gives you a different error message. I ended up creating a temporary table, dumping everything to it, and then delete where id in that temp table. – Highwrought

@thang: Post a separate question. As it is there is nothing I can do if all you're going to give me is an error message and "doesn't work". – Digiovanni

@thang: I tried it four years ago and just tried it again now, and it works. – Digiovanni

I did the exact same thing. Did you use innodb? I wonder if it depends on the engine. – Highwrought

@thang: I tested it on both MyISAM and InnoDB and it works. What is your version of MySQL? I highly recommend posting a separate question so you can provide more details. – Digiovanni

let me dig into it some more. probably something stupid i am doing. if it still doesn't work, i'll post a new question. thanks! – Highwrought

Will you please explain what does the "As p" does? – Rivarivage

@Samarth it is an alias for the derived table (SELECT id FROM posts GROUP BY id HAVING ( COUNT(id) > 1 )) – Centeno

Answer 2

DELETE 
  p1
  FROM posts AS p1 
CROSS JOIN (
  SELECT ID FROM posts GROUP BY id HAVING COUNT(id) > 1
) AS p2
USING (id)

Foreland

This appears to work, but I'm confused by the syntax and can't find any resources elsewhere to explain it. CROSS JOIN apparently performs a cartesian join, so seems like this might do unnecessary work, or perform sub-optimally? Could anyone explain? – Tetracaine

It will do a cartesian product only if there's no USING clause. With USING the product is limited to pairs having same value in id column, so it is in fact very limited. – Foreland

Could you do the same thing with inner join? I.E. DELETE p1 FROM posts AS p1 INNER JOIN ( SELECT ID FROM posts GROUP BY id HAVING COUNT(id) > 1 ) AS p2 ON p2.ID=p1.ID – Carisa

@Andrew: Yes. Functionally these joins are exactly the same. – Foreland

@Tetracaine CROSS JOIN is the same as [INNER] JOIN in MySQL; it doesn't implement CROSS JOIN in the way that ANSI SQL describes it. SELECT ... FROM a JOIN b USING (x) is syntactic sugar for SELECT ... FROM a JOIN b ON b.x = a.x. – Rounding

Answer 3

you can use inner join :

DELETE 
    ps 
FROM 
    posts ps INNER JOIN 
         (SELECT 
           distinct id 
         FROM 
             posts 
         GROUP BY id  
      HAVING COUNT(id) > 1 ) dubids on dubids.id = ps.id  

Lynch

This is great because it works for queries with a composite PK – Shirleneshirley

When I tested in MySQL 5.7, this is more faster than BoltClock's answer. – Hydrogenolysis

This should be THE accepted answer, as it is not using subquery and in general is faster. Way faster. – Matriarch

Answer 4

If you want to delete all duplicates, but one out of each set of duplicates, this is one solution:

DELETE posts
FROM posts
LEFT JOIN (
    SELECT id
    FROM posts
    GROUP BY id
    HAVING COUNT(id) = 1

    UNION

    SELECT id
    FROM posts
    GROUP BY id
    HAVING COUNT(id) != 1
) AS duplicate USING (id)
WHERE duplicate.id IS NULL;

Georgetown