Equivalent of Oracle’s RowID in MySQL

Question

is there an equivalent of oracle's rowid in mysql?

delete from my_table where rowid not in (select max(rowid) from my_table group by field1,field2)

I want to make a mysql equivalent of this query!!!

What i'm trying to do is, : The my_table has no primary key.. i'm trying to delete the duplicate values and impose a primary key (composite of field1, field2)..!!

Answer 1

mysql> set @row_num = 0;

First set rowId or row num the use it as following:

mysql>  SELECT @row_num := @row_num + 1 as row_number,id,name,salary FROM employee
ORDER BY salary;

Answer 2

In MySql you usually use session variables to achive the functionality:

SELECT @rowid:=@rowid+1 as rowid
FROM table1, (SELECT @rowid:=0) as init
ORDER BY sorter_field

But you can not make sorts on the table you are trying to delete from in subqueries.

UPD: that is you will need to create a temp table, insert the ranging subquery to the temp table and delete from the original table by joining with the temporary table (you will need some unique row identifier):

CREATE TEMPORARY TABLE duplicates ...

INSERT INTO duplicates (rowid, field1, field2, some_row_uid)
SELECT
  @rowid:=IF(@f1=field1 AND @f2=field2, @rowid+1, 0) as rowid,
  @f1:=field1 as field1,
  @f2:=field2 as field2,
  some_row_uid
FROM testruns t, (SELECT @rowid:=NULL, @f1:=NULL, @f2:=NULL) as init
ORDER BY field1, field2 DESC;

DELETE FROM my_table USING my_table JOIN duplicates
  ON my_table.some_row_uid = duplicates.some_row_uid AND duplicates.rowid > 0

Since that is one time operation, this should not bring too much overhead.

Answer 3

you can avoid the temp table using another derived table:

DELETE FROM my_table USING my_table JOIN (
    SELECT @rowid:=IF(@f1=field1 AND @f2=field2, @rowid+1, 0) as rowid,
        @f1:=field1 as field1,
        @f2:=field2 as field2,
       some_row_uid
   FROM testruns t, (SELECT @rowid:=NULL, @f1:=NULL, @f2:=NULL) as init
   ORDER BY field1, field2 DESC) as duplicates
ON my_table.some_row_uid = duplicates.some_row_uid AND duplicates.rowid > 0

Answer 4

Maybe, I am misreading the question but your query (even in Oracle) doesn't accomplish your desired goal:

delete from my_table where rowid not in (select max(rowid) from 
my_table group by field1,field2)

MySQL equivalent is

SELECT @rowid:=max(rowid) from my_table;
DELETE FROP my_table where rowid != @rowid;

This will wipe out all rows except for last one.

To perform one time cleanup (removing duplicate records) of your data you can do this:

CREATE TABLE my_table2 SELECT distinct f1, f2, f3, etc from my_table;
DROP TABLE my_table;
ALTER TABLE my_table2 RENAME my_table;

Then add whatever columns & keys necessary by ALTER TABLE. Above code might require to drop any foreign keys you might have.