Re: How to idenity duplicate rows

Peter Eisentraut wrote:

>David Inglis wrote:
>
>
>>Can anybody assist with this problem I have a table that has some
>>duplicated rows of data,  I want to place a unique constraint on  the
>>columns userid and procno to eliminate this problem in the future but
>>how do I identify and get rid of the existing duplication.
>>
>>
>
>To find them, something like
>
>SELECT a, b, c FROM table GROUP BY a, b, c HAVING count(*) > 1;
>
>comes to mind, where you have to list all columns of the table in place
>of a, b, c.
>
>As for deleting all but one row in a duplicated group, you're going to
>have to get at them by the oid or ctid columns perhaps.
>
>The other idea is to run CREATE TABLE newtable AS SELECT DISTINCT * FROM
>oldtable;.
>
>
This doesn't bring over to the new table any foreign key relationships
or triggers.

Another approach (if you don't have OID's) is to create uniqueness by
appending a column to the table, populating it with sequential integers.
Then you proceed as otherwise suggested above by using aggregation to
identify the duplicated rows.