Thread: Finding (and deleting) dupes in relation table

Finding (and deleting) dupes in relation table

From
CSN
Date:
I have a table that relates id's of two other tables:

table1id, table2id

Dupes have found their way into it (create unique
index across both fields fails). Is there a quick and
easy way to find and delete the dupes (there are tens
of thousands of records)?

Thanks,
CSN



__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

Re: Finding (and deleting) dupes in relation table

From
CSN
Date:
Nevermind, figured it out:

select distinct on (table1id, table2id) * into temp
from table3;
delete from table3;
insert into table3 select * from temp;



--- CSN <cool_screen_name90001@yahoo.com> wrote:

> I have a table that relates id's of two other
> tables:
>
> table1id, table2id
>
> Dupes have found their way into it (create unique
> index across both fields fails). Is there a quick
> and
> easy way to find and delete the dupes (there are
> tens
> of thousands of records)?
>
> Thanks,
> CSN
>
>
>
> __________________________________
> Yahoo! Mail - PC Magazine Editors' Choice 2005
> http://mail.yahoo.com
>




__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

Re: Finding (and deleting) dupes in relation table

From
"Markus Wollny"
Date:
Hello!

CSN [cool_screen_name90001@yahoo.com] wrote:
> I have a table that relates id's of two other tables:
>
> table1id, table2id
>
> Dupes have found their way into it (create unique index
> across both fields fails). Is there a quick and easy way to
> find and delete the dupes (there are tens of thousands of records)?
>
> Thanks,
> CSN

If your table was created WITH OIDS you could identify the duplicates
thus:

select a.table1id
     , a.table12d
     , max(a.oid) as maxoid
     , count(a.oid) as coid
  from schema.mytable a,
       schema.mytable b
 where a.table1id = b.table1id
   and a.table2id=b.table2id
   and a.oid <> b.oid
 group by a.table1id, a.table2id
 order by a.table1id;

If you wish to delete surplus rows, you might do the following:

delete from schema.mytable where oid in (
     select maxoid from (
             select a.table1id, a.table12d, max(a.oid) as
maxoid, count(a.oid) as coid
                 from schema.mytable a,
                      schema.mytable b
               where a.table1id = b.table1id
                and a.table2id=b.table2id
                and a.oid <> b.oid
              group by a.table1id, a.table2id
              order by a.table1id ) as foo
       where coid >1 );

This will delete the oldest tuple of a duplicate set of rows; if there
are more than two tuples in a set, you'll want to execute this a couple
of times until there's no duplicate left, as the delete will only reduce
a set by one tuple at a time. I'd also recommend to apply a PRIMARY KEY
constraint afterwards instead of just a unique index - this will prevent
NULL-entries as well as creating the desired unique index - and I think
it's good practice to have a primary key on about every table there is,
except when it's just a junk data table like a logging table where
content is regularly evaluated and discarded.

Kind regards

   Markus