Query running out of temp

Question

Hi,
on my 9.2.0.8 I've got query like this
select  DISTINCT T.ID_KONTR as ID_KONTR,
                 T.PESEL as PESEL,
                 T.NAZWISKO as NAZWISKO,
                 T.IMIE as IMIE,
                 T.DATA_REJ as DATA_REJ 
from T,
    (SELECT DISTINCT MAX(T.ID_KONTR) as ID_KONTR  from T  ,
        (select  DISTINCT T.ID_KONTR as ID_KONTR,T.PESEL as PESEL,
                          T.NAZWISKO as NAZWISKO,
                          T.IMIE as IMIE,
                          T.DATA_REJ as DATA_REJ  
                from T
                where
                    (T.OK =1)
                   and (T.PESEL is not null)
                   and  T.data_rej between to_date('2011-10-21' , 'YYYY-MM-DD') and to_date('2011-10-25','YYYY-MM-DD')
         ) aaa 
    where
        (T.OK =1)
        and (T.PESEL is not null) 
        and T.pesel = aaa.pesel
        group by  T.PESEL 
        having count(*) >1
   ) bbb 
where T.ID_KONTR=bbb.ID_KONTR order by T.DATA_REJ desc;
with plan
lan
SELECT STATEMENT  RULE                                                            
     15 SORT UNIQUE                                                         
          14 NESTED LOOPS                                                    
               11 VIEW INSTALL.                                              
                    10 SORT UNIQUE                                          
                         9 FILTER                                     
                              8 SORT GROUP BY                                
                                   7 VIEW SYS.                          
                                        6 SORT UNIQUE                      
                                             5 TABLE ACCESS BY INDEX ROWID T                
                                                  4 NESTED LOOPS            
                                                       2 TABLE ACCESS BY INDEX ROWID T      
                                                            1 INDEX RANGE SCAN NON-UNIQUE TK_DATA_REJ
                                                       3 INDEX RANGE SCAN NON-UNIQUE TK_PESEL      
               13 TABLE ACCESS BY INDEX ROWID T                                              
                    12 INDEX UNIQUE SCAN UNIQUE TK_PK                                         
          Unfortunately this query fills up temp space (because of huge sorts) , any ideas how that query can be rewritten to
decrease temp usage ?
Regards
GregG

Answer 1

Hello
Just to follow on from what Don is saying, I think you can start to rewrite the query like so
SELECT
      id_kontr,
      pesel,
      nazwisko,
      imie,
      data_rej     
FROM
    (   SELECT
              t.id_kontr AS id_kontr,
              t.pesel    AS pesel,
              t.nazwisko AS nazwisko,
              t.imie     AS imie,
              t.data_rej AS data_rej
              COUNT(CASE
                         WHEN t.data_rej BETWEEN to_date('2011-10-21', 'YYYY-MM-DD')
                                             AND to_date('2011-10-25', 'YYYY-MM-DD')
                         THEN 1
                    END
                   ) OVER(PARTITION BY t.pesel) p_count,
              MAX(t.id_kontr) OVER (PARTITION BY t.pesel) max_id_kontr
        FROM
             t
        WHERE
             t.ok = 1
        AND
             t.pesel IS NOT NULL
WHERE
     p_count > 1
AND
   id_kontr = max_id_kontr
ORDER BY
      data_rej DESCWithout understanding the need for the distinct, I can't say whether it does exactly what you need but it should be a start and it also means you're only accessing "t" once.
HTH
David