Help with optimize analitics query .

Question

HI,
I've got Oracle 10.2.0.3 query like :
select * from
     select rank() over (partition by t1.customer_key order by
     func(t1.klt_dlugosc,t1.klt_szerokosc,t2.gemfo_dlugosc,t2.gemfo_szerokosc),rownum) as id,
     t1.customer_key,
     func(t1.klt_dlugosc,t1.klt_szerokosc,t2.gemfo_dlugosc,t2.gemfo_szerokosc) odleg_gemfo,
     t2.gemfo_id             
     from tmp_22_geo_0309 t1,
(select * from adresy_gemfo_geo t2)) where id<3;
PLAN_TABLE_OUTPUT
Plan hash value: 187241240
| Id  | Operation                | Name                           | Rows  | Bytes |TempSpc| Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT         |                                |    10M|   486M|       |   109K  (2)| 00:25:28 |
|*  1 |  VIEW                    |                                |    10M|   486M|       |   109K  (2)| 00:25:28 |
|*  2 |   WINDOW SORT PUSHED RANK|                                |    10M|   630M|  1455M|   109K  (2)| 00:25:28 |
|   3 |    COUNT                 |                                |       |       |       |            |          |
|   4 |     MERGE JOIN CARTESIAN |                                |    10M|   630M|       |  9720   (3)| 00:02:17 |
|   5 |      TABLE ACCESS FULL   | ADRESY_GEMFO_GEO               |   151 |  5738 |       |     3   (0)| 00:00:01 |
|   6 |      BUFFER SORT         |                                | 66293 |  1812K|       |   109K  (2)| 00:25:28 |
|   7 |       TABLE ACCESS FULL  | TMP_22_GEO_0309                       | 66293 |  1812K|       |    64   (2)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter("ID"<3)
   2 - filter(RANK() OVER ( PARTITION BY "CUSTOMER_KEY" ORDER BY
              func("KLT_DLUGOSC","KLT_SZEROKOSC","ADRESY_GEMFO_GEO"."GEMFO_DLUGOSC","ADRESY_GEMFO_GEO"."
              GEMFO_SZEROKOSC"),ROWNUM)<3)
Note
   - dynamic sampling used for this statement
   is burning CPU , an d runs like 2h
t1 is 66293 rows
and t2 is 151 rows
is there any way to rewrite this ?
function func is doing some geolocation related math like sin/cos abs .
Regards.
Greg

Answer 1

Yes, thats expected (cartesian join), tables are small .
Function is doing geolocation calculations like:
CREATE OR REPLACE function func (dlugosc_geo_1 number,
                                      szerokosc_geo_1 number,
                                      dlugosc_geo_2 number,
                                      szerokosc_geo_2 number) return number
is
  a_1 number;
  c_1 number;
  d_1 number;
  metry number := 1000;
  r_ziemi number := 6371;
  pi number := 3.1415926535897932384626433832795028841972;
  dLat number;
  dLon number;
  szer_1 number;
  szer_2 number;
begin
  dLat := abs(szerokosc_geo_2-szerokosc_geo_1)*pi/180;
  dLon := abs(dlugosc_geo_2-dlugosc_geo_1)*pi/180;
  szer_1 := szerokosc_geo_1 * pi/180;
  szer_2 := szerokosc_geo_2 * pi/180;
  a_1 := sin(dLat/2) * sin(dLat/2) +
         cos(szer_1) * cos(szer_2) *
         sin(dLon/2) * sin(dLon/2);
  c_1 := 2 * atan2(sqrt(a_1), sqrt(1-a_1));
  d_1 := round(metry * r_ziemi * c_1);
RETURN d_1;
end;
/