Help needed with a SELECT statement. How can I make it run faster?

Question

Hi,
not sure if my brain is just too tired but I can't seem to crack this problem today.
Here is my scenario.
I have 2 tables
TABLE1 (searchId INTEGER, routeId INTEGER);
TABLE2 (routeId INTEGER, cityId INTEGER);
There are indexes on all 4 columns.
(routeId on TABLE1 is a primary key).
In the data I am using, a given search has more than 500 routes, each route has between 10 and 300 cities among more than 4000 possible different cities.
Now, what I want to create is the list of route couple, within a certain search, that do not have a single city in common.
That list should populate a table with the following structure
TABLE3 (searchId INTEGER, routeId1 INTEGER, routeId2 INTEGER)
Here is the fastest select statement I have found so far.
SELECT :searchId, t1.routeId, t2.routeId FROM table1 t1, table1 t2
WHERE t1.searchId=:searchId AND t2.searchId=:searchId
AND t1.routeId>t2.routeId
AND NOT EXISTS (
SELECT cityId FROM table2
WHERE routeId=t1.routeId
INTERSECT
SELECT cityId FROM table2
WHERE routeId=t2.routeId);
But it still seem really slow to me.
Any suggestion for an improved version is welcome.
Thanks,
Martin.
Title was edited by:
user453358

Answer 1

I originaly posted this thread because I tought I was missing something "obvious" that would perform better that would make my SELECT statement perform better.
So I did not want to go as deep as using TKPROOF yet.
Here is the statistics I gets on my statement.
  1   SELECT t1.searchId,t1.routeId, t2.routeId id2
  2      FROM table1 t1, table1 t2
  3     WHERE t1.searchid=t2.searchid
  4      AND t1.searchId=91
  5      AND t1.routeId>t2.routeId
  6      AND NOT EXISTS (
  7             SELECT cityId FROM table2
  8              WHERE routeId=t1.routeId
  9             INTERSECT
10             SELECT cityId FROM table2
11*            WHERE routeId=t2.routeId)
SQL> /
43302 rows.
Tidsåtgång: 00:01:55.02
Körschema
   0      SELECT STATEMENT Optimizer=ALL_ROWS (Cost=9 Card=1 Bytes=14)         
   1    0   FILTER                                                             
   2    1     TABLE ACCESS (BY INDEX ROWID) OF 'table1' (TABLE) (Cost=1 Card=1 Bytes=7)                                                
   3    2       NESTED LOOPS (Cost=3 Card=1 Bytes=14)                          
   4    3         TABLE ACCESS (BY INDEX ROWID) OF 'table1' (TABLE) (Cost=2 Card=1 Bytes=7)
   5    4           INDEX (RANGE SCAN) OF 'table1_IDX1' (INDEX) (Cost=1 Card=1)                                                    
   6    3         INDEX (RANGE SCAN) OF 'table1_IDX1' (INDEX) (Cost=0 Card=1)                                                      
   7    1     INTERSECTION                                                     
   8    7       SORT (UNIQUE) (Cost=3 Card=108 Bytes=864)                      
   9    8         TABLE ACCESS (BY INDEX ROWID) OF 'table2' (TABLE) (Cost=2 Card=108 Bytes=864)                                 
  10    9           INDEX (RANGE SCAN) OF 'table2_IDX1' (INDEX) (Cost=1 Card=108)                                               
  11    7       SORT (UNIQUE) (Cost=3 Card=108 Bytes=864)                      
  12   11         TABLE ACCESS (BY INDEX ROWID) OF 'table2' (TABLE) (Cost=2 Card=108 Bytes=864)                                 
  13   12           INDEX (RANGE SCAN) OF 'table2_IDX1' (INDEX) (Cost=1 Card=108)                                               
Statistik
          1  recursive calls                                                   
          0  db block gets                                                     
    2872765  consistent gets                                                   
          0  physical reads                                                    
        812  redo size                                                         
     964172  bytes sent via SQL*Net to client                                  
      32245  bytes received via SQL*Net from client                            
       2888  SQL*Net roundtrips to/from client                                 
     860256  sorts (memory)                                                    
          0  sorts (disk)                                                      
      43302  rows processed  Looks like a big number of consistent gets! Any idea how to improve on that?
Martin.