Do I need another index?

Question

Hi!
I have a table:CREATE TABLE G5DZE (
    "ID"     VARCHAR2(100 BYTE),
    "STATUS" NUMBER(2,0) DEFAULT 1,
    "G5IDD"  VARCHAR2(200 BYTE),
    "G5IDR"  VARCHAR2(200 BYTE),
    "G5NOS"  VARCHAR2(200 BYTE),
    "G5WRT"  NUMBER(32,0),
    "G5DWR"  DATE,
    "G5PEW"  NUMBER,
    "G5DZP"  NUMBER(1,0),
    "G5RZN"  VARCHAR2(200 BYTE),
    "G5DWW"  DATE,
    "G5DTW"  DATE,
    "G5DTU"  DATE,
    "ID_G5JDR_RJDR" VARCHAR2(100 BYTE),
    "IDR"           VARCHAR2(100 BYTE),
    "PLS_ID"        NUMBER
  );and similar table with ROG_TEMP_G5DZE name. There are indexes:CREATE UNIQUE INDEX "G5DZE_PK" ON "G5DZE" ("PLS_ID", "ID");
CREATE UNIQUE INDEX "T_G5DZE_PK" ON "ROG_TEMP_G5DZE" ("PLS_ID", "ID");
CREATE INDEX "G5DZE_JDR_FK" ON "G5DZE" ("PLS_ID", "ID_G5JDR_RJDR");
CREATE INDEX "T_G5DZE_JDR_FK" ON "ROG_TEMP_G5DZE" ("PLS_ID", "ID_G5JDR_RJDR");PK is the pair (PLS_ID, ID).
Table G5DZE contains current data and ROG_TEMP_G5DZE contains previous data.
I need to find what records were added to G5DZE and what records were in ROG_TEMP_G5DZE but now they aren't in G5DZE.
SELECT null, null, null,
              g5.id    AS new_ID,
              g5.G5DTU AS new_DTU,
              g5.G5DTW AS new_DTW
         FROM G5DZE g5
        WHERE g5.STATUS = 1
          AND NOT EXISTS (SELECT 1
                            FROM ROG_TEMP_G5DZE
                           WHERE G5DTU = g5.G5DTU AND -- ***
                                 ID = g5.ID AND
                                 status=1
                             AND pls_id = 530)
        UNION ALL
       SELECT rt.id AS old_ID,
              rt.G5DTU AS old_DTU,
              rt.G5DTW AS old_DTW,
              null, null, null
         FROM ROG_TEMP_G5DZE rt
        WHERE rt.STATUS = 1
          AND pls_id = 530
          AND NOT EXISTS (SELECT 1
                            FROM G5DZE
                           WHERE G5DTU = rt.G5DTU AND
                                 rt.ID = ID
                             AND status=1);Now I have about 60k rows in both tables. The first SELECT (before the UNION ALL) first fast when I comment the WHERE condition marked with "***".
The explain plan for first SELECT is:PLAN_TABLE_OUTPUT
Plan hash value: 3325319594
| Id  | Operation                    | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |                |     1 |   170 |     2   (0)| 00:00:01 |
|   1 |  NESTED LOOPS ANTI           |                |     1 |   170 |     2   (0)| 00:00:01 |
|*  2 |   TABLE ACCESS FULL          | G5DZE          |     1 |    83 |     2   (0)| 00:00:01 |
|*  3 |   TABLE ACCESS BY INDEX ROWID| ROG_TEMP_G5DZE |     1 |    87 |     0   (0)| 00:00:01 |
|*  4 |    INDEX RANGE SCAN          | T_G5DZE_JDR_FK |     1 |       |     0   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - filter(G5.STATUS=1)
   3 - filter(STATUS=1 AND G5DTU=G5.G5DTU AND ID=G5.ID)
   4 - access(PLS_ID=530)and when I comment the marked WHERE condition:PLAN_TABLE_OUTPUT
Plan hash value: 1041062869
| Id  | Operation                     | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT              |                |     1 |   161 |     2  (50)| 00:00:01 |
|   1 |  MERGE JOIN ANTI              |                |     1 |   161 |     2  (50)| 00:00:01 |
|*  2 |   TABLE ACCESS BY INDEX ROWID | G5DZE          |     1 |    83 |     0   (0)| 00:00:01 |
|   3 |    INDEX FULL SCAN            | G5DZE_PK       |     1 |       |     0   (0)| 00:00:01 |
|*  4 |   SORT UNIQUE                 |                |     1 |    78 |     2  (50)| 00:00:01 |
|*  5 |    TABLE ACCESS BY INDEX ROWID| ROG_TEMP_G5DZE |     1 |    78 |     1   (0)| 00:00:01 |
|*  6 |     INDEX RANGE SCAN          | T_G5DZE_JDR_FK |     1 |       |     1   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - filter(G5.STATUS=1)
   4 - access(ID=G5.ID)
       filter(ID=G5.ID)
   5 - filter(STATUS=1)
   6 - access(PLS_ID=530)Do I need to create another index like:CREATE UNIQUE INDEX G5DZE_3I ON G5DZE (PLS_ID, ID, G5DTU);to make the query run faster or is there better sollution? The index would have rather redundant data.

Answer 1

Karthick_Arp wrote:
Did you try using MINUS operator?No, I didn't. I can't imagine how may the MINUS be used here. May I ask for an example?
When I add index:create index G5DZE_I4 on G5DZE (status, ID, g5dtu,g5dtw);the plan for statementSELECT null, null, null,
        g5.id    AS new_ID,
        g5.G5DTU AS new_DTU,
        g5.G5DTW AS new_DTW
   FROM G5DZE g5
  WHERE g5.STATUS = 1
    AND NOT EXISTS (SELECT 1
                      FROM ROG_TEMP_G5DZE
                     WHERE G5DTU = g5.G5DTU AND
                           ID = g5.ID AND
                           status=1
                       AND pls_id = 530);isPLAN_TABLE_OUTPUT
Plan hash value: 3290283461
| Id  | Operation                    | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |                |     1 |   170 |     1   (0)| 00:00:01 |
|   1 |  NESTED LOOPS ANTI           |                |     1 |   170 |     1   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN           | G5DZE_I4       |     1 |    83 |     1   (0)| 00:00:01 |
|*  3 |   TABLE ACCESS BY INDEX ROWID| ROG_TEMP_G5DZE |     1 |    87 |     0   (0)| 00:00:01 |
|*  4 |    INDEX RANGE SCAN          | T_G5DZE_JDR_FK |     1 |       |     0   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   2 - access(G5.STATUS=1)
   3 - filter(STATUS=1 AND G5DTU=G5.G5DTU AND ID=G5.ID)
   4 - access(PLS_ID=530)but the query still runs VEEEERRRRY slow :(
Statistics are not gathered yet, however. I tried to do that but DBA need fix some datablocks first.