Question

我要改进此查询：

    SELECT sub.SDDOCO
    FROM( 
      SELECT SDSHAN, SDDOCO, SDMCU, SDLITM, SDIR03, SDVR01
      FROM PRODDTA.F4211 

      UNION ALL

      SELECT SDSHAN, SDDOCO, SDMCU, SDLITM, SDIR03, SDVR01
      FROM PRODDTA.F42119 
    )sub 
    INNER JOIN PRODDTA.F0101 tab 
         ON sub.SDSHAN = tab.ABAN8
    WHERE 
        sub.SDMCU LIKE  CONCAT(CONCAT('%',nvl(:pMCU,SDMCU)),'%') 
    AND sub.SDLITM LIKE  CONCAT(CONCAT('%',nvl(:pLITM,SDLITM)),'%') 
    AND sub.SDIR03 LIKE  CONCAT(CONCAT('%',nvl(:pIR03,SDIR03)),'%')
    AND tab.ABALKY LIKE  CONCAT(CONCAT('%',nvl(:pALKY,ABALKY)),'%')
    AND sub.SDVR01 LIKE  CONCAT(CONCAT('%',nvl(:pVR01,SDVR01)),'%')

我已经以这种方式修改了WHERE条件：

    ((:pMCU IS NOT NULL AND sub.SDMCU LIKE '%' || :pMCU || '%') OR :pMCU IS NULL)
AND ((:pLITM IS NOT NULL AND sub.SDLITM LIKE '%' || :pLITM || '%') OR :pLITM IS NULL)
AND ((:pIR03 IS NOT NULL AND sub.SDIR03 LIKE '%' || :pIR03 || '%') OR :pIR03 IS NULL)
AND ((:pVR01 IS NOT NULL AND sub.SDVR01 LIKE '%' || :pVR01 || '%') OR :pVR01 IS NULL)
AND ((:pALKY IS NOT NULL AND tab.ABALKY LIKE '%' || :pALKY || '%') OR :pALKY IS NULL)

在this link，您可以阅读两个查询的执行计划。哪一个更好？有什么改进查询的建议吗？我需要有结果的所有记录

修改：哪里＆＃34; F4211＆＃34; table有71,758条记录，F42119有4,606,154条记录

编辑：和F0101表有634,187条记录。 Whith所有参数都设置为NULL，结果是在不合理的时间（几分钟）内有4,677,925条记录。没有where子句的查询在58秒内占用4,677,925条记录

修改：插入代码片段我超过了char的最大数量:( 有关发布执行计划的任何建议吗？

修改：这是执行计划：

"PLAN_TABLE_OUTPUT"
"Plan hash value: 3925457500"
" "
"-----------------------------------------------------------------------------------------------------------"
"| Id  | Operation                              | Name             | Rows  | Bytes | Cost (%CPU)| Time     |"
"-----------------------------------------------------------------------------------------------------------"
"|   0 | SELECT STATEMENT                       |                  |     2 |  1224 |   614K  (1)| 00:00:24 |"
"|   1 |  NESTED LOOPS                          |                  |     2 |  1224 |   614K  (1)| 00:00:24 |"
"|   2 |   VIEW                                 | index$_join$_004 | 31721 |  1455K| 11379   (1)| 00:00:01 |"
"|*  3 |    HASH JOIN                           |                  |       |       |            |          |"
"|   4 |     INDEX FAST FULL SCAN               | F0101_0          | 31721 |  1455K|  2414   (1)| 00:00:01 |"
"|*  5 |     INDEX FAST FULL SCAN               | F0101_3          | 31721 |  1455K|  8390   (1)| 00:00:01 |"
"|   6 |   VIEW                                 |                  |     1 |   565 |    19   (0)| 00:00:01 |"
"|   7 |    UNION ALL PUSHED PREDICATE          |                  |       |       |            |          |"
"|*  8 |     TABLE ACCESS BY INDEX ROWID BATCHED| F4211            |     1 |   417 |     9   (0)| 00:00:01 |"
"|*  9 |      INDEX RANGE SCAN                  | F4211_4          |     6 |       |     3   (0)| 00:00:01 |"
"|* 10 |     TABLE ACCESS BY INDEX ROWID BATCHED| F42119           |     1 |   422 |    10   (0)| 00:00:01 |"
"|* 11 |      INDEX RANGE SCAN                  | F42119_9         |    10 |       |     3   (0)| 00:00:01 |"
"-----------------------------------------------------------------------------------------------------------"
" "
"Predicate Information (identified by operation id):"
"---------------------------------------------------"
" "
"   3 - access(ROWID=ROWID)"
"   5 - filter(""GGG"".""ABALKY"" LIKE U'%'||NVL(:PALKY,""GGG"".""ABALKY"")||U'%')"
"   8 - filter(""SDMCU"" LIKE U'%'||NVL(:PMCU,""SDMCU"")||U'%' AND ""SDLITM"" LIKE "
"              U'%'||NVL(:PLITM,""SDLITM"")||U'%' AND ""SDIR03"" LIKE U'%'||NVL(:PIR03,""SDIR03"")||U'%' AND ""SDVR01"" "
"              LIKE U'%'||NVL(:PVR01,""SDVR01"")||U'%')"
"   9 - access(""SDSHAN""=""GGG"".""ABAN8"")"
"  10 - filter(""SDMCU"" LIKE U'%'||NVL(:PMCU,""SDMCU"")||U'%' AND ""SDLITM"" LIKE "
"              U'%'||NVL(:PLITM,""SDLITM"")||U'%' AND ""SDIR03"" LIKE U'%'||NVL(:PIR03,""SDIR03"")||U'%' AND ""SDVR01"" "
"              LIKE U'%'||NVL(:PVR01,""SDVR01"")||U'%')"
"  11 - access(""SDSHAN""=""GGG"".""ABAN8"")"

Answer 1

1 Oracle具有提高查询性能的良好工具，例如DBMS_SQLTUNE https://docs.oracle.com/database/121/TGSQL/tgsql_sqltune.htm#TGSQL586 在开始时，您可以创建调整任务

declare
  l_task_name varchar2(4000);
begin
 l_task_name := DBMS_SQLTUNE.CREATE_TUNING_TASK(
      sql_text => q'!SELECT sub.SDDOCO
FROM( 
  SELECT SDSHAN, SDDOCO, SDMCU, SDLITM, SDIR03, SDVR01
  FROM PRODDTA.F4211 
  UNION ALL
  SELECT SDSHAN, SDDOCO, SDMCU, SDLITM, SDIR03, SDVR01
  FROM PRODDTA.F42119 
)sub 
INNER JOIN PRODDTA.F0101 tab 
     ON sub.SDSHAN = tab.ABAN8
WHERE 
    sub.SDMCU LIKE  CONCAT(CONCAT('%',nvl(:pMCU,SDMCU)),'%') 
AND sub.SDLITM LIKE  CONCAT(CONCAT('%',nvl(:pLITM,SDLITM)),'%') 
AND sub.SDIR03 LIKE  CONCAT(CONCAT('%',nvl(:pIR03,SDIR03)),'%')
AND tab.ABALKY LIKE  CONCAT(CONCAT('%',nvl(:pALKY,ABALKY)),'%')
AND sub.SDVR01 LIKE  CONCAT(CONCAT('%',nvl(:pVR01,SDVR01)),'%')!'
, bind_list   => sql_binds( anydata.ConvertVarchar2('ABC'), 
            anydata.ConvertVarchar2('BCD'),
            anydata.ConvertVarchar2('CDE'),
            anydata.ConvertVarchar2('DEF'),
            anydata.ConvertVarchar2('EFG'))
, task_name   => 'TEST_TUNING_SQL_SO' 
, description => 'Task to tune a query from StackOverflow');
end;
/

执行后

begin
  DBMS_SQLTUNE.EXECUTE_TUNING_TASK (
   task_name => 'TEST_TUNING_SQL_SO');
end;
/

检索报告并执行建议

select DBMS_SQLTUNE.REPORT_TUNING_TASK(
   task_name => 'TEST_TUNING_SQL_SO') from dual;

2如何公布计划

您可以以文本格式（而非HTML格式）生成它。并尝试仅添加包含谓词信息的计划表。如果不需要，您可以排除任何其他信息。

Oracle Optimization：哪个执行计划更好？

1 个答案: