• sql优化案例(filter)


    explain plan for  select  '33101' org_no, nvl(count(*), 0) outNum
       from CPSS.k_id_staff_tqjl_v a
      where a.pda_status_code = '002'
        and a.gds_no in (SELECT b.dim_value
                           FROM CPSS.k_ic_dim_value b
                          WHERE b.level_id = '5'
                            AND b.dim_code = 'org'
                          START WITH b.dim_value = '33101'
                         CONNECT BY PRIOR b.dim_value = b.p_dim_id)
        and not exists (
            select 1
               from CPSS.K_PDA_HEARTBEAT  k
              where date_str = '20200714'
                and GMT_CREATE >= sysdate - 15 / 24 / 60
                and k.user_no = a.auth_user_no
                and k.org_no = a.gds_no
                );
    
    
    Plan hash value: 1785689430                                                                                                                  
                                                                                                                                                 
    ---------------------------------------------------------------------------------------------------------------------------------------------
    | Id  | Operation                                        | Name                     | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
    ---------------------------------------------------------------------------------------------------------------------------------------------
    |   0 | SELECT STATEMENT                                 |                          |     1 |    94 |    39   (3)| 00:00:01 |       |       |
    |   1 |  SORT AGGREGATE                                  |                          |     1 |    94 |            |          |       |       |
    |*  2 |   FILTER                                         |                          |       |       |            |          |       |       |
    |*  3 |    HASH JOIN                                     |                          |     2 |   188 |    37   (3)| 00:00:01 |       |       |
    |   4 |     NESTED LOOPS                                 |                          |    53 |  3445 |    36   (3)| 00:00:01 |       |       |
    |   5 |      NESTED LOOPS                                |                          |   294 |  3445 |    36   (3)| 00:00:01 |       |       |
    |   6 |       VIEW                                       | VW_NSO_1                 |     6 |   204 |    14   (8)| 00:00:01 |       |       |
    |   7 |        HASH UNIQUE                               |                          |     6 |   618 |    14   (8)| 00:00:01 |       |       |
    |*  8 |         FILTER                                   |                          |       |       |            |          |       |       |
    |*  9 |          CONNECT BY NO FILTERING WITH SW (UNIQUE)|                          |       |       |            |          |       |       |
    |  10 |           TABLE ACCESS FULL                      | K_IC_DIM_VALUE           |  2579 |    98K|    13   (0)| 00:00:01 |       |       |
    |* 11 |       INDEX RANGE SCAN                           | IDX_K_ID_STAFF_N1        |    49 |       |     1   (0)| 00:00:01 |       |       |
    |* 12 |      TABLE ACCESS BY INDEX ROWID                 | K_ID_STAFF               |     9 |   279 |     4   (0)| 00:00:01 |       |       |
    |  13 |     TABLE ACCESS BY INDEX ROWID                  | K_IC_DIM_VALUE           |    96 |  2784 |     1   (0)| 00:00:01 |       |       |
    |* 14 |      INDEX RANGE SCAN                            | IDX_K_IC_DIM_VALUE_N1    |    96 |       |     1   (0)| 00:00:01 |       |       |
    |  15 |    PARTITION RANGE SINGLE                        |                          |     1 |    37 |     1   (0)| 00:00:01 |     8 |     8 |
    |* 16 |     TABLE ACCESS BY LOCAL INDEX ROWID            | K_PDA_HEARTBEAT          |     1 |    37 |     1   (0)| 00:00:01 |     8 |     8 |
    |* 17 |      INDEX RANGE SCAN                            | IDX_GMT_CREATE_USER_ORG2 |     1 |       |     1   (0)| 00:00:01 |     8 |     8 |
    |* 18 |    COUNT STOPKEY                                 |                          |       |       |            |          |       |       |
    |  19 |     TABLE ACCESS BY INDEX ROWID                  | P_PDA_DEVICE             |     2 |    20 |     1   (0)| 00:00:01 |       |       |
    |* 20 |      INDEX RANGE SCAN                            | IDX_P_PDA_DEVICE_N1      |     1 |       |     1   (0)| 00:00:01 |       |       |
    ---------------------------------------------------------------------------------------------------------------------------------------------
                                                                                                                                                 
    Predicate Information (identified by operation id):                                                                                          
    ---------------------------------------------------                                                                                          
                                                                                                                                                 
       2 - filter( NOT EXISTS (SELECT 0 FROM "CPSS"."K_PDA_HEARTBEAT" "K" WHERE "GMT_CREATE" IS NOT NULL AND                                     
                  "GMT_CREATE">=SYSDATE@!-.0104166666666666666666666666666666666667 AND "DATE_STR"='20200714' AND "K"."USER_NO"=:B1 AND          
                  "K"."ORG_NO"=:B2) AND  (SELECT "D"."STATUS_CODE" FROM CPSS."P_PDA_DEVICE" "D" WHERE ROWNUM=1 AND "D"."FETCHER_NO"=:B3)='002')  
       3 - access("S"."ORG_NO"="D"."DIM_VALUE")                                                                                                  
       8 - filter("B"."LEVEL_ID"=5 AND "B"."DIM_CODE"='org')                                                                                     
       9 - access("B"."P_DIM_ID"=PRIOR "B"."DIM_VALUE")                                                                                          
           filter("B"."DIM_VALUE"='33101')                                                                                                       
      11 - access("S"."ORG_NO"="DIM_VALUE")                                                                                                      
      12 - filter("S"."POST"='17' AND "S"."WORK_MODE"='01' AND "S"."STAFF_STATE"='01' AND "S"."DATA_OPER_TYPE"<>'D')                             
      14 - access("D"."DIM_CODE"='org' AND "D"."LEVEL_ID"=5)                                                                                     
      16 - filter("DATE_STR"='20200714')                                                                                                         
      17 - access("GMT_CREATE">=SYSDATE@!-.0104166666666666666666666666666666666667 AND "K"."USER_NO"=:B1 AND "K"."ORG_NO"=:B2 AND               
                  "GMT_CREATE" IS NOT NULL)                                                                                                      
           filter("K"."USER_NO"=:B1 AND "K"."ORG_NO"=:B2)                                                                                        
      18 - filter(ROWNUM=1)                                                                                                                      
      20 - access("D"."FETCHER_NO"=:B1)   
      
    explain plan for select '33101' org_no, nvl(count(*), 0) outNum from CPSS.k_id_staff_tqjl_v a where a.pda_status_code = '002' and a.gds_no in (SELECT b.dim_value FROM CPSS.k_ic_dim_value b WHERE b.level_id = '5' AND b.dim_code = 'org' START WITH b.dim_value = '33101' CONNECT BY PRIOR b.dim_value = b.p_dim_id) and (auth_user_no,gds_no) not in (select/*+qb_name(zz)*/ user_no,org_no from CPSS.K_PDA_HEARTBEAT k where date_str = '20200714' and GMT_CREATE >= sysdate - 15 / 24 / 60); Plan hash value: 3515795332 ----------------------------------------------------------------------------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop | ----------------------------------------------------------------------------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 131 | 44 (5)| 00:00:01 | | | | 1 | SORT AGGREGATE | | 1 | 131 | | | | | |* 2 | FILTER | | | | | | | | |* 3 | HASH JOIN | | 9 | 1179 | 39 (6)| 00:00:01 | | | | 4 | MERGE JOIN ANTI NA | | 53 | 5406 | 38 (6)| 00:00:01 | | | | 5 | SORT JOIN | | 53 | 3445 | 36 (3)| 00:00:01 | | | | 6 | NESTED LOOPS | | 53 | 3445 | 36 (3)| 00:00:01 | | | | 7 | NESTED LOOPS | | 294 | 3445 | 36 (3)| 00:00:01 | | | | 8 | VIEW | VW_NSO_1 | 6 | 204 | 14 (8)| 00:00:01 | | | | 9 | HASH UNIQUE | | 6 | 618 | 14 (8)| 00:00:01 | | | |* 10 | FILTER | | | | | | | | |* 11 | CONNECT BY NO FILTERING WITH SW (UNIQUE)| | | | | | | | | 12 | TABLE ACCESS FULL | K_IC_DIM_VALUE | 2579 | 98K| 13 (0)| 00:00:01 | | | |* 13 | INDEX RANGE SCAN | IDX_K_ID_STAFF_N1 | 49 | | 1 (0)| 00:00:01 | | | |* 14 | TABLE ACCESS BY INDEX ROWID | K_ID_STAFF | 9 | 279 | 4 (0)| 00:00:01 | | | |* 15 | SORT UNIQUE | | 2 | 74 | 3 (34)| 00:00:01 | | | | 16 | PARTITION RANGE SINGLE | | 2 | 74 | 2 (0)| 00:00:01 | 8 | 8 | |* 17 | TABLE ACCESS BY LOCAL INDEX ROWID | K_PDA_HEARTBEAT | 2 | 74 | 2 (0)| 00:00:01 | 8 | 8 | |* 18 | INDEX RANGE SCAN | IDX_GMT_CREATE_USER_ORG2 | 7 | | 1 (0)| 00:00:01 | 8 | 8 | | 19 | TABLE ACCESS BY INDEX ROWID | K_IC_DIM_VALUE | 96 | 2784 | 1 (0)| 00:00:01 | | | |* 20 | INDEX RANGE SCAN | IDX_K_IC_DIM_VALUE_N1 | 96 | | 1 (0)| 00:00:01 | | | |* 21 | COUNT STOPKEY | | | | | | | | | 22 | TABLE ACCESS BY INDEX ROWID | P_PDA_DEVICE | 2 | 20 | 1 (0)| 00:00:01 | | | |* 23 | INDEX RANGE SCAN | IDX_P_PDA_DEVICE_N1 | 1 | | 1 (0)| 00:00:01 | | | ----------------------------------------------------------------------------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 2 - filter( (SELECT "D"."STATUS_CODE" FROM CPSS."P_PDA_DEVICE" "D" WHERE ROWNUM=1 AND "D"."FETCHER_NO"=:B1)='002') 3 - access("S"."ORG_NO"="D"."DIM_VALUE") 10 - filter("B"."LEVEL_ID"=5 AND "B"."DIM_CODE"='org') 11 - access("B"."P_DIM_ID"=PRIOR "B"."DIM_VALUE") filter("B"."DIM_VALUE"='33101') 13 - access("S"."ORG_NO"="DIM_VALUE") 14 - filter("S"."POST"='17' AND "S"."WORK_MODE"='01' AND "S"."STAFF_STATE"='01' AND "S"."DATA_OPER_TYPE"<>'D') 15 - access(INTERNAL_FUNCTION("S"."ORG_NO")=INTERNAL_FUNCTION("ORG_NO") AND INTERNAL_FUNCTION("S"."AUTH_USER_NO")=INTERNAL_FUNCTION("USER_NO")) filter(INTERNAL_FUNCTION("S"."ORG_NO")=INTERNAL_FUNCTION("ORG_NO") AND INTERNAL_FUNCTION("S"."AUTH_USER_NO")=INTERNAL_FUNCTION("USER_NO")) 17 - filter("DATE_STR"='20200714') 18 - access("GMT_CREATE">=SYSDATE@!-.0104166666666666666666666666666666666667 AND "GMT_CREATE" IS NOT NULL) 20 - access("D"."DIM_CODE"='org' AND "D"."LEVEL_ID"=5) 21 - filter(ROWNUM=1) 23 - access("D"."FETCHER_NO"=:B1)

      

    实际生产环境中常常能见到执行计划中出现filter

    filter有哪些坏处:

    1.出现后驱动表固定,优化器不能选择合适的驱动表

    2.filter的表关联方式类似与nesloop,对于驱动表数据集比较大的情况下,不适合用该关联方式

    上述sql

    select                                                                                                                                          
      '33101' org_no, nvl(count(*), 0) outNum                                  
       from CPSS.k_id_staff_tqjl_v a                                           
      where a.pda_status_code = '002'                                          
        and a.gds_no in (SELECT b.dim_value                                    
                           FROM CPSS.k_ic_dim_value b                          
                          WHERE b.level_id = '5'                               
                            AND b.dim_code = 'org'                             
                          START WITH b.dim_value = '33101'                     
                         CONNECT BY PRIOR b.dim_value = b.p_dim_id) 
    改sql执行效率ok,返回结果集有1万多(毫秒级别)。

    加入not exists后,查询效率大幅度降低,需要三秒多。
    执行计划我们看到走的是filter表关联方式

    用not in改写后走的是排序合并连接,查询耗时又恢复导毫秒级别。
  • 相关阅读:
    print(end=" ") 滚动输出到屏幕
    写入到csv文件的两种方式(pd.DaaFrame 和 csv.writerow)
    pytorch 损失函数(nn.BCELoss 和 nn.CrossEntropyLoss)(思考多标签分类问题)
    pytorch实战(二)hw2——预测收入是否高于50000,分类问题
    信用卡新颖的攻击方式,黑客可非接触式卡进行交易
    微软结束对SolarWinds黑客的内部调查
    Apple M1芯片首款恶意软件曝光
    新的Office恶意软件,黑客可远程窃取信息
    如何审核Active Directory中的密码更改
    特斯拉恶意软件使用新的传送和规避技术
  • 原文地址:https://www.cnblogs.com/muzisanshi/p/13370806.html
Copyright © 2020-2023  润新知