SELECT子句使用IN ...非常慢？

Question

你们可以查看以下对Oracle DB的查询并指出错误:

SELECT t1.name FROM t1, t2 WHERE t1.id = t2.id AND t2.empno IN (1, 2, 3, …, 200)

查询统计:

• 所用时间:10.53秒.

指数:

• t2.empno 被编入索引.

• t1.id 被编入索引.

• t2.id 被编入索引.

更新

---

上面的查询只是我使用的查询的示例副本.下面是一个更真实的形式

解释计划

查询:

SELECT 
    PRODUCT_REPRESENTATION_SK 
FROM 
    Product_Representation pr 
    , Design_Object do
    , Files files  
    ,EPS_STATUS epsStatus 
    ,EPS_ERROR_CODES epsError
    ,VIEW_TYPE viewTable  
WHERE  
    pr.DESIGN_OBJECT_SK = do.DESIGN_OBJECT_SK 
    AND  pr.LAYER_NAME !='Layer 0' 
    AND  epsStatus.EPS_STATUS_SK = pr.EPS_STATUS  
    AND epsError.EPS_ERROR_CODE = pr.EPS_ERROR_CODE 
    AND viewTable.VIEW_TYPE_ID = pr.VIEW_TYPE_ID 
    AND  files.pim_id = do.PIM_ID 
    AND  do.DESIGN_OBJECT_ID IN
        (
147086,149924,140458,135068,145197,134774,141837,138568,141731,138772,143769,141739,149113,148809,141072,141732,143974,147076,143972,141078,141925,134643,139701,141729,147078,139120,137097,147072,138261,149700,149701,139127,147070,149702,136766,146829,135762,140155,148459,138061,138762.............................................  200 such numbers
        )

Colled Colums:

pr.DESIGN_OBJECT_SK
do.DESIGN_OBJECT_SK
do.DESIGN_OBJECT_ID
files.pim_id

表

---

TABLE "PIM"."DESIGN_OBJECT" 
(   
"DESIGN_OBJECT_SK" NUMBER(*,0) NOT NULL ENABLE, 
"PIM_ID" NUMBER(*,0) NOT NULL ENABLE, 
"DESIGN_OBJECT_TYPE_SK" NUMBER(*,0) NOT NULL ENABLE, 
"DESIGN_OBJECT_ID" VARCHAR2(40 BYTE) NOT NULL ENABLE, 
"DIVISION_CD" NUMBER(*,0), 
"STAT_IND" NUMBER(*,0) NOT NULL ENABLE, 
"STAT_CHNG_TMST" TIMESTAMP (6), 
"CRTD_BY" VARCHAR2(45 BYTE), 
"CRT_TMST" TIMESTAMP (6), 
"MDFD_BY" VARCHAR2(45 BYTE), 
"CHNG_TMST" TIMESTAMP (6), 
"UPDATE_CNT" NUMBER(*,0), 
"GENDER" VARCHAR2(1 BYTE), 

 PRIMARY KEY ("DESIGN_OBJECT_SK")
)
TABLESPACE "PIM"  ENABLE, 

FOREIGN KEY ("DESIGN_OBJECT_TYPE_SK")
    REFERENCES "PIM"."DESIGN_OBJECT_TYPE" ("DESIGN_OBJECT_TYPE_SK")
        ON DELETE CASCADE ENABLE, 

FOREIGN KEY ("PIM_ID")
    REFERENCES "PIM"."FILES" ("PIM_ID")
        ON DELETE CASCADE ENABLE

)

表2

---

CREATE TABLE "PIM"."PRODUCT_REPRESENTATION" 
(
"PRODUCT_REPRESENTATION_SK" NUMBER(*,0) NOT NULL ENABLE, 
"DESIGN_OBJECT_SK" NUMBER(*,0) NOT NULL ENABLE, 
"VIEW_TYPE_ID" NUMBER(*,0) NOT NULL ENABLE, 
"LAYER_NAME" VARCHAR2(255 BYTE), 
"STAT_IND" NUMBER(*,0) NOT NULL ENABLE, 
"STAT_CHNG_TMST" TIMESTAMP (6), 
"CRTD_BY" VARCHAR2(45 BYTE), 
"CRT_TMST" TIMESTAMP (6), 
"MDFD_BY" VARCHAR2(45 BYTE), 
"CHNG_TMST" TIMESTAMP (6), 
"UPDATE_CNT" NUMBER(*,0), 
"EPS_STATUS" VARCHAR2(30 BYTE) NOT NULL ENABLE, 
"EPS_GENERATED_TIME" TIMESTAMP (6), 
"EPS_ERROR_CODE" NUMBER, 
"EPS_ERROR_DETAILS" VARCHAR2(500 BYTE), 
"DEEPSERVER_ASSET_LAYER_ID" VARCHAR2(255 BYTE), 
"PRODUCT_REPRESENTATION_LOC" VARCHAR2(255 BYTE), 

 PRIMARY KEY ("PRODUCT_REPRESENTATION_SK")
)
TABLESPACE "PIM"  ENABLE, 

FOREIGN KEY ("DESIGN_OBJECT_SK")
    REFERENCES "PIM"."DESIGN_OBJECT" ("DESIGN_OBJECT_SK") 
        ON DELETE CASCADE ENABLE, 
FOREIGN KEY ("VIEW_TYPE_ID")
    REFERENCES "PIM"."VIEW_TYPE" ("VIEW_TYPE_ID")
        ON DELETE CASCADE ENABLE, 

CONSTRAINT "EPS_ERROR_CODE_FK"
FOREIGN KEY ("EPS_ERROR_CODE")
     REFERENCES "PIM"."EPS_ERROR_CODES" ("EPS_ERROR_CODE") 
        ON DELETE CASCADE ENABLE, 
CONSTRAINT "EPS_STATUS_FK" 
FOREIGN KEY ("EPS_STATUS")
    REFERENCES "PIM"."EPS_STATUS" ("EPS_STATUS_SK") 
        ON DELETE CASCADE ENABLE
) 

Answer 1

首先出错的是使用implict join语法.这是一个SQL反模式.

如果你在IN子句中有一个大的列表,你是否尝试过将它们放在一个表中并使用连接？

什么数据库？您是否查看过您的解释计划或执行计划,以了解减速的位置？

Answer 2

让我们忘记一下这个empno BETWEEN 1 and 200建议,并假设你有t2.empno IN (3,7,...,5209)200个参赛作品.

您还可以将查询(这是一个隐藏的JOIN查询)写入非等效的 EXISTS查询,该查询将显示相同的结果(但可能更少的行)并且应该比JOIN更快:

SELECT
    t1.name
FROM
    t1
WHERE EXISTS
      ( SELECT *
        FROM t2
        WHERE t2.id = t1.id
          AND t2.empno IN (3,7,...,5209)
      )

(狂野猜测)

如果另一方面,它甚至t2.empno IN (3,7,...,5209)不是,t2.empno IN (SELECT tx.empno FROM tx WHERE someConditions) 而你正在使用MySQL,那么这就是你问题的根源(众所周知,MySQL无法field IN (SELECT f FROM x)以最佳方式处理).因此,您可以将查询更改为:

SELECT
    t1.name
FROM
    t1
  JOIN t2
    ON t2.id = t2.id
  JOIN tx
    ON tx.empno = t2.empno
WHERE
    someConditions

甚至:

SELECT
    t1.name
FROM
    t1
WHERE EXISTS
      ( SELECT *
        FROM t2
          JOIN tx
            ON tx.empno = t2.empno
        WHERE t2.id = t1.id
          AND someConditions
      )