Question

请你们查看以下对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)

查询统计信息：

所用时间：10.53秒。

指数：

t2.empno已编入索引。
t1.id已编入索引。
t2.id已编入索引。

更新

以上查询只是我使用的查询的示例副本。以下是更真实的表格

解释计划 Explain Plan

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
        )

索引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

首先出现的问题是使用implict join语法。这是一个SQL反模式。

如果您在IN子句中有一个大型列表，您是否尝试将它们放入表中并使用连接？

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

让我们暂时忘记empno BETWEEN 1 and 200的建议，并假设你有t2.empno IN (3,7,...,5209)（200个参赛作品）。

您还可以将您的查询（这是一个隐藏的JOIN查询）写入非等效 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，它甚至不是t2.empno IN (3,7,...,5209)而是t2.empno IN (SELECT tx.empno FROM tx WHERE someConditions) 和你正在使用MySQL，那么这就是你问题的根源（MySQL已知不是以最好的方式处理field IN (SELECT f FROM x)。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

甚至是：

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

Answer 3

请勿使用交叉连接。

试试这个

SELECT
    t1.name
FROM
    t1
JOIN t2
    ON t2.id = t1.id
WHERE
    t2.empno IN (1,...,200)

编辑：编辑完成后，查看笛卡尔产品中的多个表格，使用正确的JOIN语法可能非常重要。

SELECT子句使用IN ...非常慢？

3 个答案: