SQL 为什么在添加更多 where 条件时下面的查询很慢

如何解决SQL 为什么在添加更多 where 条件时下面的查询很慢

我遇到了一个奇怪的问题，当我添加更多条件时，我的 SQL 查询很慢。

我的表：

CREATE TABLE[dbo].[ASSETS_INFO]
(
[ASSET_ID][int] IDENTITY(1,1) NOT NULL,[OBJECT_ID] [int] NOT NULL,[EQUIPMENT_ID] [nvarchar] (20) NOT NULL,[EQUIPMENT_NAME] [nvarchar] (100) NULL,[USER_NO] [nvarchar] (20) NULL,[PRODUCTER] [int] NULL,[ASSET_TYPE] [nvarchar] (20) NULL,CONSTRAINT[PK_ASSETS_INFO] PRIMARY KEY CLUSTERED
(
[ASSET_ID] ASC
)WITH(PAD_INDEX = OFF,STATISTICS_NORECOMPUTE = OFF,IGNORE_DUP_KEY = OFF,ALLOW_ROW_LOCKS = ON,ALLOW_PAGE_LOCKS = ON) ON[PRIMARY]
) ON[PRIMARY] TEXTIMAGE_ON[PRIMARY]
GO

CREATE TABLE [dbo].[BM_METERINFO](
[METER_ID] [varchar](40) NOT NULL,[CUSTOMER_ID] [varchar](20) NOT NULL,[CUSTOMER_NAME] [varchar](200) NULL,CONSTRAINT [PK_BM_METERINFO] PRIMARY KEY CLUSTERED 
(
  [METER_ID] ASC,[CUSTOMER_ID] ASC
)WITH (PAD_INDEX = OFF,ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY]
GO

CREATE TABLE [dbo].[WSMP_OBJECT](
    [OBJECT_ID] [int] IDENTITY(1,[OBJECT_NAME] [nvarchar](50) NOT NULL,[OBJECT_STATE] [int] NOT NULL,[OBJECT_INDEX] [int] NULL,CONSTRAINT [PK_WSMP_OBJECT] PRIMARY KEY CLUSTERED 
(
   [OBJECT_ID] ASC
)WITH (PAD_INDEX = OFF,ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]
GO

我的 SQL 查询 1：

declare @table_count_f1 table
(
OBJECT_ID INT
);
insert into @table_count_f1
select OBJECT_ID from[FNNC_WAPCONFIG_AUTHOR_AREA_WSMPOBJECTIDS] ('2','lams_v2_dev','')  order by OBJECT_ID   -- get all object_ids which is very quick (more than 60,000 rows)

declare @table_count_f2 table
(
OBJECT_ID INT
);
insert into @table_count_f2
SELECT DISTINCT obj.OBJECT_ID FROM[ASSETS_INFO] A
INNER JOIN[WSMP_OBJECT] OBJ ON OBJ.OBJECT_ID = A.OBJECT_ID
LEFT JOIN dbo.BM_METERINFO bm ON A.USER_NO = BM.CUSTOMER_ID and a.EQUIPMENT_ID = bm.METER_ID
WHERE OBJ.OBJECT_STATE in (0,1,-2) AND(A.SYS_DATA_STATE <> -1 OR A.SYS_DATA_STATE IS NULL)
    AND A.ASSET_TYPE in('22')
    And case  WHEN bm.METER_ID IS  null THEN 1 WHEN bm.METER_ID = '' THEN 1 ELSE 2 END = 1
    --here is the problem: it got the result when I use either of 'AND',but it runs very slowly when I use both of them.I have added index already
order by OBJECT_ID

select count(*) from @table_count_f1 t1,@table_count_f2 t2
where t1.OBJECT_ID = t2.OBJECT_ID

执行计划

我的 SQL 查询 2：

declare @table_count_f0 table
(
OBJECT_ID INT
);
insert into @table_count_f0
select OBJECT_ID from[FNNC_WAPCONFIG_AUTHOR_AREA_WSMPOBJECTIDS]('2','') order by OBJECT_ID

SELECT A.*,OBJ.OBJECT_INDEX,OBJ.OBJECT_STATE,case  WHEN bm.METER_ID IS  null THEN 1 WHEN bm.METER_ID = '' THEN 1 ELSE 2 end setState FROM[ASSETS_INFO] A
INNER JOIN[WSMP_OBJECT] OBJ ON OBJ.OBJECT_ID = A.OBJECT_ID
LEFT JOIN dbo.BM_METERINFO bm ON a.EQUIPMENT_ID = bm.METER_ID AND A.USER_NO = BM.CUSTOMER_ID
WHERE OBJ.OBJECT_STATE IN(0,-2) AND(A.SYS_DATA_STATE <> -1 OR A.SYS_DATA_STATE IS NULL)

and OBJ.OBJECT_ID in (select OBJECT_ID from @table_count_f0 )
        and case  WHEN bm.METER_ID IS  null THEN 1 WHEN bm.METER_ID = '' THEN 1 ELSE 2 END = 2
ORDER BY A.EQUIPMENT_NAME asc
        OFFSET 20 ROWS
        FETCH NEXT 20 ROWS ONLY

问题出在ORDER BY，当我使用ORDER BY EQUIPMENT_ID时它很快，但是当我使用ORDER BY EQUIPMENT_NAME 时它很慢，我在ASSETS_INFO中添加了一个索引SSMS 但它没有效果。

感谢您抽出宝贵时间。

解决方法

我假设您已经有一个关于 WSMP_OBJECT (OBJECT_ID) 的覆盖索引，以及包含覆盖查询的 EQUIPMENT_NAME 的新索引。

您在这里的主要问题是事物的外观排序。您在整个结果集上区分的事实与 SQL Server 不了解您尝试使用左连接做什么的组合意味着它会再次排序，即使它可能已经对数据进行了排序。

本质上，您想要的是针对 WSMP_OBJECT 的半连接和针对 BM_METERINFO 的反连接。

SELECT
    obj.OBJECT_ID
FROM [ASSETS_INFO] A
INNER JOIN [WSMP_OBJECT] OBJ ON OBJ.OBJECT_ID = A.OBJECT_ID
WHERE OBJ.OBJECT_STATE in (0,1,-2)
    AND (A.SYS_DATA_STATE <> -1 OR A.SYS_DATA_STATE IS NULL)
    AND A.ASSET_TYPE in('22')
    AND NOT EXISTS (SELECT 1
        FROM dbo.BM_METERINFO bm
        WHERE A.USER_NO = BM.CUSTOMER_ID and a.EQUIPMENT_ID = bm.METER_ID
        AND bm.METER_ID <> '')
order by OBJECT_ID

同样（这里我相信 setState 永远是 2）：

SELECT
    A.*,OBJ.OBJECT_INDEX,OBJ.OBJECT_STATE,2 setState
FROM [ASSETS_INFO] A
INNER JOIN [WSMP_OBJECT] OBJ ON OBJ.OBJECT_ID = A.OBJECT_ID
WHERE OBJ.OBJECT_STATE IN(0,-2) AND(A.SYS_DATA_STATE <> -1 OR A.SYS_DATA_STATE IS NULL)
and OBJ.OBJECT_ID in (select OBJECT_ID from @table_count_f0 )
    AND NOT EXISTS (SELECT 1
        FROM dbo.BM_METERINFO bm
        WHERE A.USER_NO = BM.CUSTOMER_ID and a.EQUIPMENT_ID = bm.METER_ID
        AND bm.METER_ID <> '')
ORDER BY A.EQUIPMENT_NAME asc
        OFFSET 20 ROWS
        FETCH NEXT 20 ROWS ONLY

,

我不知道为什么下面的代码更快，它似乎仍然与索引有关。

修改后的 SQL 查询 1：

declare @table_count_f1 table
(
OBJECT_ID INT
);
insert into @table_count_f1
select OBJECT_ID from [FNNC_WAPCONFIG_AUTHOR_AREA_WSMPOBJECTIDS] ('2','lams_v2_dev','')  order by OBJECT_ID

declare @table_count_f2 table
(
OBJECT_ID INT,PRODUCTER INT,ASSET_TYPE INT
);
insert into @table_count_f2
SELECT DISTINCT obj.OBJECT_ID,A.PRODUCTER,A.ASSET_TYPE FROM [ASSETS_INFO] A 
INNER JOIN[WSMP_OBJECT] OBJ ON OBJ.OBJECT_ID = A.OBJECT_ID
LEFT JOIN dbo.BM_METERINFO bm ON  A.USER_NO = BM.CUSTOMER_ID and a.EQUIPMENT_ID = bm.METER_ID 
WHERE OBJ.OBJECT_STATE in (0,-2) AND(A.SYS_DATA_STATE <> -1 OR A.SYS_DATA_STATE IS NULL) 
--AND A.ASSET_TYPE in('22') 
-- AND A.PRODUCTER in ('1')
and case  WHEN bm.METER_ID IS  null THEN 1 WHEN bm.METER_ID='' THEN 1 ELSE 2 END=1 
 AND OBJ.OBJECT_STATE in(1) 
order by OBJECT_ID


select count(*) from @table_count_f1 t1,@table_count_f2 t2
where t1.OBJECT_ID=t2.OBJECT_ID 
AND t2.PRODUCTER in ('1')
AND t2.ASSET_TYPE in ('22')