提高 SQL 服务器中慢速子查询的性能

Question

I have a table with 1.3 million rows.

Query #1 takes 29 seconds to run in SQL Server 2016 Management Studio.

Query #1:

select 
    *,
    (select Count(*)
     from [dbo].[Results] t2
     where t2.RaceDate < t1.RaceDate
       and t1.HorseName = t2.HorseName
       and t2.Position = '1'
       and t1.CourseName = t2.CourseName
       and t2.CountryCode = 'GB') as [CourseDistanceWinners]
from 
    [dbo].[Results] t1

But query #2 takes takes several hours with the only difference being t1.HorseName = t2.HorseName vs t1.TrainerName = t2.TrainerName . There will be many more matches but on TrainerName than HorseName but I wasn't expecting several hours.

Query #2:

select 
    *,
    (select Count(*)
     from [dbo].[Results] t2
     where t2.RaceDate < t1.RaceDate
       and t1.TrainerName = t2.TrainerName
       and t2.Position = '1'
       and t1.CourseName = t2.CourseName
       and t2.CountryCode = 'GB') as [CourseDistanceWinners]
from 
    [dbo].[Results] t1

I've managed to get the query down to 15 minutes using the techniques below but I still think this is a very long time. Is there anything else I can do to improve performance of Query2 or a way to rewrite it for performance?

What I have tried so far

1. I've changed [TrainerName] [nvarchar](255) NULL, to [TrainerName] [nvarchar](50) NULL,

2. I've added a composite index and several non clustered indexes

   CREATE INDEX idx_HorseName ON [dbo].[Results] (HorseName); CREATE INDEX idx_TrainerName ON [dbo].[Results] (TrainerName); CREATE INDEX idx_CourseName ON [dbo].[Results] (CourseName); CREATE INDEX idx_Position ON [dbo].[Results] (Position); CREATE INDEX idx_JockeyName ON [dbo].[Results] (JockeyName); CREATE INDEX idx_RaceDate ON [dbo].[Results] (RaceDate); CREATE INDEX idx_TrainerComposite ON [dbo].[Results] (TrainerName, RaceDate, CourseName);

Further info:

Table structure:

SET QUOTED_IDENTIFIER ON
GO

CREATE TABLE [dbo].[Results]
(
    [CountryCode] [NVARCHAR](50) NULL,
    [CourseName] [NVARCHAR](50) NULL,
    [HorseName] [NVARCHAR](50) NOT NULL,
    [HorseSuffix] [NVARCHAR](5) NOT NULL,
    [JockeyName] [NVARCHAR](255) NULL,
    [OwnerName] [NVARCHAR](255) NULL,
    [Position] [NVARCHAR](255) NULL,
    [PublishedTime] [NVARCHAR](6) NOT NULL,
    [RaceDate] [DATETIME] NOT NULL,
    [RaceTitle] [NVARCHAR](255) NULL,
    [StallPosition] [NVARCHAR](255) NULL,
    [TrainerName] [NVARCHAR](50) NULL,
    [Rating] [INT] NULL,

    CONSTRAINT [PK_Results_1] 
        PRIMARY KEY CLUSTERED ([HorseName] ASC,
                               [HorseSuffix] ASC,
                               [PublishedTime] ASC,
                               [RaceDate] ASC)
                    WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
                          IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, 
                          ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY]
GO

Query #1 execution plan:

Query #2 execution plan:

Answer 1

Use a window function!

select r.*,
       sum(case when position = 1 and country_code = 'GB' then 1 else 0 end) over
           (partition by horsename, coursename
            order by racedate
            rows between unbounded preceding and 1 preceding
           ) as CourseDistanceWinners
from [dbo].[Results] r