Question

我需要一些帮助，试着让这个LIKE语句更加优雅和高效。它表现了我对它的期望，但它很慢，即使在小型数据集上也是如此。我在win 10机器上使用带有SSMS 2017的SQL-Server 2016。提前谢谢！

where (a.Description like '% ' + @event + ' %'  
        or a.Description like '% ' + @event + '.'   
        or a.Description like '% ' + @event + '.%'  
        or a.Description like '% ' + @event + ',%'  
        or a.Description like '% ' + @event + ';%'
        or a.Description like '%'  + @event + '/%'
        or a.Description like '%/' + @event + '%'
        or a.Description like '% ' + @event 
        or a.Description like @event + ' %' 
        or a.Description like @event 
        )

@event只是varchar喜欢＆＃39;死亡＆＃39;或者＆＃39;绊倒＆＃39;。如果你有更好的方法，一定要上学吧！

Answer 1

T-SQL支持LIKE中的字符类，因此您可以简化：

where (a.Description like '%[ /]' + @event + '[ .,;/]%'
        or a.Description like '%[ /]' + @event
        or a.Description like @event + '[ .,;/]%'
        or a.Description like @event 
        )

如果您愿意接受任何非字母分隔符（数字相关），您可以执行以下操作：

where (a.Description like '%[^a-z0-9]'+@event+'[^a-z0-9]%'
       or a.Description like @event+'[^a-z0-9]%'
       or a.Description like '%[^a-z0-9]'+@event
       or a.Description like @event
       )

Answer 2

假设描述是非索引列的一种方法可能使它更快，那就是从原始描述字符串中创建较小的字符串，并对这些较小的字符串进行比较。

除非说明栏中的数据较少，否则这很好。

如果您可以预先存储并索引存储的数据列，则可以尝试使用下面的动态

我做了标记，发现在N = 13之后，％string％的表现变差了

if exists (select object_id('t') )
drop table t
create table t (description varchar(max));

insert into t  values
('Implemantion of the CR systems at dealership level for understanding of the customer concern & faster resolution'),
('Trend analysis for Customer Complaint Ratio at Regional / Zonal / Area level both for Sales and Service operations');

declare @counter int
set @counter=1

WHILE @counter<15
begin
insert into t
select * from t
set @counter=@counter+1
end
declare @start datetime,@end1 datetime,@end2 datetime


declare @lookingfor varchar(100)
set @lookingfor ='level'

select 
description,
cast(substring(t.description,N,LEN(@lookingfor)) as varchar(100)) as d into #t from t
outer apply
(select top (select LEN(t.description)-LEN(@lookingfor)+1)
 row_number() over(order by (select null)) N
  from sys.objects s1 cross join sys.objects s2
)T1

CREATE NONCLUSTERED INDEX IDX_C_temp_d ON #t(d)

set @start=getdate()
select count(distinct description) from t
where description like '%'+@lookingfor+'%'

select @end1=getdate()

select count(distinct description) from #t  where d like @lookingfor

select @end2=getdate()

select datediff(ss,@start,@end1)

select datediff(ss,@end1,@end2)

select @start,@end1,@end2

drop table #t

PS：您需要对解决方案进行基准测试。

处理这个复杂的LIKE语句的更有效/更优雅的方式？

2 个答案: