计数函数中的 SQL 动态分区

Question

我有一个与报告服务一起使用的存储过程，并且该报告有一些将值传递给存储过程的用户过滤器。由于这些参数，显示的计数会发生变化，所以我想在分区中使用它们来更改计数。我尝试了以下不起作用。

编辑：

计数是在 ss_number 上完成的，而不是在 @SearchBy 上完成的。所以如果我有

SS#       |    Name    |    City  |  Amount

123456789 | Mike Smith | Trenton | 100.00

123456789 | Mike Smith | Trenton | 200.00

123456789 | Mike Smith | Jackson | 100.00

我的计数是 3，即使 @SearchBy = City 并且我在特伦顿进行过滤。

CASE WHEN @SearchBy = 'Product Name' THEN count(ss_number) OVER (PARTITION BY ss_number, @SearchBy) 

现在我正在使用 case 语句，但它大大减慢了我的查询速度。

这是尝试在我的分区中使用变量的没有 case 语句的代码。

CREATE PROCEDURE [dbo].[sp_My_sp]
@SearchBy VARCHAR(MAX), 
@SearchString VARCHAR(MAX),
@SearchNum Int,
@ClaimDate Datetime2

AS
WITH MyCTE AS
(   
SELECT val.Claim_date
, val.Claim_Status
, val.Status_Desc
, ES_Claim_Status
, val.ss_number
, val.name_field1
, val.street_add1
, val.street_add2
, val.city
, val.state
, val.zip_code_pre
, val.reference_number
, val.Game_Name 
, val.val_agent
, val.home_number
, val.work_phone
, val.county_desc
, o.agent_num
, count(ss_number) OVER (PARTITION BY ss_number, @SearchBy) as count
, prize_amount
FROM Sec_Claims val left outer join vw_owners_concat_agent_num o
ON val.SS_NUMBER = convert(varchar(15),o.SS_NO)
where convert(numeric,prize_amount) >= 600)
SELECT Claim_date
, CASE WHEN agent_num IS NULL THEN 'NO' ELSE 'YES' END as "IsRetailer"
, Claim_Status
, Status_Desc
, ES_Claim_Status
, ss_number
, name_field1
, street_add1
, street_add2
, city
, state
, zip_code_pre
, reference_number
, Game_Name 
, val_agent
, home_number
, work_phone
, county_desc
, agent_num
, count
, ROW_NUMBER() OVER(PARTITION BY Name_Field1 ORDER BY Name_Field1) As RowNumber
, convert(decimal(10,2),prize_amount) as prize_amount
, sum(Convert(decimal(9,2),prize_amount)) OVER (PARTITION BY ss_number, Name_Field1) AS prizesum
FROM MyCTE 
WHERE 
    (CASE
        WHEN @SearchBy = 'Agent Number' THEN agent_num
        WHEN @SearchBy = 'SS#' THEN SS_NUMBER
        WHEN @SearchBy = 'Name' THEN Name_Field1
        WHEN @SearchBy = 'Address' THEN STREET_ADD1
        WHEN @SearchBy = 'City' THEN City
        WHEN @SearchBy = 'Claim#' THEN convert(varchar(max),REFERENCE_NUMBER)
        WHEN @SearchBy = 'Validating Retailer' THEN convert(varchar(max),VAL_AGENT)
        WHEN @SearchBy = 'County' THEN COUNTY_DESC
        WHEN @SearchBy = 'Home Phone' THEN convert(varchar(max),HOME_NUMBER)
        WHEN @SearchBy = 'Work Phone' THEN convert(varchar(max),WORK_PHONE)
        WHEN @SearchBy = 'Game Name' THEN GAME_NAME
    END 
    like (@SearchString))
    and
    count>= @SearchNum 
    and
    claim_date > @ClaimDate 
ORDER BY ss_number

Answer 1

您的查询速度变慢了，因为您可以想象，需要在 Select 语句中对每一行评估“case”语句。
这意味着，如果您的查询加载了许多行，则会产生大量开销。

我看到两种不同的方法来解决您的问题。我不喜欢它们中的任何一个，但它们很有效:)

第一个解决方案
使用 IF ... ELSE 覆盖所有可能的输入。 这意味着您必须多次编写相同的查询才能在一行代码中区分它。
这意味着代码变得冗余并且基本上不可维护

第二个解决方案
构建动态查询。
你可以这样写

DECLARE @myQuery nvarchar(max)
SET @myQuery = 'Select '
if @SearchBy = 'Product Name'
  SET @myQuery= @myQuery + 'count(ss_number) OVER (PARTITION BY ss_number, @SearchBy)'
ELSE
  .... --put here your second case
END

SET @myQuery= @myQuery + 'FROM ... WHERE ...'
EXECUTE spexecute_sql @myQuery

Answer 2

尝试在 Over 语句中使用“Case”：

Declare @SearchBy varchar(255)
SET @SearchBy = 'Product Name'

Select count(ss_number)
            OVER (PARTITION BY ss_number,
                               Case @SearchBy
                                    When 'Product Name' then [Name]
                                    When 'City Name'    then [City]
                                    When 'Product Name' then cast([Amount] as varchar(255))
                               END) t     
    from (VALUES (123456789,'Mike Smith','Trenton',100.00)
                ,(123456789,'Mike Smith','Trenton',200.00)
                ,(123456789,'Mike Smith','Jackson',100.00)
         ) as t([ss_number],Name,City,Amount)

Answer 3

根据您的需求，您首先需要动态 SQL。话虽如此，您可以使用row_number()。我仍然不明白您的额外分区，但您可以将其添加到下面的示例中。你可以运行它，它会打印出要执行的命令。

--create procedure yourProc(
--                          @SearchBY varchar(64)           --Column to search / filter by
--                          ,@SearchString varchar(256)     --Values to limit column above by
--                          ,@Count int                     --limit the count per each SSN
--                          ,@Date datetime)                
--as

--begin

declare     @SearchBY varchar(64)       = 'City'            --Column to search / filter by
declare     @SearchString varchar(256)  = 'Trenton'         --Values to limit column above by
declare     @Count int                  = 3                 --limit the count per each SSN
declare     @Date datetime              = '12/25/2017'

declare @SQL nvarchar(max) = 

'with cte as(
    select
        ss_number
        ,Name
        ,City
        ,Amount
        ,row_number() over (partition by ss_number order by (select 1)) as RN
    from
        YourTable
    where
        SomeDateColumn >= ''' + convert(varchar(10),@Date,112) +
        ''' and ' + @SearchBY + ' = ''' + @SearchString +
        ''')
 select *
 from cte
 where ss_number in 
 (select distinct ss_number from cte where RN >= ' + cast(@Count as varchar(16)) + ')
 '

print(@SQL)
--exec(@SQL)
--end

这是一个工作示例：http://rextester.com/BPWOXH7777