SQL 一对多在一行答案

【问题标题】：SQL one to many on one rowSQL 一对多在一行
【发布时间】：2013-12-11 12:48:25
【问题描述】：

我有一张表格，记录每次就诊的患者诊断情况。患者可以有不止一种诊断。

我希望每个出勤人员在一行中包含所有诊断（最多 12 个诊断）。下面带回不止一行。

select
    dg.AttendanceID
    , dg.PatientNumber
    , dg.DiagnosisDate
    , dg.Diagnosis
from
    Diagnosis dg


AttendanceID    PatientNumber   DiagnosisDate   Diagnosis
10001           123456          01-Oct-13   A
10001           123456          01-Oct-13   B
10002           123456          20-Oct-13   D

结果应该是这样的：

AttendanceID    PatientNumber   DiagnosisDate   Diagnosis 1 Diagnosis 2 Diagnosis 3
10001           123456          01-Oct-13   A            B       
10002           123456          20-Oct-13   D

有人可以帮忙吗？

【问题讨论】：

我相信 PIVOT 可以在这里完成这项工作？

标签： sql sql-server sql-server-2005 pivot

【解决方案1】：

您可以通过实现 PIVOT 函数来获得结果，但我也建议使用窗口函数row_number() 来生成每个attendanceid 和patientnumber 的诊断数：

select attendanceid, patientnumber,
  diagnosisdate,
  Diagnosis1, Diagnosis2, Diagnosis3
from
(
  select attendanceid, patientnumber,
    diagnosisdate, diagnosis,
    'diagnosis'+
      cast(row_number() over(partition by attendanceid, patientnumber
                              order by diagnosis) as varchar(2)) seq
  from diagnosis
) d
pivot
(
  max(diagnosis)
  for seq in (Diagnosis1, Diagnosis2, Diagnosis3)
) piv;

见SQL Fiddle with Demo。由于您知道每位患者/就诊最多会有 12 个诊断，因此您可以轻松地对查询进行硬编码以添加其他列。

但如果您需要代码的动态版本，那么您可以使用类似于以下内容的代码：

DECLARE @cols AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT distinct ',' + QUOTENAME(seq) 
                    from
                    (
                       select 'diagnosis'+
                          cast(row_number() over(partition by attendanceid, patientnumber
                               order by diagnosis) as varchar(2)) seq
                       from diagnosis
                    ) d
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query = 'SELECT attendanceid, patientnumber,
               diagnosisdate,' + @cols + ' 
            from 
            (
              select attendanceid, patientnumber,
                diagnosisdate, diagnosis,
                ''diagnosis''+
                  cast(row_number() over(partition by attendanceid, patientnumber
                                          order by diagnosis) as varchar(2)) seq
              from diagnosis
            ) x
            pivot 
            (
                max(diagnosis)
                for seq in (' + @cols + ')
            ) p '

execute sp_executesql @query;

见SQL Fiddle with Demo

两个版本都给出了结果：

| ATTENDANCEID | PATIENTNUMBER |                  DIAGNOSISDATE | DIAGNOSIS1 | DIAGNOSIS2 | DIAGNOSIS3 |
|--------------|---------------|--------------------------------|------------|------------|------------|
|        10001 |        123456 | October, 01 2013 00:00:00+0000 |          A |          B |     (null) |
|        10002 |        123456 | October, 20 2013 00:00:00+0000 |          D |     (null) |     (null) |

【讨论】：

【解决方案2】：

我不是 100% 确定，但应该是这样的。

;with cte as(
select
dg.AttendanceID
, dg.PatientNumber
, dg.DiagnosisDate
, dg.Diagnosis
,ROW_NUMBER() over (partition by dg.attendanceID order by attendanceID) as seq1
,ROW_NUMBER() over (partition by dg.attendanceID, patient_no order by diagnosis) as seq2
from
Diagnosis dg
)

 select
 t1.attendanceID
 ,t1.patientNumber
 ,t1.diagnosisDate
 ,(select Diagnosis from cte t2 where t1.attendanceID=t2.attendanceID and             t1.patientNumber=t2.patientNumber and t2.seq2='1') as diag1
 ,(select Diagnosis from cte t2 where t1.attendanceID=t2.attendanceID and    t1.patientNumber=t2.patientNumber and t2.seq2='2') as diag2
 ,(select Diagnosis from cte t2 where t1.attendanceID=t2.attendanceID and t1.patientNumber=t2.patientNumber and t2.seq2='3') as diag3
 ,(select Diagnosis from cte t2 where t1.attendanceID=t2.attendanceID and t1.patientNumber=t2.patientNumber and t2.seq2='4') as diag4
 ,(select Diagnosis from cte t2 where t1.attendanceID=t2.attendanceID and t1.patientNumber=t2.patientNumber and t2.seq2='5') as diag5
  from cte t1
  where seq1= 1

【讨论】：

你不认为你需要分组吗？
这行得通，但我不明白复杂的语法。谢谢，

【解决方案3】：

以下是使用光标的可能解决方案

DECLARE @combinedString VARCHAR(MAX),
@id int,
@Diagnosis VARCHAR(MAX)


Declare @CONCATRESULT TABLE (AttendanceID int, Question VARCHAR(MAX) )

Declare Dia cursor for 
SELECT  AttendanceID FROM  Diagnosis 

open Dia 
Fetch next from Dia into @id
while @@FETCH_STATUS=0 
begin

SELECT   @combinedString = COALESCE(@combinedString + ', ', '') + Diagnosis,         
@id=AttendanceID  FROM [Prod_PostData].[dbo].[KMSFPostDataDenorm] d
WHERE AttendanceID = @id

insert into @CONCATRESULT values (  @id ,@combinedString )

SET @combinedString = null

Fetch next from Dia into @id
end
close Dia
deallocate Dia


select
dg.AttendanceID
, dg.PatientNumber
, dg.DiagnosisDate
, dg.Diagnosis
from
Diagnosis dg join (select * from @CONCATRESULT) Y on dg.AttendanceID=Y.AttendanceID

【讨论】：