Rotate/pivot 1000 columns to rows - sql

Question

I inherited a DB from which I need to extract data in columns that is currently arranged in rows. The query to retrieve the data is:
select distinct
LIMSNo, p250.*, p500.*, p750.*, p1000.*
from
results as r
inner join Points250 as p250
on r.resultsid = p250.resultsid and p250.channel = 1
inner join Points500 as p500
on r.resultsid = p500.resultsid and p500.channel = 1
inner join Points750 as p750
on r.resultsid = p750.resultsid and p750.channel = 1
inner join Points1000 as p1000
on r.resultsid = p1000.resultsid and p1000.channel = 1
where
r.limsno between '120053698' and '120053704'
which produces
LIMSNo P0001 P0002 P0003 ... P1000
120053698 251.6667 302.0196 302.2353 305.9608
120053699 291.6667 342.6545 347.9635 353.6236
120053700 243.3333 298.3206 296.7235 299.5342
120053701 308.3333 365.8397 365.4071 368.3206
120053702 315 363.4153 366.6052 373.1695
Note that there are 1000 columns (P0001...P1000)
I want to pivot/rotate the data so that I get this output:
120053698 120053699 120053700 120053701 120053702
P0001 251.6667 291.6667 243.3333 308.3333 315
P0002 302.0196 342.6545 298.3206 365.8397 363.4153
...
P1000 305.9608 353.6236 299.5342 368.3206 373.1695
How do I structure the SQL query? The DB is SQL 2000.
I have tried a number of approaches to this problem. My latest attempt was:
create table #tempCols
(
ColName nchar(15)
)
insert into #tempCols
select
column_name
from
information_schema.columns
where
table_name = 'Points250'
and column_name like 'P%'
insert into #tempCols
select
column_name
from
information_schema.columns
where
table_name = 'Points500'
and column_name like 'P%'
insert into #tempCols
select
column_name
from
information_schema.columns
where
table_name = 'Points750'
and column_name like 'P%'
insert into #tempCols
select
column_name
from
information_schema.columns
where
table_name = 'Points1000'
and column_name like 'P%'
create table #LIMSList
(
LIMSNumber nchar(9)
)
insert into #LIMSList
select LimsNo from results
where LimsNo between '100030460' and '100030500'
declare #colList varchar(max)
select #colList = COALESCE(#colList + ',' ,'') + ColName from #tempCols
declare #command varchar(max)
set #command =
'
select LimsNo, Point, Data
from (
select LimsNo, p250.*, p500.*, p750.*, p1000.*
from
results as r
inner join Points250 as p250
on r.resultsid = p250.resultsid and p250.channel = 1
inner join Points500 as p500
on r.resultsid = p500.resultsid and p500.channel = 1
inner join Points750 as p750
on r.resultsid = p750.resultsid and p750.channel = 1
inner join Points1000 as p1000
on r.resultsid = p1000.resultsid and p1000.channel = 1
where
limsno in (select LimsNo from #LIMSList)) d
unpivot (Data for Point in (' + #colList + ')) as unpvt
'
print #command
exec(#command)
drop table #tempCols
drop table #LIMSList
But the error I get is:
Msg 8156, Level 16, State 1, Line 1
The column 'ResultsID' was specified multiple times for 'd'.
Msg 8156, Level 16, State 1, Line 1
The column 'ResultsID' was specified multiple times for 'unpvt'.

Answer 1

Since SQL Server 2000 does not have either an UNPIVOT or PIVOT operator, you will have to use UNION ALL for the UNPIVOT and then a CASE statement for the PIVOT. Here is a dynamic solution that should work:
DECLARE #query AS NVARCHAR(MAX),
#rowCount as int = 1,
#unpivotCount as int = 0,
#pivotCount as int,
#unRow as varchar(10) = '',
#pivotRow as varchar(10) = ''
select c.Name col
into #colsUnpivot
from sys.columns as C
where C.object_id = object_id('yourtable') and
C.name LIKE 'P%'
set #unpivotCount = (select count(*) from #colsUnpivot)
-- unpivot the data
while #rowCount <= #unpivotCount
begin
set #unRow = (select Top 1 col from #colsUnpivot)
set #query =' insert into unpivotData (LIMSNo, Val, Col)
select LIMSNo, ' + #unRow + ' as val, '''
+ #unRow + ''' as col
from yourtable'
exec(#query)
delete from #colsUnpivot where col = #unRow
if #rowCount <= #unpivotCount
set #rowCount = #rowCount + 1
end
--select *
--from unpivotData
-- pivot the data
select distinct LIMSNo
into #colsPivot
from yourtable
set #pivotCount= (select COUNT(*) from #colsPivot)
-- reset rowcount
set #rowCount = 1
set #query = ''
---- create the CASE string
while #rowCount <= #pivotCount
begin
set #pivotRow = (select Top 1 LIMSNo from #colsPivot)
set #query = #query + ', max(case when LIMSNo = ' + #pivotRow + ' then val end) as ''' + #pivotRow + ''''
delete from #colsPivot where LIMSNo = #pivotRow
if #rowCount <= #pivotCount
begin
set #rowCount = #rowCount + 1
print #rowCount
end
end
-- add the rest of the SQL Statement
set #query = 'SELECT col ' + #query + ' from dbo.unpivotData group by col'
exec(#query)
delete from dbo.unpivotData
drop table #colsUnpivot
drop table #colspivot
I created a table called UnpivotData to hold the data for use with in the PIVOT process.
See SQL Fiddle with Demo

Answer 2

pivot and unpivot operators are not available in SQL Server 2000. So you have to pivot and unpivot the data using other techniques.
I am listing static methods of unpivot and pivot and i hope you can convert it into dynamic solution if you want.
assuming you have inserted the data of your query into a table you can unpivot the data using UNION ALL and insert it into temporary table.
SELECT
*
INTO dbo.temp
FROM
(
SELECT LIMSNo, P0001 AS [Point],'P0001' AS [ColName] FROM dbo.myTable
UNION ALL
SELECT LIMSNo, P0002 AS [Point],'P0002' AS [ColName] FROM dbo.myTable
UNION ALL
SELECT LIMSNo, P0003 AS [Point],'P0003' AS [ColName] FROM dbo.myTable
UNION ALL
SELECT LIMSNo, P0004 AS [Point],'P0004' AS [ColName] FROM dbo.myTable
....
....
UNION ALL
SELECT LIMSNo, P1000 AS [Point],'P1000' AS [ColName] FROM dbo.myTable
) AS t
then you can pivot the data again
SELECT
ColName
,MAX(CASE LIMSNo WHEN '120053698' THEN [Point] ELSE NULL END ) AS [120053698]
,MAX(CASE LIMSNo WHEN '120053699' THEN [Point] ELSE NULL END ) AS [120053699]
,MAX(CASE LIMSNo WHEN '120053700' THEN [Point] ELSE NULL END ) AS [120053700]
....
....
,MAX(CASE LIMSNo WHEN '120053704' THEN [Point] ELSE NULL END ) AS [120053704]
FROM dbo.temp
GROUP BY ColName