Convert rows to constant columns in SQL Server? - sql

Question

I have the following query with its result:
SELECT * FROM dbo.DeviceView AS dv
WHERE DeviceId = 5
Result:
Id Name AttachId ColorId Date
--- ------- ---------- ------- -------
5 Apple iPhone 5s A1533 NULL 1 2013-09-10 00:00:00.000
5 Apple iPhone 5s A1533 NULL 8 2013-09-10 00:00:00.000
5 Apple iPhone 5s A1533 NULL 19 2013-09-10 00:00:00.000
ColorId is within diffrent values and it can be more or less than 3 values
I want to convert ColorId to 3 columns, such as first value in ColorId1 and second value in ColorId2 and third value in ColorId3.
eg:
Id Name AttachId ColorId1 ColorId2 ColorId3 Date
--- ------- ---------- ---------- ---------- ---------- -------
5 Apple iPhone 5s A1533 NULL 1 8 19 2013-09-10 00:00:00.000
How can I convert it to the following?
Edit:
All other fields except ColorId are the same.

Answer 1

You can use this generic query based on PIVOT and CTE's that can easily be extended for any number of colors and that performs very well:
-- First, we assign unique numbers to each of the ColorId's. These will become column names
;WITH NumberedColors (ColorId,ColorNumber) AS (
SELECT ColorId,'Color'+CAST((ROW_NUMBER() OVER (ORDER BY ColorId)) AS VARCHAR) AS ColorNumber
FROM dbo.DeviceView
GROUP BY ColorId
),
-- Here we return the dbo.DeviceView extended with ColorNumber column name
DeviceViewWithNumberedColors (Id,Name,AttachId,[Date],ColorNumber,ColorId) AS (
SELECT Id,Name,AttachId,Date,NC.ColorNumber,NC.ColorId
FROM dbo.DeviceView DV
INNER JOIN NumberedColors NC ON DV.ColorId=NC.ColorId
)
-- Finally, we use the PIVOT to assign color's to the appropriate columns
SELECT *
FROM (
SELECT Id,Name,AttachId,[Date],ColorId,ColorNumber
FROM DeviceViewWithNumberedColors D
) AS Source
PIVOT (
SUM(ColorId) FOR ColorNumber IN ([Color1],[Color2],[Color3],[Color4],[Color5],[Color6],[Color7],[Color8],[Color9],[Color10])
) Piv
In the PIVOT clause, make sure you have enough Color columns. If this can not be hard-coded, i.e. the number of colors could grow beyond a fixed number, then use dynamic SQL to generate this query.

Answer 2

If you want exactly three colors, you can use conditional aggregation or pivot:
select id, name, attachid,
max(case when seqnum = 1 then color end) as color1,
max(case when seqnum = 2 then color end) as color2,
max(case when seqnum = 3 then color end) as color3,
date
from (select t.*,
row_number() over (partition by id order by (select null)) as seqnum
from t
) t
group by id, name, attachid, date;

Answer 3

If count of ColorId values is not a fixed number you must use dynamic sql.
DECLARE #sql nvarchar(MAX) = 'SELECT Id, Name, AttachId'
SELECT #sql = #sql
+ ', MAX(IIF(ValueNum = ' + LTRIM(STR(ColumnNum)) + ', ColorId, NULL)) AS ColorId' + LTRIM(STR(ColumnNum))
FROM (
SELECT DISTINCT
DENSE_RANK() OVER (ORDER BY ColorId) AS ColumnNum
FROM dbo.DeviceView
WHERE Id IN (
SELECT TOP 1 Id
FROM dbo.DeviceView
GROUP BY Id
ORDER BY COUNT(DISTINCT ColorId) DESC)
) AS c
EXEC (#sql
+ ', [Date] FROM ('
+ 'SELECT Id, Name, AttachId, ColorId, [Date], DENSE_RANK() OVER (PARTITION BY Id ORDER BY ColorId) AS ValueNum FROM dbo.DeviceView'
+ ') AS it '
+ 'GROUP BY Id, Name, AttachId, [Date]');