Remove duplicates with row pivot - sql

Question

I want to get rid of duplicates in one column (device_name) but keep related data from another column (app_id). Each device can have couple of applications (1 -> x, usually between 1-5) so I want to put those app identifiers into a new columns which I want to call [APP1],[APP2],[APP3] and so on. The best option would be kind of dynamic Pivot, but any static solution will be welcome as well.
Thanks for help in advance.
PS
I came up with below code what so ever, but in only concatenate APP id's separated by comas into one column.
USE tempdb;
SELECT DEVICE_NAME,
NoOfApps,
STUFF(( SELECT ', ' + APP_ID
FROM dbo.Aperture_full_test apps
WHERE apps.DEVICE_NAME = Aperture_full_test.DEVICE_NAME
FOR XML PATH(''), TYPE
).value('.', 'VARCHAR(MAX)'), 1, 2, '') AS Appid
FROM ( SELECT DEVICE_NAME, COUNT(DEVICE_NAME) AS NoOfApps
FROM dbo.Aperture_full_test
GROUP BY DEVICE_NAME
) Aperture_full_test
ORDER BY NoOfApps DESC
Data sample:
USE tempdb;
GO
IF OBJECT_ID('dbo.Aperture_full_test') IS NOT NULL
DROP TABLE dbo.Aperture_full_test;
GO
CREATE TABLE dbo.Aperture_full_test
(
DEVICE_NAME varchar(30) NOT NULL,
APP_ID varchar(10) NOT NULL
);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('LDNSQLF700', 157848);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('LDNSQLF700', 155439);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('LDNSQLF700', 635533);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('NYSQL502', 189164);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('NYSQL502', 188641);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('AUSSQL140', 537990);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('AUSSQL140', 1349605);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('JAP543X2', 5646789);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('EU456CLX', 6545789);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('EUCTX654', 5637965);
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('EUCTX654', 6464367) ;
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('EUCTX654', 1323123) ;
INSERT INTO dbo.Aperture_full_test(DEVICE_NAME, APP_ID)
VALUES('EUCTX654', 1004326) ;
GO

Answer 1

Since you are using SQL Server you can implement the PIVOT function.
If you have a known number of values then you can hard code the query using the following:
select device_name, App1, App2, App3, App4, App5
from
(
select device_name, app_id,
'App'+
cast(row_number() over(partition by device_name
order by device_name) as varchar(10)) col
from Aperture_full_test
) d
pivot
(
max(app_id)
for col in (App1, App2, App3, App4, App5)
) piv;
See SQL Fiddle with Demo.
But if you are going to have an unknown number of app_ids for each device, then you can use dynamic SQL:
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX)
select #cols = STUFF((SELECT ', ' + QUOTENAME('App'+cast(rn as varchar(10)))
from
(
select row_number() over(partition by device_name
order by device_name) rn
from Aperture_full_test
) d
group by rn
order by rn
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT device_name, ' + #cols + ' from
(
select device_name, app_id,
''App''+
cast(row_number() over(partition by device_name
order by device_name) as varchar(10)) col
from Aperture_full_test
) x
pivot
(
max(app_id)
for col in (' + #cols + ')
) p '
execute(#query)
See SQL Fiddle with Demo. Both give the result:
| DEVICE_NAME | APP1 | APP2 | APP3 | APP4 |
-------------------------------------------------------
| AUSSQL140 | 537990 | 1349605 | (null) | (null) |
| EU456CLX | 6545789 | (null) | (null) | (null) |
| EUCTX654 | 5637965 | 6464367 | 1323123 | 1004326 |
| JAP543X2 | 5646789 | (null) | (null) | (null) |
| LDNSQLF700 | 157848 | 155439 | 635533 | (null) |
| NYSQL502 | 189164 | 188641 | (null) | (null) |
Edit, If you want to count the total number of devices for each server, then you can use count() over(). The hard-coded version will be:
select device_name, TotalDevices, App1, App2, App3, App4, App5
from
(
select device_name, app_id,
'App'+
cast(row_number() over(partition by device_name
order by device_name) as varchar(10)) col,
count(app_id) over(partition by device_name) TotalDevices -- add this line
from Aperture_full_test
) d
pivot
(
max(app_id)
for col in (App1, App2, App3, App4, App5)
) piv;
See SQL Fiddle with Demo

Answer 2

This is a pivot query, which can be done in the group by. What you need is a sequence number for the Apps.
select aft.deviceName, COUNT(*) as NumApps,
MAX(case when seqnum = 1 then App_id end) as App1,
MAX(case when seqnum = 2 then App_id end) as App2,
MAX(case when seqnum = 3 then App_id end) as App3,
MAX(case when seqnum = 4 then App_id end) as App4,
MAX(case when seqnum = 5 then App_id end) as App5
from (select aft.*,
ROW_NUMBER() over (partition by device_name order by (select NULL)) as seqnum
from Aperture_full_test aft
) aft
group by aft.deviceName
order by NumApps desc
This version does not have the Apps in a particular order becasue the xml code did not have them in a particular order.