Combing multiple rows into one row - sql

Question

I have the following table
Index BookNumber
2 51
2 52
2 53
1 41
1 42
1 43
I am trying to come up with the following output
Index BookNumber1 Booknumber2 Booknumber3
----------------------------------------------
1 41 42 43
2 51 52 53
I was able to come up with the following query , however the output is unexpected
SELECT DISTINCT
index,
CASE WHEN index = 1 THEN Booknumber END AS BookNumber1,
CASE WHEN index = 2 THEN Booknumber END AS BookNumber2,
CASE WHEN index = 3 THEN Booknumber END AS BookNumber3
FROM Mytable;
I get following output
Index BN1 BN2 BN3
------------------------------
1 41 null null
1 null 42 null
1 null null 43
2 51 null null
2 null 52 null
2 null null 53
Is there a way to compress this to only 2 rows?

Answer 1

I am not quite sure how the index in your query matches the index column in your data. But the query that you want is:
SELECT index,
max(CASE WHEN index = 1 THEN Booknumber END) AS BookNumber1 ,
max(CASE WHEN index = 2 THEN Booknumber END) AS BookNumber2,
max(CASE WHEN index = 3 THEN Booknumber END) AS BookNumber3
FROM Mytable
GROUP BY index;
Give your data, the query seems more like:
SELECT index,
max(CASE WHEN ind = 1 THEN Booknumber END) AS BookNumber1 ,
max(CASE WHEN ind = 2 THEN Booknumber END) AS BookNumber2,
max(CASE WHEN ind = 3 THEN Booknumber END) AS BookNumber3
FROM (select mt.*, row_number() over (partition by index order by BookNumber) as ind
from Mytable mt
) mt
GROUP BY index;
By the way, "index" is a reserved word, so I assume that it is just a placeholder for another column name. Otherwise, you need to escape it with double quotes or square braces.

Answer 2

Assuming there are always 3 or fewer book numbers for each index, you could use:
with data as
(select idx,
booknumber as bn1,
lag(booknumber, 1) over(partition by idx order by idx, booknumber) as bn2,
lag(booknumber, 2) over(partition by idx order by idx, booknumber) as bn3
from books)
select *
from data
where data.bn1 = (select max(x.bn1) from data x where x.idx = data.idx)
sqlfiddle demo is here: http://sqlfiddle.com/#!6/8dc82/5/0

Answer 3

Don't forget that index is a reserved word. Personally I prefer not to use reserved words as column names, but you can compensate by using square brackets like in my example.
This will work from sqlserver 2008+
declare #t table([Index] int, BookNumber int)
insert #t values
(2,51),(2,52),(2,53),(1,41),(1,42),(1,43)
;with cte as
(
select [Index], BookNumber,
row_number() over (partition by [Index] order by BookNumber) rn
from #t
)
select [Index], [1] as Booknumber1, [2] as Booknumber2, [3] as Booknumber3
from cte
pivot (max([booknumber]) FOR [rn] IN ([1],[2],[3])) AS pvt
Result:
Index Booknumber1 Booknumber2 Booknumber3
1 41 42 43
2 51 52 53