Display Rows as Columns - sql-server

Question

I have seen few similar questions asked on Stackoverflow with same subject, However this is not fitting into any of those.
I have 3 Tables Purchase, Hotel, Car Purchase to Hotel and Car is 1 to 0 or many relationship. MS SQL 2008 Server.
Purchase Table
pid bookingdate ...
1
2
3
Hotel Table
hid pid amount rooms location brand...
1 1
2 1
3 1
4 3
4 3
Car Table
cid pid make model ...
1 1
2 2
3 2
What I want is display the hotel data in columns
pid bookingdate cid make model hid1 amount1 rooms1 location1 brand1 hid2 amount2 rooms2 location2 brand2 hid3 amount3 rooms3 location3 brand3 hid4 amount4 rooms4 location4 brand4
If there is only one hotel for given purchase id, other columns should be null. If more than 4 hotel ignore the other hotels(5th 6th ect)
Please assume Car table can have 0 or 1 at this stage.
(Advance::>> If Car table can 0 to many (relationship to Purchase) 1.) take only first record to account 2.) get the sum of records.)
Purpose of this project is generate a csv file which can upload to a some third party product. Currently we have about 100 columns for Purchase and Car about 30 Columns for Hotel (130 all). With Hotel rows displaying in this way it will be 220 (100 + 30 x 4) columns. And about 100K rows. Performance is an issue as well.
I tried few things, but nothing successful not even close to getting stuck. This was the closest I got but I have 30 columns to duplicate not just a one, In-fact I have a feeling that PIVOT is not the way to go. I'm thinking using DENSE_RANK() or something like that
SELECT pid
, hid
, DENSE_RANK() OVER(ORDER BY pid)
FROM HOTEL
WHERE pid IN (
SELECT pid
FROM Hotel
WHERE pid IN (
SELECT pid
FROM Purchase
WHERE rdate BETWEEN '2014-04-01' AND '2014-12-31'
)
GROUP BY pid
HAVING COUNT(pid) > 1
)
Assuming then we might able to process hotels with more than 1 attached to Purchase first, using a cursor (ugly) or something like that

Answer 1

Okay I have found a solution. You are welcome to refine and add your thoughts. Dynamically name columns would be handy. (ie product1,product2,product3,product4)
Schema
create table Purchase
(
pid int,
purchasedate date,
currency varchar(10),
paymenttype varchar(10),
creditcard varchar(10),
creditcardtype varchar(10)
);
create table Hotel
(
hid int,
pid int,
product varchar(30),
country varchar(10),
city varchar(10),
rooms int,
starrating int
);
create table Car
(
cid int,
pid int,
product varchar(30),
country varchar(10),
city varchar(10),
cancel int,
starrating int
);
insert into Purchase values (1, '2015-01-15','AUD', 'CC','12345678','AMEX')
insert into Purchase values (2, '2015-01-15','AUD', 'CC','12345678','AMEX')
insert into Purchase values (3, '2015-01-15','AUD', 'CC','12345678','AMEX')
insert into Purchase values (4, '2015-01-15','AUD', 'CC','12345678','AMEX')
insert into Purchase values (5, '2015-01-15','AUD', 'CC','12345678','AMEX')
insert into Hotel values (1,1, 'Five for Two','Australia', 'Melbourne','1','3')
insert into Hotel values (2,1, 'Five for None','Australia', 'Sydney','1','3')
insert into Hotel values (3,1, 'Five for Five','Australia', 'Melbourne','1','3')
insert into Hotel values (4,1, 'Five for Two','Australia', 'Jamboora','1','3')
insert into Hotel values (5,2, 'Five for Three','Australia', 'Sydney','1','3')
insert into Hotel values (6,2, 'Five for Love','Australia', 'Cook','1','3')
insert into Hotel values (7,2, 'Five for Grease','Australia', 'Darwin','1','3')
insert into Hotel values (8,3, 'Love Me','Australia', 'Darwin','1','3')
insert into Hotel values (9,4, 'Live for Grease','Australia', 'Footscray','1','3')
insert into Hotel values (10,4, 'Love Grease','Australia', 'Officer','1','3')
insert into Car values (1,1, 'Love Grease','Australia', 'Officer','1','3')
insert into Car values (2,2, 'Love Grease','Australia', 'Cook','1','3')
insert into Car values (3,4, 'Live Grease','Australia', 'Jamboora','1','3')
-- For Advance insert into Car values (4,4, 'Cove Grease','Australia', 'Melbourne','1','3')
And the Code is
SELECT *, DENSE_RANK() OVER(PARTITION BY pid ORDER BY hid DESC) AS Ranking
INTO #TempHotel
FROM Hotel
SELECT P.pid, P.purchasedate, P.currency, P.paymenttype, P.creditcard, P.creditcardtype
,C.cid, C.product, C.country, C.city, C.cancel, C.starrating
,H1.hid, H1.product, H1.country, H1.city,H1.rooms,H1.starrating
,H2.hid, H2.product, H2.country, H2.city,H2.rooms,H2.starrating
,H3.hid, H3.product, H3.country, H3.city,H3.rooms,H3.starrating
,H4.hid, H4.product, H4.country, H4.city,H4.rooms,H4.starrating
FROM Purchase P
left join Car C on P.pid = C.pid
left join #TempHotel H1 on P.pid = H1.pid and H1.Ranking = 1
left join #TempHotel H2 on P.pid = H2.pid and H2.Ranking = 2
left join #TempHotel H3 on P.pid = H3.pid and H3.Ranking = 3
left join #TempHotel H4 on P.pid = H4.pid and H4.Ranking = 4
DROP Table #TempHotel
You can also find this in SQL fiddle Here.

Answer 2

If the column count never changes you could use Pivot / Unpivot
Here is a technet article which says it all: http://technet.microsoft.com/en-us/library/ms177410(v=sql.105).aspx
and a sample from the article.
-- Pivot table with one row and five columns
SELECT 'AverageCost' AS Cost_Sorted_By_Production_Days,
[0], [1], [2], [3], [4]
FROM
(SELECT DaysToManufacture, StandardCost
FROM Production.Product) AS SourceTable
PIVOT
(
AVG(StandardCost)
FOR DaysToManufacture IN ([0], [1], [2], [3], [4])
) AS PivotTable;