SQL Server split a single column multiple times - sql

Question

I have a database table that has a column with stacked data with two levels with a column that I want to break a part. Here is the example of the data (data changed to protect the innocent :) :
Table
ID = varchar(100)
CarData = varchar(1000)
ID CarData
1 Nissan:blue:20000,Ford:green:10000
2 Nissan:steel:20001,Ford:blue:10001,Chevy:blue:10000,Ford:olive:10000
** Note that cardata can is not fixed, and can have many cars in it
Output Desired:
ID Manufacture Color Cost
1 Nissan Blue 20000
1 Ford green 10000
2 Nissan steel 20001
... and on
So to say it plainly I need to break the first stacked field which is a comma and create a row for that, then break the second stacked field which is a colon into columns.
Any help would be greatly appreciated.

Answer 1

-- Sample data
declare #T table(ID int, CarData varchar(100))
insert into #T values
(1, 'Nissan:blue:20000,Ford:green:10000'),
(2, 'Nissan:steel:20001,Ford:blue:10001,Chevy:blue:10000,Ford:olive:10000')
-- Recursice CTE to get one row for each car
;with cte(ID, Car, CarData) as
(
select ID,
cast(substring(CarData+',', 1, charindex(',', CarData+',')-1) as varchar(100)),
stuff(CarData, 1, charindex(',', CarData), '')+','
from #T
where len(CarData) > 0
union all
select ID,
cast(substring(CarData, 1, charindex(',', CarData)-1) as varchar(100)),
stuff(CarData, 1, charindex(',', CarData), '')
from cte
where len(CarData) > 0
)
-- Use parsename to split the car data
select ID,
parsename(replace(Car, ':', '.'), 3) as Manufacture,
parsename(replace(Car, ':', '.'), 2) as Color,
parsename(replace(Car, ':', '.'), 1) as Cost
from cte
order by ID
Result:
ID Manufacture Color Cost
-- ----------- ------ -----
1 Nissan blue 20000
1 Ford green 10000
2 Nissan steel 20001
2 Ford blue 10001
2 Chevy blue 10000
2 Ford olive 10000
Edit 1
You will have trouble with parsename if color, cost or a manufacturer name contains a .. If that is the case you should try this instead.
-- Sample data
declare #T table(ID int, CarData varchar(100))
insert into #T values
(1, 'Nissan:blue:20000,Ford:green:10000'),
(2, 'Nissan:steel:20001,Ford:blue:10001,Chevy:blue:10000,Ford:olive:10000')
-- Recursice CTE to get one row for each car
;with cte(ID, Car, CarData) as
(
select ID,
cast(substring(CarData+',', 1, charindex(',', CarData+',')-1) as varchar(100)),
stuff(CarData, 1, charindex(',', CarData), '')+','
from #T
where len(CarData) > 0
union all
select ID,
cast(substring(CarData, 1, charindex(',', CarData)-1) as varchar(100)),
stuff(CarData, 1, charindex(',', CarData), '')
from cte
where len(CarData) > 0
)
-- Split the car data with substring
select ID,
substring(Car, 1, P1.Pos-1) as Manufacture,
substring(Car, P1.Pos+1, P2.Pos-P1.Pos-1) as Color,
substring(Car, P2.Pos+1, len(Car)-P2.Pos) as Cost
from cte
cross apply (select charindex(':', Car)) as P1(Pos)
cross apply (select charindex(':', Car, P1.Pos+1)) as P2(Pos)
order by ID

Answer 2

Use this string splitting function to produce a table of results.
I would first call dbo.split() using a , as the separator character. Then you'll have a list of items like:
Nissan:blue:20000
Ford:green:10000
Nissan:steel:20001
Ford:blue:10001
Chevy:blue:10000
Ford:olive:10000
From there you can call dbo.split() again using : as your separator. Each call will result in exactly three records (assuming your design as at least that "normal").
As #JNK mentioned in his comment, hopefully this is not something you'd want to be running regularly.
EDIT:
Some sample code to get you started:
SELECT *
INTO #YuckyCar
FROM (
SELECT 1 ID, 'Nissan:blue:20000,Ford:green:10000' CarData
UNION
SELECT 2, 'Nissan:steel:20001,Ford:blue:10001,Chevy:blue:10000,Ford:olive:10000'
) T;
-- Shows logical step #1
SELECT ID, X.items MoreCarData
FROM #YuckyCar CROSS APPLY dbo.Split(CarData, ',') X;
-- Shows logical step #2
SELECT Q.ID, Y.items
FROM (
SELECT ID, X.items MoreCarData
FROM #YuckyCar CROSS APPLY dbo.Split(CarData, ',') X) Q CROSS APPLY dbo.Split(Q.MoreCarData, ':') Y
DROP TABLE #YuckyCar;
The problem in the last part is that you can't guarantee row 1 = Manufacturer, row 2 = Color, row 3 = Cost.

Answer 3

This should solve your problem:
[EDIT] Your ID is a varchar(100) and you do not specify if it is a primary key, so I made some changes ... ID does't have to be primary key in this case.
declare #T table(ID varchar(100), CarData varchar(1000))
declare #OUT table(pk INT IDENTITY(1,1), ID varchar(100), Manufacture varchar(100), Color VARCHAR(100), Cost INT)
insert into #T (ID, CarData) values
('1', 'Nissan:blue:20000,Ford:green:10000'),
('2', 'Nissan:steel:20001,Ford:blue:10001,Chevy:blue:10000,Ford:olive:10000')
DECLARE #x XML, #i INT, #ID VARCHAR(100), #maxi INT;
;WITH list AS (SELECT pk = ROW_NUMBER() OVER(ORDER BY ID), * FROM #T)
SELECT #i=1, #maxi=MAX(pk) FROM list;
WHILE #i <= #maxi
BEGIN
;WITH list AS (SELECT pk = ROW_NUMBER() OVER(ORDER BY ID), * FROM #T)
SELECT
#x = CAST( '<root><car><prop>' +
REPLACE(
REPLACE(
CarData
,':'
,'</prop><prop>'
)
,','
,'</prop></car><car><prop>'
) +
'</prop></car></root>'
AS XML)
, #ID = ID
FROM list
WHERE pk = #i
INSERT INTO #OUT
SELECT
ID = #ID
,Manufacture = x.value('./prop[1]','VARCHAR(100)')
,Color = x.value('./prop[2]','VARCHAR(100)')
,Cost = x.value('./prop[3]','INT')
FROM #x.nodes('/root/car') AS T(x)
SET #i = #i + 1;
END
SELECT * FROM #OUT
/* -- OUTPUT
ID Manufacture Color Cost
--------------------------------
1 Nissan blue 20000
1 Ford green 10000
2 Nissan steel 20001
2 Ford blue 10001
2 Chevy blue 10000
2 Ford olive 10000
*/