left join of 3 tables and display like a dynamic pivot - sql

Question

i have 3 tables:
payorderType :
---------
typeID | TypeName |
1 | accounting |
2 | budget |
----------
step:
----------
StepID | StepName | typeID
1 | payorder | 1
2 | cheque | 1
3 | cheque | 2
----------
user:
----------
userID | StepName | StepID
7878 | payorder | 1
4547 | cheque | 2
6538 | cheque | 1
----------
I want to make a table which users exists in row and columns includes with concat of step and payorderType. same as below:
users | accounting_payorder | accounting_cheque | budget_cheque |
7878 | 1 | 0 | 0 |
4547 | 0 | 1 | 0 |
6538 | 0 | 1 | 0 |
My quesdtion is : if i don't know number of payorderType rows and number of step rows, how should i write it?
My Script is here :
First I create a table in cursor for concat payorderType and step:
CREATE PROC sp_payOrderType
AS
BEGIN
DECLARE a CURSOR
FOR SELECT DISTINCT p.TypeName,s.StepName
FROM
dbo.PayOrderType p LEFT JOIN
dbo.vStep s ON s.TypeID = p.TypeID
FOR READ ONLY
DECLARE #payOrderType NVARCHAR(50),#stepName NVARCHAR(50)
DECLARE #SQL NVARCHAR(MAX)=''
OPEN a
FETCH NEXT FROM a INTO #payOrderType, #stepName
WHILE ##FETCH_STATUS=0
BEGIN
DECLARE #b VARCHAR(max), #b2 VARCHAR(max)
SELECT #b = ISNULL(#b ,'') +'['+ ISNULL(#payOrderType ,'')+ '____'+ISNULL(#stepName ,'')+ ']'+ ' NVARCHAR(1000) ,'
FETCH NEXT FROM a INTO #payOrderType,#stepName
END
CLOSE a
DEALLOCATE a
SELECT #SQL = 'ALTER table AA(' + SUBSTRING(#b,1, LEN(#b)-1) + ')'
SELECT #SQL
END
but i don't know how i should relate rows(userID) with columns ?

Answer 1

You should have determined output structure. Its little risky to have variable output structure.
But here we go:
Make your structure simple (remove most of variables) - create view (or use derived table) payorderType + step
-- should be inner join probably instead of left join
-- if you use left join you have to isnull s.StepName
SELECT
u.userID,
s.StepID,
p.TypeName,
s.StepName,
p.TypeName + '_' + s.StepName StepType,
-- Your column can be like `coalesce(p.TypeName + '_' + s.StepName, p.TypeName, s.StepName) StepType` for left joins
1 Point
FROM dbo.PayOrderType p
INNER JOIN dbo.vStep s ON s.TypeID = p.TypeID
INNER JOIN dbo.user u ON u.StepID = s.StepID
Make your queries more clear (can all yours fields have null values?. Now you can work with only one column/variable.
Now is time for pivot:
SELECT
userID,
[accounting_payorder],
[accounting_cheque],
[budget_cheque]
FROM newview v
PIVOT(MAX(point) FOR StepType in ([accounting_payorder], [accounting_cheque], [budget_cheque])
If its necessary you can use dynamic query:
declare #header varchar(max),
#columns varchar(max)
select #header = coalesce(#header + ', ', '') + 'isnull(''' + StepType + ''', 0) ' + '[' + StepType + ']',
#columns = coalesce(#columns + ', ', '') + '[' + StepType + ']'
from newview
group by StepType
declare #sqlpvt varchar(4000) -- limited by lenght of exec statement
set #sqlpvt = 'select userID, $HEADER FROM newview v PIVOT(MAX(point) FOR StepType in ($COLUMNS)'
-- replace pseudovariables
set #sqlpvt = replace(#sqlpvt, '$HEADER', #header)
set #sqlpvt = replace(#sqlpvt, '$COLUMNS', #columns)
print #sqlpvt
exec (#sqlpvt)
Sorry if somethink is wrong (writed on blind), but i think for guide it's enough. But you should prefer end on step 2 (non-static code is dangerous).