Transpose table and save as View - sql-server

Question

I have a dynamic pivot/unpivot script that transposes a table. This is dynamic enough to return certain columns that I want and using dynamic columns.
What I am looking for is rather to convert this into either a UDF or a VIEW so that I can join it to other tables.
Please help.
ALTER PROC [dbo].[uspGetUserByValues]
(
#Select NVARCHAR(4000) = '*',
#Where NVARCHAR(4000) = NULL,
#OrderBy NVARCHAR(4000) = NULL
)
AS
BEGIN
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX)
select #cols = STUFF((SELECT distinct ',P.' + QUOTENAME(PropertyDescription)
from System_Properties
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT ' + #cols + ', M.Email, C.Company_Name, C.Company_Type_ID, U.UserName, ISNULL(SMS.SMSProfiles,0) SMSProfiles, U.UserID
from
(
select PropertyDescription, UP.UserID, PropertyValue
from User_Properties UP
JOIN System_Properties SP ON UP.PropertyID = SP.PropertyID
JOIN aspnet_Membership M ON UP.UserID = M.UserID
) X
pivot
(
min(PropertyValue)
for PropertyDescription in (' + REPLACE(#cols,'P.','') + ')
) P
JOIN aspnet_Membership M ON P.UserID = M.UserID
JOIN aspnet_Users U on P.UserID = U.UserID
JOIN Companies C ON C.Company_ID = P.Company_ID
LEFT JOIN (SELECT UserId, COUNT(Users_SMS_Profile_ID) SMSProfiles
FROM Users_SMS_Profile GROUP BY UserID ) SMS ON SMS.UserID = P.UserID
'
SET #query = 'SELECT ' + #Select + ' FROM ('+ #query +') A'
IF ISNULL(#Where,'NULL') != 'NULL'
BEGIN
SET #query = #query + ' WHERE ' + #Where
END
IF ISNULL(#OrderBy,'NULL') != 'NULL'
BEGIN
SET #query = #query + ' ORDER BY ' + #OrderBy
END
execute(#query)
--PRINT(#query)
END

Answer 1

OH wow I made it.
I know this is with "known" column names but actually I didn't have to know them.
Firstly, this is the query I used to create the View. I will need to drop the view at least every I add a new Property or I can actually write a job that checks if all the properties from System_Properties are represented in the view, if not then drop the view and run this code.
CREATE PROC [dbo].[uspCreateViewUsers]
AS
BEGIN
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX)
select #cols = STUFF((SELECT distinct ',P.' + QUOTENAME(PropertyDescription)
from System_Properties
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'CREATE VIEW vwUsers AS SELECT ' + #cols + ', M.Email, C.Company_Name, C.Company_Type_ID, U.UserName, ISNULL(SMS.SMSProfiles,0) SMSProfiles, U.UserID
from
(
select PropertyDescription, UP.UserID, PropertyValue
from User_Properties UP
JOIN System_Properties SP ON UP.PropertyID = SP.PropertyID
JOIN aspnet_Membership M ON UP.UserID = M.UserID
) X
pivot
(
min(PropertyValue)
for PropertyDescription in (' + REPLACE(#cols,'P.','') + ')
) P
JOIN aspnet_Membership M ON P.UserID = M.UserID
JOIN aspnet_Users U on P.UserID = U.UserID
JOIN Companies C ON C.Company_ID = P.Company_ID
LEFT JOIN (SELECT UserId, COUNT(Users_SMS_Profile_ID) SMSProfiles
FROM Users_SMS_Profile GROUP BY UserID ) SMS ON SMS.UserID = P.UserID
'
execute(#query)
END
Them the View, which can't be represented graphically by table joins looks like this:
SELECT P.[Company_ID], P.[Created_Date], P.[Created_User], P.[Cust_ID], P.[FirstName], P.[IPCheck], P.[JobTitle], P.[LastLogin], P.[LastModified_Date], P.[LastModified_User],
P.[LastName], P.[Newsletter_OptIn], P.[Password_Change], P.[SupAdmin], P.[SysAccess], P.[SysAdmin], P.[User_Cat_1], P.[User_Cat_10], P.[User_Cat_2],
P.[User_Cat_3], P.[User_Cat_4], P.[User_Cat_5], P.[User_Cat_6], P.[User_Cat_7], P.[User_Cat_8], P.[User_Cat_9], P.[UserClient_ID], M.Email, C.Company_Name,
C.Company_Type_ID, U.UserName, ISNULL(SMS.SMSProfiles, 0) SMSProfiles, U.UserID
FROM (SELECT PropertyDescription, UP.UserID, PropertyValue
FROM User_Properties UP JOIN
System_Properties SP ON UP.PropertyID = SP.PropertyID JOIN
aspnet_Membership M ON UP.UserID = M.UserID) X PIVOT (min(PropertyValue) FOR PropertyDescription IN ([Company_ID], [Created_Date], [Created_User],
[Cust_ID], [FirstName], [IPCheck], [JobTitle], [LastLogin], [LastModified_Date], [LastModified_User], [LastName], [Newsletter_OptIn], [Password_Change], [SupAdmin],
[SysAccess], [SysAdmin], [User_Cat_1], [User_Cat_10], [User_Cat_2], [User_Cat_3], [User_Cat_4], [User_Cat_5], [User_Cat_6], [User_Cat_7], [User_Cat_8],
[User_Cat_9], [UserClient_ID])) P JOIN
aspnet_Membership M ON P.UserID = M.UserID JOIN
aspnet_Users U ON P.UserID = U.UserID JOIN
Companies C ON C.Company_ID = P.Company_ID LEFT JOIN
(SELECT UserId, COUNT(Users_SMS_Profile_ID) SMSProfiles
FROM Users_SMS_Profile
GROUP BY UserID) SMS ON SMS.UserID = P.UserID
This now allows me to query the View as if it was a table.
I hope this helps someone else in the future.

Answer 2

Simply said: you cant
At least you can't do it using conventional TSQL programming. Which means you would have to use some hack. Let me explain.
Closest thing to your SP would be UDF. However, UDFs are rather restricted. One thing UDF expect is for data to stay the same while and after executing it. Of course that this means EXEC() is forbidden in that scope.
Another possibility would be a view. However, you have a number of parameters and view's schema depends on these parameters. Functionality to change view's schema based on input parameters doesn't exist in SQL server.
And now for the hacks.
One hack I can think of is:
create a CLR UDF
create new connection based on context connection (same server, same db)
exec your SP there
return result to your original pipe
But it may or may not work (it's a hack after all).
If the hack doesn't work, you can try playing it by the book. This means creating a CLR UDF, assemble select statement in there and execute it, which means that you will have to throw away your original SP. However, it is not a hack since SQL CLR UDF's are made for such (and other) situations. Only thing you will have to take care about is using SqlMetaData because UDF doesn't have a predefined resultset. See this.
In my previous answer I stated that it could be done using CLR UDF, but that was wrong. One thing I forgot is that Microsoft insists on providing a finite number of columns for UDF. This may not be obvious while developing in .NET - after all, you can return any number of columns to SqlPipe. See this (untested) code...
[SqlFunction(DataAccess = DataAccessKind.Read)]
public static void DynOutFunc(SqlString select, SqlString where, SqlString orderBy)
{
// 1: Create SQL query
string query = "select db_id(), db_name()";
// 2: Find out which colums and their types are part of the output
SqlMetaData[] metaData =
{
new SqlMetaData("ID", System.Data.SqlDbType.Int),
new SqlMetaData("Database", System.Data.SqlDbType.NVarChar, 256)
};
using (SqlConnection connection = new SqlConnection("context connection=true"))
{
connection.Open();
SqlCommand command = new SqlCommand(query, connection);
SqlDataReader reader = command.ExecuteReader();
using (reader)
{
while (reader.Read())
{
SqlDataRecord record = new SqlDataRecord(metaData);
SqlContext.Pipe.SendResultsStart(record);
for(int i = 0; i < metaData.Length; i++)
{
if(metaData[i].DbType == DbType.String)
record.SetString(i, reader.GetString(i));
else if(metaData[i].DbType == DbType.Int32)
record.SetInt32(i, reader.GetInt32(i));
// else if's should cover all supported data taypes
}
SqlContext.Pipe.SendResultsRow(record);
}
}
SqlContext.Pipe.SendResultsEnd();
}
}
Notice SqlMetaData collection that holds information about columns. What stops you from appending just another column to it?
But(!) when it comes to registering that function in the SQL Server itself, you HAVE to provide arguments, like:
CREATE FUNCTION DynOutFunc
#select [nvarchar](4000),
#where [nvarchar](4000),
#orderBy [nvarchar](4000)
RETURNS TABLE (p1 type1, p2 type2, ... pN typeN)
AS EXTERNAL NAME SqlClrTest.UserDefinedFunctions.DynOutFunc;
It turns out there are no hacks for this I can think of. Or there are just no hacks here at all.