View with dynamic columns

Question

Hello.
I'm sure this must be a FAQ but can't find it anywhere.
I'm collecting data from people in a questionnaire style but don't know what/how many questions there are. I need to create a view that displays responses to each of the questions as a view, as follows:
Name | Company | Favourite Colour | Age Range | Favourite Food
In this example, there are three 'dynamic' questions. The data is stored as follows:
People - basic information about the respondents
Questions - the names of the questions being asked
Choices - the options that can be chosen from for each question
Responses - the answers that people provide.
If I know exactly what questions there are, then I can JOIN to the questions table the appropriate number of times. But I don't! Can this be done without dynamic SQL?
Very many thanks for help.
Simplified tables and data follow:
CREATE TABLE [People](
[ID] [int] NULL,
[Name] [varchar](50) NULL,
[Company] [varchar](50) NULL
) ON [PRIMARY]
GO
CREATE TABLE [Questions](
[ID] [int] NULL,
[QuestionName] [varchar](50) NULL
) ON [PRIMARY]
GO
CREATE TABLE [Choices](
[ID] [int] NULL,
[QuestionID] [int] NULL,
[ChoiceName] [varchar](50) NULL
) ON [PRIMARY]
GO
CREATE TABLE [Responses](
[ID] [int] NULL,
[PersonID] [int] NULL,
[QuestionID] [int] NULL,
[ChoiceID] [int] NULL,
[OtherText] [varchar](50) NULL
) ON [PRIMARY]
GO
INSERT [People] ([ID], [Name], [Company]) VALUES (1, N'Ben', N'ACME')
GO
INSERT [People] ([ID], [Name], [Company]) VALUES (2, N'Dave', N'My Corp')
GO
INSERT [People] ([ID], [Name], [Company]) VALUES (3, N'Sarah', N'Newco')
GO
INSERT [Questions] ([ID], [QuestionName]) VALUES (1, N'FavouriteColour')
GO
INSERT [Questions] ([ID], [QuestionName]) VALUES (2, N'AgeRange')
GO
INSERT [Questions] ([ID], [QuestionName]) VALUES (3, N'FavouriteFood')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (1, 1, N'Red')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (2, 1, N'Blue')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (3, 1, N'Green')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (4, 2, N'<18')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (5, 2, N'18-65')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (6, 2, N'65+')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (7, 3, N'Pasta')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (8, 3, N'Curry')
GO
INSERT [Choices] ([ID], [QuestionID], [ChoiceName]) VALUES (9, 3, N'Steak')
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (1, 1, 1, 1, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (2, 1, 2, 1, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (3, 1, 3, 2, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (4, 2, 1, 4, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (5, 2, 2, 5, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (6, 2, 3, 6, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (7, 3, 1, 7, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (8, 3, 2, 9, NULL)
GO
INSERT [Responses] ([ID], [PersonID], [QuestionID], [ChoiceID], [OtherText]) VALUES (9, 3, 3, 9, NULL)
GO

Answer 1

I'm not into reporting tools, but I recall that Kalman has mentioned some component in Reporting Services that can do this. And obviously, if you receive the result set as a DataTable in .NET, you can write code to flip it.
As I said, a view or a result set in a relational database is a table that is supposed to model an enitity with a finite number of distinct attributes. For this reason, any query where the number of columns is not known until run-time has to build dynamically.
And since a relational database is not designed for handling such dynamic scenarios, all attempt to work with that dynamically generated result set will be painful.
So for the filters you mention, there are two options:
1) Filter in the database on the base tables.
2) Filter in the client on the pivoted data.
But don't get the idea to filter in SQL Server after the pivot.
If the total dataset is large, the first alternative certainly has an advantage, but else it's likely to be easier to do client side.
Erland Sommarskog, SQL Server MVP, [email protected]