Dynamic SQL Unpivoting and re - Pivoting seems overly complicated...

  • Hi All,

    It is fairly common where we work to have a questionnaire output in the form of two tables, one that contains a set of questions and another that contains a matching set of answers, along with a UserId and Timestamp. In order to display these nicely, it is useful for us to display these two as one table, where the questions become the column names and the answers become the values, all in one row per questionnaire. I have written some code which does this fine (see the example below, where the output tables correspond to #Questions and #Answers), but it seems to me that this is a lot of code, a lot of dynamic SQL and generally pretty involved looking.

    Is there a simpler way to do the same thing? Do you have any ideas of things that MIGHT make it all simpler and (maybe) quicker?

    Cheers,

    Mark

    CREATE TABLE #Questions

    (

    s01q01VARCHAR(50)

    ,s01q02VARCHAR(50)

    ,s02q01VARCHAR(50)

    ,s02q02VARCHAR(50)

    ,s02q03VARCHAR(50)

    )

    CREATE TABLE #Answers

    (

    CompletingUserVARCHAR(50)

    ,CompletingTimeVARCHAR(50)

    ,s01q01VARCHAR(50)

    ,s01q02VARCHAR(50)

    ,s02q01VARCHAR(50)

    ,s02q02VARCHAR(50)

    ,s02q03VARCHAR(50)

    )

    INSERT #Questions

    SELECT'Question One','Question Two','Question One B','Question Two B','Question Three B'

    INSERT #Answers

    SELECT'Darth Vader',GETDATE(),'definitiely 1','There are Two','I started a new section','£22.00',NULL

    ---------------------------------------------------------------------------------------------------------------------------------------------

    -- Enter an empty string so that this is no longer NULL, which means we can easily add to it

    DECLARE @ColumnList VARCHAR(MAX) = ''

    DECLARE @SQL VARCHAR(MAX)

    CREATE TABLE #AnswerList (Label VARCHAR(10), Wording VARCHAR(MAX))

    CREATE TABLE #QuestionList (Label VARCHAR(10), Wording VARCHAR(MAX))

    -- The columnlist here is our input list of columns to Pivot about

    SELECT @ColumnList = @ColumnList + ',' + COLUMN_NAME

    FROM tempdb.INFORMATION_SCHEMA.COLUMNS

    WHERE TABLE_NAME LIKE '#Questions%'

    -- lose the extaneous comma from our concatenated list of columns

    SELECT @ColumnList = SUBSTRING(@ColumnList,2,LEN(@ColumnList))

    -- make our dynamic SQL with the column list in it and any changes will be incorporated into our query

    SET @SQL =

    -- First make the list of answers, with the labels in one column and the wording in the other

    'INSERT #AnswerList

    SELECT Label,Answer FROM

    (SELECT CompletingUser, CompletingTime,' + @ColumnList + ' FROM #Answers) AS PVT

    UNPIVOT (Answer FOR Label IN (' + @ColumnList +')) AS UnPvt

    ' +

    -- Now make a similar list of questions, with the same values in the labels row as the answers (this is always true)

    'INSERT #QuestionList

    SELECT Label,Question FROM

    (SELECT ' + @ColumnList + ' FROM #Questions) AS PVT

    UNPIVOT (Question FOR Label IN (' + @ColumnList +')) AS UnPvt'

    EXEC (@SQL)

    -- Make a new column list - this is now the output column list, taken as the list of the label of al the questions

    SET @ColumnList = ''

    -- Put it all nicely in square brackets

    SELECT @ColumnList = @ColumnList + ',[' + Wording + ']'

    FROM #QuestionList

    -- Lose the comma off the front

    SET @ColumnList = SUBSTRING(@ColumnList,2,LEN(@ColumnList))

    -- Now get the one-off info from the Answers and all the dynamic questions and answers

    SET @SQL =

    'SELECT

    (SELECT CompletingUser FROM #Answers) AS CompletingUser

    ,(SELECT CompletingTime FROM #Answers) AS CompletingTime

    ,*

    FROM (SELECT Q.Wording AS Question, A.Wording AS Answer

    FROM #QuestionListAS Q

    ' +

    -- Left outer join as some questions may be unanswered, so having NULL values

    'LEFT OUTER JOIN #AnswerListAS AON Q.Label = A.Label) AS QA

    ' +

    -- it doesn't know that we have only one row, so we have to put a MAX in, but it makes no different because we DO have only one row.

    'PIVOT(MAX([Answer]) FOR Question IN ('+ @ColumnList +')) as Pvt'

    EXEC (@SQL)

    DROP TABLE #Answers

    DROP TABLE #Questions

    DROP TABLE #AnswerList

    DROP TABLE #QuestionList

    All suggestions most welcome!

  • Hi,

    Rather than storing the questions as columns, store them as rows. ie.

    create table Questions(

    QuestionID int,

    Question varchar(200)

    )

    Also, makes it a lot simpler if you need to add a new question (because user's like to do that sort of thing!)

    Store the answers in a similar way

    create table Answers(

    AnswerID int,

    QuestionID int,

    Answer varchar(50),

    UserID int,

    DateAnswered datetime

    )

    By adding a 'QuestionaireID' to the Question table, you can also store more than one questionnaire in the same table, thus avoiding having to create new tables and procedures, when the users find another questionnaire that needs automating.

  • Hi,

    thanks for the answer - unfortunately I had hoped this could be avoided, as it means that I'll have to spend a very long time changing a lot of things and asking our software providers to change things as well (we don't really get a choice about the format they supply in, and we use the separate tables in SSRS, so one row results is good for that.).

    I guess we just have to say that the complexity is a necessary evil until we can get things changed in a larger way here.

    Thanks again, and to all that have read this!

    Mark

Viewing 3 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic. Login to reply