MS BI TOOLS

MICROSOFT BUSINESS INTELLIGENCE

Saturday, February 26, 2011

Pivots with Dynamic Columns in SQL Server 2005


Pivots in SQL Server 2005 can rotate a table, i.e. they can turn rows into columns. PIVOTs are frequently used in reports, and they are reasonably simple to work with. However, I've seen quite a few questions about this operator. Most questions were about the column list in the PIVOT statement. This list is fixed, but many times the new columns are determined by the report at a later stage. This problem is easily solved when we mix pivots with dynamic SQL, so here is a very simple example about how to dynamically generate the pivot statement:
PIVOT allows you to turn data rows into columns. For example, if you have a table like this (I use only three months here for simplicity):
CREATE TABLE Sales ([Month] VARCHAR(20) ,SaleAmount INT)

INSERT INTO Sales VALUES ('January', 100)
INSERT INTO Sales VALUES ('February', 200)
INSERT INTO Sales VALUES ('March', 300)

SELECT FROM SALES
 

Month             SaleAmount
----------------  -----------
January           100
February          200
March             300 

Suppose we wanted to convert the above into this:
 
January     February    March
----------- ----------  ----------
100         200         300

We can do this using the PIVOT operator, as follows:
SELECT  [January]
      [February]
      [March]
FROM    SELECT    [Month]
                  SaleAmount
          FROM      Sales
        p PIVOT SUM(SaleAmount)
                    FOR [Month] 
                      IN ([January],[February],[March])
                  ) AS pvt

However, in the above example, I have the column names fixed as the first three months. If I want to create a result in which the columns are dynamic (for example, they are read from a table), then I need to construct the above query dynamically. To demonstrate this let’s look at the following example:
In the first table I have the column names I want to use:
CREATE TABLE Table1 (ColId INT,ColName VARCHAR(10))
INSERT INTO Table1 VALUES(1, 'Country')
INSERT INTO Table1 VALUES(2, 'Month')
INSERT INTO Table1 VALUES(3, 'Day')

In the second table I have the data. This consists of a row identifier (tID), a column ID (ColID) that refers to the column type in Table1, and a value:
CREATE TABLE Table2 (tID INT,ColID INT,Txt VARCHAR(10))

INSERT INTO Table2 VALUES (1,1, 'US')
INSERT INTO Table2 VALUES (1,2, 'July')
INSERT INTO Table2 VALUES (1,3, '4')
INSERT INTO Table2 VALUES (2,1, 'US')
INSERT INTO Table2 VALUES (2,2, 'Sep')
INSERT INTO Table2 VALUES (2,3, '11')
INSERT INTO Table2 VALUES (3,1, 'US')
INSERT INTO Table2 VALUES (3,2, 'Dec')
INSERT INTO Table2 VALUES (3,3, '25')

Now I would like to retrieve data from these two tables, in the following format:

tID         Country    Day        Month
----------- ---------- ---------- ----------
1           US         4          July
2           US         11         Sep
3           US         25         Dec 

In other words I want to turn the data rows in Table2 into columns. If I had a fixed set of columns for the result, i.e. the columns Country, Day, and Month were fixed, I could use SQL Server 2005’s PIVOT operator in a query like:
SELECT  tID
      [Country]
      [Day]
      [Month]
FROM    SELECT    t2.tID
                  t1.ColName
                  t2.Txt
          FROM      Table1 AS t1
                    JOIN Table2 
                       AS t2 ON t1.ColId t2.ColID
        p PIVOT MAX([Txt])
                    FOR ColName IN [Country][Day],
                                     [Month] ) ) AS pvt
ORDER BY tID ;

However I need to construct this query dynamically, because the column names Country, Day, and Month are specified in a table, and can be changed independently from my query. In our case these columns are given in Table1.
In the first step to generate the final pivot query I need to create the list of columns, in this case [Country], [Day], [Month].
Since there is no string concatenation aggregator in SQL (a concatenation aggregator would not be deterministic without some order restriction), and since the column names are stored in rows of a table, I need to flatten these columns into a single row or variable. There are various solutions to achieve this. One solution would be to use a query like:
DECLARE @cols NVARCHAR(2000)
SELECT  @cols COALESCE(@cols ',[' colName ']',
                         '[' colName ']')
FROM    Table1
ORDER BY colName

This query works both on SQL Server 2000 and 2005. It is efficient, but some may not like it because it uses the same variable (@cols) on both sides of an assignment. Another solution that works on SQL Server 2005 only is to use XML PATH.
DECLARE @cols NVARCHAR(2000)
SELECT  @cols STUFF(( SELECT DISTINCT TOP 100 PERCENT
                                '],[' t2.ColName
                        FROM    Table1 AS t2
                        ORDER BY '],[' t2.ColName
                        FOR XML PATH('')
                      ), 12'') + ']'

This second query (I’ve seen this posted by Peter Larsson) has the advantage that it does not use the @cols variable on the right hand side. I like this solution more, since this can be extended as a general string concatenation aggregate in more complex queries.
Both of the above queries generate, from Table1, the string: ‘[Country],[Day], [Month]’. This column list is used twice in the pivot query that we aim to construct. Once it is use in the list of columns that we want to retrieve, and once it is used as the list of values that should become columns. Having constructed this list of columns above, we can just concatenate it with the missing parts of the pivot query like:
DECLARE @query NVARCHAR(4000)
SET @query N'SELECT tID, '+
@cols +'
FROM
(SELECT  t2.tID
      , t1.ColName
      , t2.Txt
FROM    Table1 AS t1
        JOIN Table2 AS t2 ON t1.ColId = t2.ColID) p
PIVOT
(
MAX([Txt])
FOR ColName IN
( '+
@cols +' )
) AS pvt
ORDER BY tID;'
Executing this with
EXECUTE(@query)
will give us the expected result: a table that is pivoted and shows columns that were specified in a table:
tID         Country    Day        Month
----------- ---------- ---------- ----------
1           US         4          July
2           US         11         Sep
3           US         25         Dec

5 comments:

  1. Hey very nice blog!!
    Hi there,I enjoy reading through your article post, I wanted to write a little comment to support you and wish you a good continuationAll the best for all your blogging efforts.
    Appreciate the recommendation! Let me try it out.
    Keep working ,great job!
    Hadoop training

    ReplyDelete
  2. Superb. I really enjoyed very much with this article here. Really it is an amazing article I had ever read. I hope it will help a lot for all. Thank you so much for this amazing posts and please keep update like this excellent article.thank you for sharing such a great blog with us. expecting for your.
    Digital Marketing Company in India
    seo Company in India

    ReplyDelete
  3. Great Article… I love to read your articles because your writing style is too good, its is very very helpful for all of us and I never get bored while reading your article because, they are becomes a more and more interesting from the starting lines until the end.

    Hadoop Training in Chennai
    Dotnet Training in Chennai

    ReplyDelete

  4. Wonderful blog.. Thanks for sharing informative Post. Its very useful to me.

    Installment loans
    Payday loans
    Title loans

    ReplyDelete

  5. Its a wonderful post and very helpful, thanks for all this information. You are including better information regarding this topic in an effective way.Thank you so much

    Personal Installment Loans
    Payday Cash Advance loan
    Title Car loan
    Cash Advance Loan

    ReplyDelete