Excluding blank rows in CTE query - sql

Question

I am new to using CTE queries in SQL Server. I've built this query with help from the web in an effort to start building my "change log" to highlight changes made in my database. Please see example linked below. I'd like to exclude rows where there are no changes. Can you assist in how to accomplish this?
Row #3 with Nov 7 changedate has blank values. I would like for this row to not display. I also don't want to have to do something like WHERE row1 <> '' AND row2 <> '' AND row3 <> '', etc, because my final query will contain much much more rows. Is this possible?
http://sqlfiddle.com/#!6/134bd/4/0

Answer 1

Here is an option that you can use.
Below uses COALESCE function.
Using the same data and just modified your code from sqlfiddle.
The CASE statement to return NULL when match and cast data types to varchar, then use COALESCE in the where
Below modified script to include nTEXT column. You can use DATALENGTH with COALESCE in the WHERE clause.
Modified:
CREATE TABLE tblEmp
([memid] int, [empid] int, [name] varchar(50),[salary] int, [room] varchar(50), changedate datetime, ntxt ntext);
INSERT INTO tblEmp
([memid], [empid], [name], [salary], [room], [changedate], [ntxt])
VALUES
(41, 1, 'peter', 1000, 'Regency', '11/4/2012', ''),
(43, 1, 'peterz', 2000, 'Regency','11/5/2013', 'nn') ,
(44, 1, 'peterz', 2000, 'Regency','11/7/2013', '') ,
(45, 4, 'sally', 2001, 'Sheratio','11/2/2013', '') ,
(46, 4, 'sally', 2000, 'Sheraton','11/6/2013', ''),
(47, 1, 'peter', 3000, 'Regency','12/4/2013', '') ,
(48, 4, 'sallye', 2000,'Sheraton 1', '11/9/2013', '') ,
(49, 4, 'sally', 3000, 'Sheraton','11/6/2013', 'kljslkdjflkajslkjasdlkjalskjdlakjsdlkjasldjfk')
;
WITH cte AS
(
SELECT
empid,
name,
salary, room,
changedate,
ntxt,
rn=ROW_NUMBER()OVER(PARTITION BY empid ORDER BY changedate)
FROM tblemp
)
SELECT *
FROM
(
SELECT c1.empid, oldname=CASE WHEN c1.Name=c2.Name THEN NULL ELSE C1.Name END,
newname=CASE WHEN c1.Name=c2.Name THEN NULL ELSE C2.Name END,
oldsalary=CASE WHEN c1.salary=c2.salary THEN NULL ELSE C1.salary END,
newsalary=CASE WHEN c1.salary=c2.salary THEN NULL ELSE C2.salary END,
oldroom=CASE WHEN c1.Room=c2.Room THEN NULL ELSE C1.Room END,
newroom=CASE WHEN c1.room=c2.room THEN NULL ELSE C2.room END,
c2.changedate
, c2.ntxt
FROM cte c1 INNER JOIN cte c2
ON c1.empid=c2.empid AND c2.RN=c1.RN+1
) x
WHERE NOT (COALESCE(oldname, newname, CAST(oldsalary AS VARCHAR), CAST(newsalary AS VARCHAR), CAST(oldroom AS VARCHAR), CAST(newroom AS VARCHAR)) is null
AND DATALENGTH(ntxt) = 0)
ORDER BY ChangeDate DESC

Answer 2

A slightly different approach with SQL Server 2012 would be to use LAG in the common table expression to be able to detect row by row changes. The CTE basically pulls out each row along with the relevant data from the previous row, and does a straight forward compare in the outer query to generate the result.
WITH cte AS (
SELECT
empid, changeDate,
LAG(name) OVER (PARTITION BY empid ORDER BY changeDate) oldname, name,
LAG(salary) OVER (PARTITION BY empid ORDER BY changeDate) oldsalary, salary,
LAG(room) OVER (PARTITION BY empid ORDER BY changeDate) oldroom, room
FROM tblEmp
)
SELECT empid,
CASE WHEN name<>oldname THEN oldname ELSE '' END oldname,
CASE WHEN name<>oldname THEN name ELSE '' END newname,
CASE WHEN salary<>oldsalary THEN oldsalary ELSE '' END oldsalary,
CASE WHEN salary<>oldsalary THEN salary ELSE '' END newsalary,
CASE WHEN room<>oldroom THEN oldroom ELSE '' END oldroom,
CASE WHEN room<>oldroom THEN room ELSE '' END newroom,
changeDate
FROM cte
WHERE oldname<>name OR oldsalary<>salary OR oldroom<>room
ORDER BY empid, changeDate;
An SQLfiddle to test with.