Compare 2 rows in MS SQL and get the columns that differ - sql-server

Question

i have searched this one out for a while, but just don't know if there is a "silver bullet" solution to what I'm looking to do. I have a table in my DB (for the sake of this discussion the actual columns are irrelevant). I want to be able to look at 2 rows from the same table and get a list of columns that are different between the 2.
I know I could write a whole bunch of TSQL to make this happen for a specific table, but I was hoping there was a built in function in SQL Server (I'm running 2008 R2) that could do this.
I know there are simple functions like CHECKSUM that will tell me if 2 rows are different, but I need the specifics of which columns are different.
Preferably I would like to make it into a UDF and pass the table name, and primary keys of the 2 rows I want compared.
Any Thoughts or Suggestions?
--EDIT-- Here is the solution I came up with:
Well, It is certainly not the most elegant solution...but it will work in a pinch.
CREATE PROCEDURE sp_Compare_Table_Rows
(
#TablePK varchar(1000), -- The Name of the Primary Key in that Table
#TableName varchar(1000), -- The Name of the Table
#PK1 int, -- The ID of the 1st table
#PK2 int -- The ID of the 2nd table
)
AS
DECLARE #Holder table
(
Column_Name varchar(250),
Different bit
)
INSERT INTO #Holder(Column_Name,Different)
select
COLUMN_NAME,0
from
LPS_DEV.INFORMATION_SCHEMA.COLUMNS
WHERE
TABLE_NAME = #TableName
and ORDINAL_POSITION >1
DECLARE #LoopedColumnName varchar(250)
DECLARE #DynamicQuery nvarchar(max)
DECLARE #ORD1 int
DECLARE #ORD2 int
SET #DynamicQuery = ''
SET #LoopedColumnName = ''
SET #ORD1 = 0
SET #ORD2 = 0
DECLARE MYCUR CURSOR FOR SELECT Column_Name FROM #Holder
OPEN MYCUR
FETCH NEXT FROM MYCUR INTO #LoopedColumnName
WHILE ##FETCH_STATUS = 0
BEGIN
SET #DynamicQuery = 'SELECT #Outer= CHECKSUM(' + #LoopedColumnName + ') FROM ' + #TableName + ' WHERE ' + #TablePK + ' = ' + CONVERT(varchar(100),#PK1)
exec sp_executesql #DynamicQuery, N'#Outer int output',#ORD1 out
SET #DynamicQuery = 'SELECT #Outer= CHECKSUM(' + #LoopedColumnName + ') FROM ' + #TableName + ' WHERE ' + #TablePK + ' = ' + CONVERT(varchar(100),#PK2)
exec sp_executesql #DynamicQuery, N'#Outer int output',#ORD2 out
IF #ORD1 <> #ORD2
BEGIN
UPDATE #Holder SET Different = 1 WHERE Column_Name = #LoopedColumnName
END
FETCH NEXT FROM MYCUR INTO #LoopedColumnName
END
CLOSE MYCUR
DEALLOCATE MYCUR
select * from #Holder

Answer 1

John, you can use something like this. This will display the column names -
Set the #tablename and #whereclause.
declare #tablename varchar(1000),
#cols varchar(max),
#sqlstmt nvarchar(max),
#whereclause nvarchar(max);
set #tablename = 'sometable';
set #whereclause = ' where
a.col1 = ''SOMEOTHERVALUE'' and a.datecol2 = ''19910101'' and
b.col2 = ''SOMEVALUE'' and b.datecol2 = ''19910101''
'
select #cols = stuff((
select ', case when a.' + c.name + ' = b.' + c.name
+ ' then '''' else ''' + c.name + ''' end'
from sys.columns c
inner join sys.tables t on c.object_id = t.object_id
where t.name = #tablename
for xml path ('')), 1, 1, '')
set #sqlstmt = 'select ' + #cols + ' from ' + #tablename
+ ' a, ' + #tablename + ' b ' + #whereclause
exec sp_executesql #sqlstmt

Answer 2

I did some enhancements on Sriram answer, here is the new script:
declare #tablename nvarchar(MAX),
#cols varchar(max) = '',
#sqlstmt nvarchar(max) = 'DECLARE #T AS TABLE (ColumnName NVARCHAR(MAX), Value NVARCHAR(MAX), Value2 NVARCHAR(MAX))',
#whereclause nvarchar(max) = '';
set #tablename = 'TABLE_NAME';
set #whereclause = ' where a.FIRST_ROW_ID = ''NUMBER'' and b.SECOND_ROW_ID = ''NUMBER'''
select #cols = (
select ' INSERT INTO #T SELECT ''' + c.name + ''', cast(a.' + c.name + ' as nvarchar(max)), cast(b.' + c.name + ' as nvarchar(max)) from ' + #tablename + ' a, ' + #tablename + ' b ' + #whereclause + ' AND cast(a.' + c.name + ' as nvarchar(max)) != cast(b.' + c.name + ' as nvarchar(max))'
from sys.columns c
inner join sys.tables t on c.object_id = t.object_id
where t.name = #tablename
for xml path (''))
set #sqlstmt += #cols + ' SELECT * FROM #T'
exec sp_executesql #sqlstmt
You need to replace 'TABLE_NAME' with your table and change the where clause variable.
The result will be like this:
ColumnName | Value | Value2
----------------------------
Id | 1 | 2
Name | Name 1 | Name 2

Answer 3

I have done further enhancement to Sriram Chitturi and Mohamed Noor's answer and also handled datetime comparison
ALTER PROC [dbo].[IdentifyNotMatchingColumnDataBetween2Rows] (
#table_name VARCHAR(300) = 'TableName'
,#excluded_columns VARCHAR(300) = 'Id,UpdateTimeStamp,InsertTimeStamp,IsDuplicate,NotMatchingColumns'
,#compare_row_id_1 INT = 1732340
,#compare_row_id_2 INT = 1736803
)
AS
/*
EXEC [dbo].[IdentifyNotMatchingColumnDataBetween2Rows]
*/
BEGIN
SET NOCOUNT ON;
DECLARE #cols VARCHAR(max) = ''
DECLARE #sqlstmt NVARCHAR(max) = 'DECLARE #T AS TABLE (ColumnName VARCHAR(300), Value NVARCHAR(MAX), Value2 NVARCHAR(MAX))'
DECLARE #where_clause NVARCHAR(max) = ' WHERE a.Id = ''' + CAST(#compare_row_id_1 as varchar(20)) + ''' and b.Id = ''' + CAST(#compare_row_id_2 as varchar(20)) + ''''
SELECT #cols = (
SELECT CASE
WHEN c.system_type_id = 106 --DECIMAL
OR c.system_type_id = 56 --INT
OR c.system_type_id = 48 --TINYINT
OR c.system_type_id = 52 --SMALLINT
OR c.system_type_id = 127 --BIGINT
OR c.system_type_id = 62 --FLOAT
OR c.system_type_id = 108 --NUMERIC
THEN ' INSERT INTO #T SELECT ''[' + c.name + ']'', CAST(a.[' + c.name + '] AS NVARCHAR(MAX)), CAST(b.[' + c.name + '] AS NVARCHAR(MAX)) FROM [' + #table_name + '] a WITH(NOLOCK) , [' + #table_name + '] b WITH(NOLOCK) ' + #where_clause + ' AND CAST(ISNULL(a.[' + c.name + '],' + '-1' + ') AS NVARCHAR(MAX)) != CAST(ISNULL(b.[' + c.name + '],' + '-1' + ') AS NVARCHAR(MAX))'
WHEN c.system_type_id IN (61,42) -- DATETIME / DATETIME2
THEN ' INSERT INTO #T SELECT ''[' + c.name + ']'', FORMAT(a.[' + c.name + '] ,''yyyyMMddHHmmssffff''), FORMAT(b.[' + c.name + '] ,''yyyyMMddHHmmssffff'') FROM [' + #table_name + '] a WITH(NOLOCK) , [' + #table_name + '] b WITH(NOLOCK) ' + #where_clause + ' AND FORMAT(ISNULL(a.[' + c.name + '],' + '''' + '1900-01-01' + '''' + ') ,''yyyyMMddHHmmssffff'') != FORMAT(ISNULL(b.[' + c.name + '],' + '''' + '1900-01-01' + '''' + ') ,''yyyyMMddHHmmssffff'')'
WHEN c.system_type_id = 58 --SMALLDATETIME
THEN ' INSERT INTO #T SELECT ''[' + c.name + ']'', FORMAT(a.[' + c.name + '] ,''yyyyMMddHHmmss''), FORMAT(b.[' + c.name + '] ,''yyyyMMddHHmmss'') FROM [' + #table_name + '] a WITH(NOLOCK) , [' + #table_name + '] b WITH(NOLOCK) ' + #where_clause + ' AND FORMAT(ISNULL(a.[' + c.name + '],' + '''' + '1900-01-01' + '''' + ') ,''yyyyMMddHHmmss'') != FORMAT(ISNULL(b.[' + c.name + '],' + '''' + '1900-01-01' + '''' + ') ,''yyyyMMddHHmmss'')'
ELSE
' INSERT INTO #T SELECT ''[' + c.name + ']'', CAST(a.[' + c.name + '] AS NVARCHAR(MAX)), CAST(b.[' + c.name + '] AS NVARCHAR(MAX)) FROM [' + #table_name + '] a WITH(NOLOCK) , [' + #table_name + '] b WITH(NOLOCK) ' + #where_clause + ' AND CAST(ISNULL(a.[' + c.name + '],' + '''' + '''' + ') AS NVARCHAR(MAX)) != CAST(ISNULL(b.[' + c.name + '],' + '''' + '''' + ') AS NVARCHAR(MAX))'
END
FROM sys.columns c
INNER JOIN sys.tables t ON c.object_id = t.object_id
WHERE t.name = #table_name
AND c.name NOT IN (
SELECT items
FROM dbo.Split(#excluded_columns, ',')
)
FOR XML path('')
)
SET #sqlstmt += #cols + ' SELECT * FROM #T';
--PRINT #sqlstmt;
EXEC sp_executesql #sqlstmt
END
To Execute
EXECUTE dbo.IdentifyNotMatchingColumnDataBetween2Rows 'TableName','Id,UpdateTimeStamp,InsertTimeStamp,Last Checked,IsDuplicate,NotMatchingColumns,',#row_id_1,#row_id_2;