Comparing two T-SQL tables for diffs

Question

I have two instances of the same database. The first db represents data from today, the second data from 6 months ago. I need to find differences for a subset of entries in a specific table.

For entries with ids that are in both tables, I'd like to find a way to view only the rows that aren't identical.

Any ideas?

Thanks

Solution 1

SELECT t1.id
FROM table1 t1
INNER JOIN table2 t2 ON t1.id = t2.id
WHERE ISNULL(t1.field1,'') <> ISNULL(t2.field1,'')
      OR ISNULL(t1.field2,'') <> ISNULL(t2.field2,'')
      OR ...

To produce long WHERE part you can use this function:

CREATE PROCEDURE compareTables
    @db1    NVARCHAR(100),
    @table1 NVARCHAR(100),
    @db2    NVARCHAR(100),
    @table2 NVARCHAR(100)
AS
BEGIN
    SET NOCOUNT ON;
    DECLARE @where NVARCHAR(MAX)
    DECLARE @cmd NVARCHAR(MAX)

    SET @where = ''

    SELECT @where = @where + 'ISNULL(t1.' + name + ','''') <> ISNULL(t2.' + name + ','''') OR ' 
    FROM sys.columns WHERE object_id = OBJECT_ID(@table1)

    SET @where = SUBSTRING(@where,1,LEN(@where)-3)

    SET @cmd = 'SELECT t1.id FROM ' + @db1 + '.' + @table1 + ' t1 '
    SET @cmd = @cmd + 'INNER JOIN ' + @db2 + '.' + @table2 + ' t2 ON t1.id = t2.id '
    SET @cmd = @cmd + 'WHERE ' + @where

    EXEC sp_executesql @cmd
END
GO

Example usage:

EXEC compareTables 'db1_name','dbo.table1','db2_name','dbo.table1'

Remember to put schema in the table name.

Solution 2

From your text, I think you say there is an ID that can be used for matching rows.

SELECT t1.*, t2.*
FROM table1 t1 JOIN table2 t2 ON t1.id=t2.id 
WHERE BINARY_CHECKSUM(t1.*) <> BINARY_CHECKSUM(t2.*)

Untested, but should work.

Solution 3

A bit late, but:

SELECT *
 from Table1
except select *
 from Table2

will list all rows in Table1 that are not present in Table2,

SELECT *
 from Table2
except select *
 from Table1

will show all in Table2 that are not in Table1, and

SELECT *
 from Table1
intersect select *
 from Table2

will show all rows that are identical in both tables. If any colums are known to vary between tables, specify only those columns you need to compare.

Solution 4

You can use the TableDiff.exe utility that comes with SQL Server 2005 and above.

You can read more here.

Solution 5

for SQL Server 2005 and up try something like this:

declare @CurrentTable table (pk int, valuedata varchar(5))
declare @oldtable table (pk int, valuedata varchar(5))

insert into @CurrentTable values (1,'aa')
insert into @CurrentTable values(2,'bb')
insert into @CurrentTable values(3,'cc')
insert into @CurrentTable values(4,'dd')
insert into @CurrentTable values(5,'ee')

insert into @oldtable values(1,'aa')
insert into @oldtable values(1,'bb')
insert into @oldtable values(3,'zz')
insert into @oldtable values(7,'aa')
insert into @oldtable values(8,'qq')

select pk,valuedata  from @CurrentTable
except
select pk,valuedata from @oldtable

OUTPUT

pk          valuedata
----------- ---------
2           bb
3           cc
4           dd
5           ee

(4 row(s) affected)

if SQL Server 2000 or older try something like this:

SELECT
    c.*
    FROM YourTableCurrent             c
        LEFT OUTER JOIN YourTableOld  o ON c.id=o.id
    WHERE ISNULL(c.IntCol,-2147483648)!=ISNULL(o.IntCol,-2147483648)
        OR ISNULL(c.varcharCol,'||null||')!=ISNULL(o.varcharCol,'||null||')
        OR ISNULL(c.DatetimeCol,'01/01/1753')!=ISNULL(o.DatetimeCol,'01/01/1753')
        ....
UNION
SELECT
    o.*
    FROM YourTableOld                     o
        LEFT OUTER JOIN YourTableCurrent  c ON c.id=o.id
    WHERE c.id IS NULL
        ....

View more solutions

Comments

[srmark]

Yeah, I was afraid you'd say that.

[srmark]

THis is quite an impressive tool. I was however hoping for something that can be done within T-SQL code.

[cortijon]

Don't forget to test for NULL

[Lukasz Lysik]

Not good? More requirements? Just say, anything can be done here at SO ;-)

[srmark]

The tables are quite big so I assume that this won't perform very well. I was hoping for some kind of advanced function that could hash rows or something. This is however a usable solution and I thank you for it.

[Lukasz Lysik]

You can generate this big WHERE statement using Dynamic SQL. Give me few minutes, maybe I'll produce something.

[Lukasz Lysik]

I've added sp to generate statement for you.

[A-K]

It is almost correct, but not completely so. NULL and '' are different values, and your ISNULL(t1.field1,'') does not distinguish between them

[Andreas Jansson]

It's only fair to add that there's a number of data compare tools out there... (Red Gate's was not my choice). Here is a list updated a few years ago: mssqltips.com/sqlservertip/1069/sql-server-comparison-tools

[Andreas Jansson]

BINARY_CHECKSUM doesn't allow the asterisk "" symbol for all fields. I get "Incorrect syntax near ''" [asterisk] in 2008R2. It would have been a very nice comparison, if it had worked.

[Andreas Jansson]

Here is a suggestion to take not_nullable and numeric values into account (tables in the same datbase here, add db1 and db2 if needed): SELECT @where = (at)where + CASE WHEN is_nullable = 0 THEN 't1.' + name + ' <> t2.' + name + ' OR ' WHEN precision = 0 THEN 'ISNULL(t1.' + name + ','''') <> ISNULL(t2.' + name + ','''') OR ' ELSE 'ISNULL(t1.' + name + ',-999999) <> ISNULL(t2.' + name + ',-999999) OR ' END FROM sys.columns WHERE object_id = OBJECT_ID((at)table1) (I was not allowed to use more than one @ variable, since SO interpreted them as user names. Sorry for the (at):s.)

[Jonas Lincoln]

technet.microsoft.com/en-us/library/ms173784(v=sql.105).aspx says it's supported in 2008R2. Is there an other problem with the query perhaps?

[Andreas Jansson]

You are right: the asterisk is allowed, but the table names are obviously not not allowed. I guess the BINARY_CHECKSUM with (asterisk) generates a checksum for all of the fields, i.e. for the joined tables combined. See stackoverflow.com/questions/9038338/…