Question

I’ve unique rows with primary key in my emp table.

select * from EMP E;

Now I’ve created emp_backup table which takes backup of emp table like.

insert into emp_backup
select * from emp e
where not exists (select EMPNO from emp_backup E1 where E1.EMPNO=e.EMPNO);

the above query successfully copies all the rows from emp to emp_backup and it make sure when you again run above query it will not copy the existing rows from emp to emp_backup table in next run.

now my problem is when i update any record in emp table and try to run above query it gives me error of primary key violation which is excepted.

for example.

update  EMP
set JOB='worker'
where   EMPNO=14;

I update emp table after copying the record with empno 14 to emp_backup table.

when i run insert into emp_backup.... query i want this update change should refelect into emp_backup table.

How can i modify the above query so that it will copy the updated rows with existing primary key from emp to emp_backup.

I hope my question is clear, tell me how i can improve it.

Answer 1

It seems that when you created your emp_backup table, that you also created a table PK on the EMPNO table. This will prevent you from inserting the same emp record twice.

You need to rethink your ‘backup’ plan – do you want

1. a running audit of every version of every record change in the emp table
2. a mirror of the emp table exactly?

For 1:

• Drop the primary key on emp_backup table. This will allow duplicate emp records to be copied into it. If you need idempotence on the emp_backup table, you could create a new surrogate PK (e.g. emp_backup_id int identity(1,1), and possibly a new CURRENT_TIMESTAMP defaulted column to assist you in determining which is the latest

After dropping your PK, updated records can be ‘backed up’ as:

insert into emp_backup 
select * from emp e 
where exists 
  (select EMPNO 
   from emp_backup E1 
   where E1.EMPNO=e.EMPNO
   and 
   (e1.othercolumn1 <> e.othercolumn1
   or e1.othercolumn2 <> e.othercolumn2
   ...
   )
); . 

(Note that columns of certain types can’t be compared directly)

For 2, you would need to update existing records in your emp_bak table:

update eb
set eb.othercolumn1 = e.othercolumn1,
    eb.othercolumn2 = e.othercolumn2
from emp_backup eb
    inner join emp e on eb.empno = e.empno
where 
   eb.othercolumn1 <> e.othercolumn1
   or eb.othercolumn2 <> e.othercolumn2
   ...
); . 

But your table mirroring begs the question, everytime you want a backup, why not just drop your existing emp_backup table and do:

insert into emp_backup 
select * from emp e 

??