Fetching changes for a given query

Question

Hi,
I have an application that uses a query to retrieve thousands of
rows.
Initially, the application fetches all the rows. After that, the application
has to fetch only the changed rows. This can be done by using a date
column that contains the Last Update Time and adding a date condition to
the query. Something like (pseudo code):
Select * from myTbl t
where t.name like '%john%'
and lastUpdateTime > current_date - (5 seconds)
In the solution above, the application gets the updated rows (in the last 5
seconds). But the application does not know which rows have been deleted
and which rows don't meet the condition.
In the given example, initially, the application has all the rows, then, when
trying to get the changes, the application don't get a row that its "name"
has been changed to 'foo'. Thus the app still displays this rows and doesn't
know that this row is not valid due to the name change (The row doesn't
meet the query condition).
Do you have any idea how to get correct changes for a given query periodically?
Thanks
Dyahav

Answer 1

You either have to audit your data e.g. with triggers, so that whatever changes happen are logged and you can query that record of changes to determine what happened or, you can keep a copy of the data at the previous time of querying and compare that to the current set of data to determine what has changed. e.g. if you have your "historic" data and your "current" data you would do..
select * from historic
minus
select * from currentwhich would give you deleted and updated records (let's call them "du")
then
select * from current
minus
select * from historicwhich would give you inserted and updated records (let's call them "iu")
then
select +primary_key_cols+ from iu
intersect
select +primary_key_cols+ from duwhich would give you the primary keys of all the updated rows (let's call them "u")
then
select +primary_key_cols+ from iu
minus
select +primary_key_cols+ from uwhich will give you the primary keys of all the inserted rows
and finally..
select +primary_key_cols+ from du
minus
select +primary_key_cols+ from uwhich will give you the primary keys of all the deleted rows
So, then when you have all the primary keys for the "inserted", "deleted", "updated" rows you can identify them and process them as required.
If you actually want to know what columns have changed on updated rows then you're going to have to code it to check column by column (perhaps using a decode statement), as required.