Compare two records column by column in same table

Question

Hi All,
I have an address table with ID as the PK but there can be more than one ID. I need to take the latest address record and first see if that same ID has a previous address record and if so, compare each column to see if the data has changed. I only need to compare the latest record with the next latest record.
I cannot figure this out at all. I can of course use the MAX function to get the latest address records but do not know how I can pull the next latest address record that matches the ID and compare each column.
Sample table:
ID street city state zip effective_date
1 123 main chicago IL 60111 3-7-2012
2 34 N 13th new york NY 18374 3-7-2012
3 15 N main Dallas TX 47389 3-7-2012
1 89 N main chicago IL 60111 1-5-2012
1 16 East St columbus OH 47382 12-10-2011
2 34 N 13th new york NY 18374 10-7-2011
2 15 S Elm new york NY 18374 09-1-2011
3 15 N main Dallas TX 47389 10-4-2011
SO...in the table above using today as the latest record date, based of my criteria I would only want to pull record ID 1 and 2 since the next most recent record for those IDs have had at least one of their address data change. ID 3 has exactly the same address data so I would not consider that in my criteria.
Can i do this with SQL only or will i need to create a procedure or function?
Any help is appreciated
Ben

Answer 1

Ben C wrote:
Hi All,
I have an address table with ID as the PK but there can be more than one ID. You mean there can be more than one row with tha same ID, right? (There can be more than one ID, also.)
I need to take the latest address record and first see if that same ID has a previous address record and if so, compare each column to see if the data has changed. I only need to compare the latest record with the next latest record.
... Sample table:Whenever you have a problem, please post CREATE TABLE and INSERT statements for the sample data, and the exact results you want from that data.
Explain, using specific examples, how you get those results from that data.
Always say which version of Oracle you're using.
Can i do this with SQL only or will i need to create a procedure or function?You don't need PL/SQL, if I understand your requirements. You can do something like this:
WITH     got_analytics     AS
     SELECT     id, street, city, state, zip, effective_date
     ,     COUNT (*)     OVER ( PARTITION BY  id )          AS cnt
     ,     ROW_NUMBER () OVER ( PARTITION BY  id
                               ORDER BY          effective_date     DESC
                       )                           AS r_num
     FROM    table_x
--     WHERE     ...     -- If you need any filtering, put it here
SELECT    id, street, city, state, zip, effective_date
FROM       got_analytics
WHERE       cnt     > 1
AND       r_num     <= 2
ORDER BY  id
,            effective_date
;This will show the most recent 2 rows for each id that has 2 (or more) rows.
You can modify it to only include ids where there was some change in street, city, state or zip between those 2 rows.
WITH     got_analytics     AS
     SELECT     id, street, city, state, zip, effective_date
     ,     COUNT (*)     OVER ( PARTITION BY  id )          AS cnt
     ,     ROW_NUMBER () OVER ( PARTITION BY  id
                               ORDER BY          effective_date     DESC
                       )                           AS r_num
     FROM    table_x
--     WHERE     ...     -- If you need any filtering, put it here
,     got_unique_cnts     AS
     SELECT     id, street, city, state, zip, effective_date
     ,     COUNT (DISTINCT '~' || street) OVER (PARTITION BY id)     AS street_cnt
     ,     COUNT (DISTINCT '~' || city)   OVER (PARTITION BY id)     AS city_cnt
     ,     COUNT (DISTINCT '~' || state)  OVER (PARTITION BY id)     AS state_cnt
     ,     COUNT (DISTINCT '~' || zip)    OVER (PARTITION BY id)     AS zip_cnt
     FROM       got_analytics
     WHERE       cnt     > 1
     AND       r_num     <= 2
SELECT    id, street, city, state, zip, effective_date
FROM       got_unique_cnts
WHERE       street_cnt     > 1
OR       city_cnt     > 1
OR       state_cnt     > 1
OR       zip_cnt     > 1
ORDER BY  id
,            effective_date
;This assumes that street, city, state and zip are all VARCHAR2s. If any of those columns is NULL on one row, and not NULL on the other row for that ID, that counts as a difference.
Edited by: Frank Kulash on Mar 7, 2012 3:37 PM
Edited by: Frank Kulash on Mar 7, 2012 3:41 PM