Getting the latest area for each case

Question

dear all;
I have a table with the following info below
create table tmp_a
       idv number(30),
       personid number(30),
       Area varchar2(200),
       dateadded date,
       is_close number(2)
insert into tmp_a
  (idv, personid, area, dateadded, is_close)
values
  (1, 10001, 'ZONE_A', sysdate, 1);
insert into tmp_a
  (idv, personid, area, dateadded, is_close)
values
  (2, 10002, 'ZONE_B', sysdate, 0);
insert into tmp_a
  (idv, personid, area, dateadded, is_close)
values
  (3, 10003, 'ZONE_A', sysdate, 0);
insert into tmp_a
  (idv, personid, area, dateadded, is_close)
values
  (4, 10003, 'ZONE_A', sysdate, 0);
  Now I would like to get this below
idv   Personid   Area        Dateadded                           Is_clode
4       10003   Zone_A     3/3/2011 1:00:37PM               0
2       10002   Zone_B     3/3/2011 12:17;14PM              0the output is based on the fact, if there are duplicates associated with the area, the query should pick the latest one...In this case, the latest one for zone_A is idv 4.
kindly note, idv is a primary key. How do I go about doing that? All help is appreciated. Thank you.

Answer 1

Hi,
ROW_NUMBER returns a unique number for every row. If you have n rows, it will return the integers 1 through n, without skipping or duplicating any numbers, so you can be confident it will return the number 1 on exactly 1 row.
ROW_NUMBER (PARTITION BY area ...)     means that a separate set of integers, each starting with 1, will be returned for each distinct value of area.
ROW_NUMBER (PARTITION BY area ORDER BY idv DESC)     means that the numbers will be assigned in DESCending order by idv. The row with the highest idv for each area will be the row for which ROW_NUMBER returns 1.
The LAST function, as demonstrated by Solomon, is a good thing to have in your toolbox, but you have to code a somewhat complicated expression for each column (except the GROUP BY column, and the column that determines the order). When you need to pick an entire row, I find it easier to use ROW_NUMBER.
Also, ROW_NUMBER can easily be adapted if your needs change. Where I work, people are always saying things like "Remember that query you wrote for me, that returns the latest row fro each area? Can you make it return the latest 5 rows?" or "It truns out idv isn't unique. Can you make it return all the rows with the highest idv?" It's easy to change the ROW_NUMBER solution to do things like that.