How to do self join and create a new column in sql - sql

Question

I created a table from a huge database after running multiple queries like the following is the sample:
communication_day | comid | person_id | area_id | w_id
20-MAR-17 03:45 | 21 | xyz1 | y123 | S1
20-MAR-17 07:45 | 21 | xyz1 | y142 | S2
14-MAR-17 07:45 | 41 | xyz1 | y153 | S1
14-MAR-17 09:00 | 41 | xyz1 | y123 | S3
30-MAR-17 09:00 | 88 | abc1 | y180 | D1
30-MAR-17 11:00 | 88 | abc1 | y181 | D2
Basically, a person (represented by person_id) comes in and requests change in its area id, this could be done be a person multiple times. Each time a person requests this, he is given a comid. So, like in the above example person_id = xyz1 first requests a change from y153 to y123 and then on 20th march, requests a change from y123 to y142.
I want to create a new column in this table called old_w_id such that for each change request (each comid) it has that old value in it. Eg for the above sample:
old_w_id
S1
S1
S1
S1
D1
D1
Cases where request came in for that comid, the old_w_id can have the same value only. Only where the change request has taken effect, old_w_id should have the previous value.
Any idea how do I do that?? Thanks in advance!!
UPDATE: Each comid might not necessarily have 2 rows i.e, a person might have requested the change but it might not have been acknowledged yet (that gives just one row for that comid at this time).

Answer 1

Yes of course I can explain. partition by clause tells analytic function to do calculation partitioned by column you specified, in our case it's the person_id, otherwise lag function will work for all rows through your returned row set. And also I'll remove default value for first record, I'll change lag(w_id, 1, w_id) to lag(w_id, 1) it's more correct to show null, because there is no lag w_id for it. Something like this.
with my_table as
(
select to_date('20-MAR-17 03:45', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 21 as comid,
'xyz1' as person_id, 'y123' as area_id, 'S1' w_id
from dual
union all
select to_date('20-MAR-17 03:45', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 21 as comid,
'xyz1' as person_id, 'y142' as area_id, 'S2' w_id
from dual
union all
select to_date('14-MAR-17 07:45', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 41 as comid,
'xyz1' as person_id, 'y153' as area_id, 'S1' w_id
from dual
union all
select to_date('14-MAR-17 09:00', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 41 as comid,
'xyz1' as person_id, 'y123' as area_id, 'S3' w_id
from dual
union all
select to_date('30-MAR-17 09:05', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 88 as comid,
'abc1' as person_id, 'y180' as area_id, 'D1' w_id
from dual
union all
select to_date('30-MAR-17 11:00', 'DD-MON-YY HH24:MI',
'nls_date_language=english') as communication_day, 88 as comid,
'abc1' as person_id, 'y181' as area_id, 'D2' w_id
from dual)
select t.*,
lag(w_id, 1) over(partition by person_id, comid order by communication_day) as old_w_id
from my_table t
out put will be
COMMUNICATION_DAY COMID PERSON_ID AREA_ID W_ID OLD_W_ID
1 30/03/17 09:05:00 88 abc1 y180 D1
2 30/03/17 11:00:00 88 abc1 y181 D2 D1
3 14/03/17 07:45:00 41 xyz1 y153 S1
4 14/03/17 09:00:00 41 xyz1 y123 S3 S1
5 20/03/17 03:45:00 21 xyz1 y123 S1
6 20/03/17 03:45:00 21 xyz1 y142 S2 S1