LAG Function not returning the default value in sql query

Question

I have an audit table per_addresses_f_

address_id              effective_start_date            effective_end_date          last_update_date    audit_change_bit_map_       audit_action_type_ address_line_2
300000009360612         2000-03-27                      2020-03-31                  2020-07-06          00100000000000000           Update                  
300000009360612         2020-04-01                      4712-12-31                  2020-07-06          01101100000000000           Update              Street12
300000009360612         2000-03-27                      4712-12-31                  2020-03-13          0                           History     
300000009360612         2000-03-27                      2020-03-31                  2020-03-13          0                           History         

I have used the below query to find out the change in address_line_2

select * from (
select 'Address' as Object
  , address_id s1
  , effective_start_date s2
  , effective_end_date s3
  , (select  max(person_id) from PER_PERSON_ADDR_USAGES_F_ pauf where pauf.address_id = t.address_id) s4
  , last_update_date s8
  ,audit_change_bit_map_ s9 
  ,case when audit_action_type_ = 'UPDATE' then 
    case substr(audit_change_bit_map_, 2, 2) when '11' then 'Update' 
    when '00' then 'Correction' 
    when '01' then 'Update'  end 
  when audit_action_type_ = 'HISTORY' and creation_date = last_update_date then 'Creation' 
  when audit_action_type_ = 'HISTORY' and creation_date <> last_update_date then 'History' 
    when audit_action_type_ = 'INSERT' then 'New'
  when audit_action_type_ = 'DELETE' then 'Deleted' end Audit_Action
 , 'Address Line 2' change
 , ADDRESS_LINE_2 current_value
 , LAG(t.address_line_2, 1,0) OVER (partition by  (select  max(person_id) from PER_PERSON_ADDR_USAGES_F_ pauf where pauf.address_id = t.address_id), address_id ORDER BY last_update_date)  Prev_value
from fusion.per_addresses_f_ t 
where address_id = 300000009360612 
order by last_update_date desc) 

In the above query the LAG function is not working correctly, if the previous value for the column is null, then 0 should be returned, but it not returning anything. The output the above query is giving -

address_id          effective_start_date    effective_end_date      last_update_date    audit_change_bit_map_       audit_action_type_ CURRENT_VALUE        prev_value  
300000009360612     2000-03-27              2020-03-31              2020-07-06          00100000000000000           Update                  
300000009360612     2020-04-01              4712-12-31              2020-07-06          01101100000000000           Update             Street12                 
300000009360612     2000-03-27              4712-12-31              2020-03-13          0                           History                                   0 
300000009360612     2000-03-27              2020-03-31              2020-03-13          0                           History         

so when i add the condition in the above query where current_value <> prev_value , it is not returning anything. Ideally it should have returned-

address_id          effective_start_date    effective_end_date      last_update_date    audit_change_bit_map_       audit_action_type_ CURRENT_VALUE        prev_value                  
300000009360612     2020-04-01              4712-12-31              2020-07-06          01101100000000000           Update          Street12            0                   

Is the lag function i am using incorrect?

Answer 1

You provided just part of your data: I see just 1 value in the address_line_2 - line #2 You can use simple lag without subqieries:

lag(ADDRESS_LINE_2) over(partition by address_id order by effective_start_date) Prev_value

Full test case on data you provided: (corrected a bit to make it clearer)

with per_addresses_f_(
    address_id
   ,effective_start_date
   ,effective_end_date
   ,last_update_date
   ,audit_change_bit_map_
   ,audit_action_type_
   ,address_line_2 -- just 1 'Street12' in your example, so I filled other values to make it more clear
   ,creation_date -- you didn't provide this field, so I filled it with sysdate
   ) as (
select 300000009360612 , date'2000-03-27' , date'2020-03-31', date'2020-07-06' , '00100000000000000', 'Update ','Street11',sysdate from dual union all
select 300000009360612 , date'2020-04-01' , date'4712-12-31', date'2020-07-06' , '01101100000000000', 'Update ','Street12',sysdate from dual union all
select 300000009360612 , date'2000-03-27' , date'4712-12-31', date'2020-03-13' , '0                ', 'History','Street13',sysdate from dual union all
select 300000009360612 , date'2000-03-27' , date'2020-03-31', date'2020-03-13' , '0                ', 'History','Street14',sysdate from dual 
)
select * from (
select 'Address' as Object
  , address_id s1
  , effective_start_date s2
  , effective_end_date s3
--  , (select  max(person_id) from PER_PERSON_ADDR_USAGES_F_ pauf where pauf.address_id = t.address_id) s4
  , last_update_date s8
  ,audit_change_bit_map_ s9 
  ,case when audit_action_type_ = 'UPDATE' then 
    case substr(audit_change_bit_map_, 2, 2) when '11' then 'Update' 
    when '00' then 'Correction' 
    when '01' then 'Update'  end 
  when audit_action_type_ = 'HISTORY' and creation_date = last_update_date then 'Creation' 
  when audit_action_type_ = 'HISTORY' and creation_date <> last_update_date then 'History' 
    when audit_action_type_ = 'INSERT' then 'New'
  when audit_action_type_ = 'DELETE' then 'Deleted' end Audit_Action
 , 'Address Line 2' change
 , ADDRESS_LINE_2 current_value
-- , LAG(t.address_line_2, 1,0) OVER (partition by  (select  max(person_id) from PER_PERSON_ADDR_USAGES_F_ pauf where pauf.address_id = t.address_id), address_id ORDER BY last_update_date)  Prev_value
 , lag(ADDRESS_LINE_2) over(partition by address_id order by effective_start_date) Prev_value
from fusion.per_addresses_f_ t 
where address_id = 300000009360612 
order by last_update_date desc)