Reputation: 3
I need to grab the last rundate when then churn value changed. I would need the result to return this:
> customer billto banner rundate churn lastchurndate
> 976193 976193 GexPro 12/04/2019 true 11/26/2019
> 976193 976193 GexPro 11/26/2019 true 11/26/2019
> 976193 976193 GexPro 11/19/2019 false 11/26/2019
This is the current raw data set:
> customer billto banner rundate churn
> 976193 976193 GexPro 12/04/2019 true
>976193 976193 GexPro 11/26/2019 true
>976193 976193 GexPro 11/19/2019 false
The churn value changed on 11/26/2019 from False to True.
I need to know when the last date was when churn <> current churn date if I were to query on current rundate as of today.
I have a query where I was able to get 11/19/2019, but it's not quit right.
select
v.st_cust_no
,v.bt_cust_no
,v.banner
,v.rundate
,v.churn
,case when c."last churn date" is null then v.rundate
else c."last churn date"
end as "last churn date"
,c."last churn"
from
dwstage v
left join
(
select
z.st_cust_no
,z.BT_CUST_NO
,z.banner
,z.rundate
,z.churn as "current churn"
,x.latest as "last churn date"
,x.churn "last churn"
from dwstage z
inner join
(select
st_cust_no
,BT_CUST_NO
,banner
,churn
,max(rundate) as latest
from dwstage
group by
st_cust_no
,BT_CUST_NO
,banner
,churn) x
on z.st_cust_no = x.st_cust_no
and z.bt_cust_no = x.bt_cust_no
and z.BANNER = x.BANNER
where z.churn <> x.churn
) c
on v.st_cust_no = c.st_cust_no
and v.bt_cust_no = c.bt_cust_no
and v.BANNER = c.BANNER
and v.rundate = c.rundate
-- where v.st_cust_no = '14025'
order by
v.st_cust_no
,v.banner
,v.rundate desc
;
Upvotes: 0
Views: 155
Reputation:
Oracle 12.1 introduced the match_recognize
clause, which can do very quick work on this kind of problem. Here is one way. I start with creating a small test table, with two different customers to test that the query works correctly on that more general scenario.
Table creation:
create table dwstage (customer, billto, banner, rundate, churn) as
select 976193, 976193, 'GexPro', to_date('12/04/2019', 'mm/dd/yyyy'), 'true' from dual union all
select 976193, 976193, 'GexPro', to_date('11/26/2019', 'mm/dd/yyyy'), 'true' from dual union all
select 976193, 976193, 'GexPro', to_date('11/19/2019', 'mm/dd/yyyy'), 'false' from dual union all
select 999999, 999999, 'Banner', to_date('03/21/2019', 'mm/dd/yyyy'), 'false' from dual union all
select 999999, 999999, 'Banner', to_date('04/21/2019', 'mm/dd/yyyy'), 'true' from dual union all
select 999999, 999999, 'Banner', to_date('08/03/2019', 'mm/dd/yyyy'), 'false' from dual union all
select 999999, 999999, 'Other' , to_date('08/31/2019', 'mm/dd/yyyy'), 'false' from dual
;
Query and output:
select customer, billto, banner, rundate, churn, lastchurndate
from dwstage
match_recognize(
partition by customer
order by rundate desc
measures final last(a.rundate) as lastchurndate
all rows per match
pattern (^ a+ b*)
define a as churn = first(churn)
)
;
CUSTOMER BILLTO BANNER RUNDATE CHURN LASTCHURND
---------- ---------- ------ ---------- ----- ----------
976193 976193 GexPro 12/04/2019 true 11/26/2019
976193 976193 GexPro 11/26/2019 true 11/26/2019
976193 976193 GexPro 11/19/2019 false 11/26/2019
999999 999999 Other 08/31/2019 false 08/03/2019
999999 999999 Banner 08/03/2019 false 08/03/2019
999999 999999 Banner 04/21/2019 true 08/03/2019
999999 999999 Banner 03/21/2019 false 08/03/2019
Upvotes: 0
Reputation: 1269773
You want the date on every row, so you do not want any filtering. This suggests window functions:
select s.*,
max(case when prev_churn <> churn then rundate) over (partition by customer, billto) as last_churndate
from (select s.*,
lag(churn) over (order by customer, billto order by rundate) as prev_churn
from dwstage s
) s;
Upvotes: 0
Reputation: 35900
You can take advantage of analytical functions as following:
Select * from
(Select t.*,
Row_number() over (partition by customer order by rundate desc nulls last) as rn
from
(Select t.*,
case when lag(chrun) over (partition by cutomer order by rundate) <> chrun then 1 end as chrunchanged
From your_table t) t
Where chrunchanged = 1)
Where rn = 1
Cheers!!
Upvotes: 1