Squashing values in a column by another timestamp colum - sql

Question

I have a following data which shows the status of a support ticket:
Edit:
More concise and generic example:
STATUS SEQ_NO
New 1
Open 2
Open 3
Open 4
Queued 5
Open 6
Open 7
Open 8
Completed 9
Completed 10
Completed 11
Closed 12
From this, I would like to extract the records,
STATUS SEQ_NO
New 1
Open 2
Queued 5
Open 6
Completed 9
Closed 12
Original question:
-- SELECT status, start_time FROM events_tab ORDER BY start_time;
STATUS START_TIME
New 30/09/2014 3:48:10 PM -- I want this record,
Open 30/09/2014 3:48:10 PM -- and this,
Open 1/10/2014 10:41:57 AM
Open 4/03/2015 9:59:04 AM
Queued 18/06/2015 1:31:30 PM -- and this,
Open 20/06/2015 10:10:47 PM -- and this,
Open 20/06/2015 11:20:11 PM
Open 27/06/2015 1:18:50 PM
Completed 27/06/2015 1:22:08 PM -- and this,
Completed 28/09/2015 9:31:55 AM
Completed 5/10/2015 11:57:38 AM
Closed 11/01/2016 9:31:26 AM -- and this.
These are events that happened in each state. I want to make a timeline of state changes from it.
I want to squash these records such that only the very first row of a group is show. However, notice that there are actually two groups of Open status. So I should get two records with Open status.
Basically I want the following result:
STATUS START_TIME
New 30/09/2014 3:48:10 PM
Open 30/09/2014 3:48:10 PM
Queued 18/06/2015 1:31:30 PM
Open 20/06/2015 10:10:47 PM
Completed 27/06/2015 1:22:08 PM
Closed 11/01/2016 9:31:26 AM
How can I achieve this with an SQL statement?
I have tried,
SELECT status, MIN(start_time)
FROM events_tab
GROUP BY status;
But this does not include multiple records in Open status, as my intention above.

Answer 1

You can use the Tabibitosan technique to achieve this goal:
WITH your_table AS (SELECT 'New' status, to_date('30/09/2014 03:48:10 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('30/09/2014 03:48:10 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('1/10/2014 10:41:57 AM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('4/03/2015 09:59:04 AM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Queued' status, to_date('18/06/2015 01:31:30 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('20/06/2015 10:10:47 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('20/06/2015 11:20:11 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Open' status, to_date('27/06/2015 01:18:50 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Completed' status, to_date('27/06/2015 01:22:08 PM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Completed' status, to_date('28/09/2015 09:31:55 AM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Completed' status, to_date('5/10/2015 11:57:38 AM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual UNION ALL
SELECT 'Closed' status, to_date('11/01/2016 09:31:26 AM', 'dd/mm/yyyy hh:mi:ss AM') start_time FROM dual)
SELECT status,
MIN(start_time) start_time
FROM (SELECT status,
start_time,
row_number() OVER (ORDER BY start_time, status) - row_number() OVER (PARTITION BY status ORDER BY start_time, status) grp
FROM your_table)
GROUP BY status, grp
ORDER BY start_time, status;
STATUS START_TIME
--------- -------------------
New 30/09/2014 15:48:10
Open 30/09/2014 15:48:10
Queued 18/06/2015 13:31:30
Open 20/06/2015 22:10:47
Completed 27/06/2015 13:22:08
Closed 11/01/2016 09:31:26
N.B. Since you have rows with different statuses having the same start_time, I have added status into the order by, in order to get the results you were after. I don't know if that was a typo, or whether multiple rows really can have the same date.
Also, I assume that the data in your example refers to one "thing", but in your real table, you can have multiple "things" each with their own set of statuses etc.
In that case, you would need to add the column(s) that differentiate the "things" (e.g. id or event_name or etc) into both row_number() analytic functions. (e.g. row_number() over (partition by <thing column(s)> order by start_time, status))

Answer 2

You can also try the SQL for Pattern Matching
WITH tickets(STATUS, START_TIME) AS (
SELECT 'New', TO_DATE('30/09/2014 3:48:10 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('30/09/2014 3:48:10 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('1/10/2014 10:41:57 AM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('4/03/2015 9:59:04 AM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Queued', TO_DATE('18/06/2015 1:31:30 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('20/06/2015 10:10:47 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('20/06/2015 11:20:11 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Open', TO_DATE('27/06/2015 1:18:50 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Completed', TO_DATE('27/06/2015 1:22:08 PM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Completed', TO_DATE('28/09/2015 9:31:55 AM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Completed', TO_DATE('5/10/2015 11:57:38 AM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual UNION ALL
SELECT 'Closed', TO_DATE('11/01/2016 9:31:26 AM', 'dd/mm/yyyy hh:mi:ss AM') FROM dual)
SELECT STATUS, START_TIME
FROM tickets
MATCH_RECOGNIZE (
ORDER BY START_TIME
MEASURES
START_TIME AS START_TIME,
STATUS as STATUS
PATTERN ( CHNG )
DEFINE
CHNG AS CHNG.STATUS <> PREV(CHNG.STATUS) OR PREV(CHNG.STATUS) IS NULL
)
STATUS START_TIME
========== ====================
New 30.09.2014 15:48:10
Open 30.09.2014 15:48:10
Queued 18.06.2015 13:31:30
Open 20.06.2015 22:10:47
Completed 27.06.2015 13:22:08
Closed 11.01.2016 09:31:26
CHNG.STATUS <> PREV(CHNG.STATUS) matches each row where STATUS is different to previous row. PREV(CHNG.STATUS) IS NULL is used to get also the very first row.

Answer 3

use row_number window function
select STATUS ,START_TIME from
(
select STATUS,START_TIME,
row_number() over (partition by STATUS,EXTRACT(YEAR FROM START_TIME) order by START_TIME) rn
from events_tab
) t where rn=1

Answer 4

Use LAG Function as you need to track the change in status:
https://dbfiddle.uk/?rdbms=oracle_11.2&fiddle=38a991b698c858f6f0417c7d4c0dc9d3
with cte1 (st,dt) as
(
select 'New' as st, '30/09/2014 3:48:10 PM' as dt from dual
union all
select 'Open' as st, '30/09/2014 3:48:10 PM' as dt from dual
union all
select 'Open' as st, '20/09/2014 3:48:10 PM' as dt from dual
union all
select 'Qued' as st, '18/06/2015 1:31:30' as dt from dual
)
select st, min(case when st<>prev_order_date then dt else dt end) as d
from
(
SELECT st, dt,
LAG (st,1) OVER (ORDER BY st) AS prev_order_date
FROM cte1
)a
group by st