I have trouble using crosstab() in postgresql-11.
Here is my table,
CREATE TABLE monitor(tz timestamptz, level int, table_name text, status text);
The table monitors events on other tables. It contains
table_name (table on which the event occurred)
timestamp(time at which the event occurred)
level (level of the event)
status of the event (start/end of the event)
Here is the sample data to it.
tz | level | status | table_name
----------------------------------+-------+--------+--------------
2019-10-24 16:18:34.89435+05:30 | 2 | start | test_table_2
2019-10-24 16:18:58.922523+05:30 | 2 | end | test_table_2
2019-11-01 10:31:08.948459+05:30 | 3 | start | test_table_3
2019-11-01 10:41:22.863529+05:30 | 3 | end | test_table_3
2019-11-01 10:51:44.009129+05:30 | 3 | start | test_table_3
2019-11-01 12:35:23.280294+05:30 | 3 | end | test_table_3
Given a timestamp, I want to list out all current events at that time. It could be done using the criteria,
start_time >= 'given_timestamp' and end_time <= 'given_timestamp'
So I tried to use crosstab() to pivot the table over columns table_name,status and timestamp. My query is,
with q1 (table_name, start_time,end_time) as
(select * from crosstab
('select table_name, status, tz from monitor ')
as finalresult (table_name text, start_time timestamptz, end_time timestamptz)),
q2 (level,start_time,end_time) as
(select * from crosstab('select level, status, tz from monitor ')
as finalresult (level int, start_time timestamptz, end_time timestamptz))
select q1.table_name,q2.level,q1.start_time,q1.end_time
from q1,q2
where q1.start_time=q2.start_time;
The output of the query is,
table_name | level | start_time | end_time
--------------+-------+----------------------------------+----------------------------------
test_table_2 | 2 | 2019-10-24 16:18:34.89435+05:30 | 2019-10-24 16:18:58.922523+05:30
test_table_3 | 3 | 2019-11-01 10:31:08.948459+05:30 | 2019-11-01 10:41:22.863529+05:30
But my expected output is,
table_name | level | start_time | end_time
--------------+-------+----------------------------------+----------------------------------
test_table_2 | 2 | 2019-10-24 16:18:34.89435+05:30 | 2019-10-24 16:18:58.922523+05:30
test_table_3 | 3 | 2019-11-01 10:31:08.948459+05:30 | 2019-11-01 10:41:22.863529+05:30
test_table_3 | 3 | 2019-11-01 10:51:44.009129+05:30 | 2019-11-01 12:35:23.280294+05:30
How do I achieve the expected output? Or is there any better way other than crosstab?
I would use a self join for this. To keep the rows on the same level and table together you can use a window function to assign numbers to them so they can be distinguished.
with numbered as (
select tz, level, table_name, status,
row_number() over (partition by table_name, status order by tz) as rn
from monitor
)
select st.table_name, st.level, st.tz as start_time, et.tz as end_time
from numbered as st
join numbered as et on st.table_name = et.table_name
and et.status = 'end'
and et.level = st.level
and et.rn = st.rn
where st.status = 'start'
order by st.table_name, st.level;
This assumes that there will never be a row with status = 'end' and an earlier timestamp then the corresponding row with status = 'start'
Online example: https://rextester.com/QYJK57764