【Question】
I have a table structured like this:

 

index   date time status

1 2015-01-01 13:00:00 start

2 2015-01-01 13:10:00 continue

3 2015-01-01 13:20:00 continue

4 2015-01-01 13:30:00 end

5 2015-01-01 13:30:00 ready

6 2015-01-01 13:40:00 start

7 2015-01-01 13:50:00 continue

8 2015-01-01 15:00:00 end`

 

What I would like to do is to count the time between start and end (ie. index 1-4 is 30min, 6-8 is 20min) by taking into account only the first start and first end. So the query doesn’t choose the time difference of index 1-8. I need a query to calculate time difference between two statuses (start-end) and show the result for multiple instances of start-end without them getting batched into one event?

 

A solution:

SELECT x.*

 

 , TIMEDIFF(MIN(y.dt),x.dt)diff

 

 FROM my_table x

 

 JOIN my_table y

 

 ON y.dt >= x.dt

 

WHERE x.status = 'start'

 

 AND y.status = 'end'

 

GROUP

 

 BY x.id;

 

【Answer】

It’s complicated to handle order-based calculations in SQL. You can do it in esProc SPL (Structured Process Language) with simple and easy code:

 

	A
1	$select * from tb1
2	=A1.group((#-1)\2)
3	=A2.new(~(1).time:start,~(2).time:end,interval@s(start,end)/60:diff)

A3: Calculate time differences to generate a new table sequence.

The SPL script can be easily embedded into a Java application via esProc JDBC. See How to Call an SPL Script in Java.