Sum Multiple Row Date Difference Mysql - mysql

Question

Table car_log
Speed LogDate
5 2013-04-30 10:10:09 ->row1
6 2013-04-30 10:12:15 ->row2
4 2013-04-30 10:13:44 ->row3
17 2013-04-30 10:15:32 ->row4
22 2013-04-30 10:18:19 ->row5
3 2013-04-30 10:22:33 ->row6
4 2013-04-30 10:24:14 ->row7
15 2013-04-30 10:26:59 ->row8
2 2013-04-30 10:29:19 ->row9
I want to know how long the car get speed under 10.
In my mind, i will count the LogDate difference between row 1 - row4 (because in 10:14:44 => between row4 and row3, the speed is 4) + (sum) LogDate difference between row6 - row8. I am doubt if it right or no.
How can i count it in mysql queries. Thank you.

Answer 1

For every row, find a first row with higher (later) LogDate. If the speed in this row is less than 10, count date difference between this row's date and next row's date, else put 0.
A query that would give a list of the values counted this way should look like:
SELECT ( SELECT IF( c1.speed <10, unix_timestamp( c2.LogDate ) - unix_timestamp( c1.logdate ) , 0 )
FROM car_log c2
WHERE c2.LogDate > c1.LogDate
LIMIT 1
) AS seconds_below_10
FROM car_log c1
Now its just a matter of summing it up:
SELECT sum( seconds_below_10) FROM
( SELECT ( SELECT IF( c1.speed <10, unix_timestamp( c2.LogDate ) - unix_timestamp( c1.logdate ) , 0 )
FROM car_log c2
WHERE c2.LogDate > c1.LogDate
LIMIT 1
) AS seconds_below_10
FROM car_log c1 ) seconds_between_logs
Update after comment about adding CarId:
When you have more than 1 car you need to add one more WHERE condition inside dependent subquery (we want next log for that exact car, not just any next log) and group whole rowset by CarId, possibly adding said CarId to the select to show it too.
SELECT sbl.carId, sum( sbl.seconds_below_10 ) as `seconds_with_speed_less_than_10` FROM
( SELECT c1.carId,
( SELECT IF( c1.speed <10, unix_timestamp( c2.LogDate ) - unix_timestamp( c1.logdate ) , 0 )
FROM car_log c2
WHERE c2.LogDate > c1.LogDate AND c2.carId = c1.carId
LIMIT 1 ) AS seconds_below_10
FROM car_log c1 ) sbl
GROUP BY sbl.carId
See an example at Sqlfiddle.

Answer 2

If the type of column 'LogDate' is a MySQL DATETIME type, you can use the timestampdiff() function in your select statement to get the difference between timestamps. The timestampdiff function is documented in the manual at:
http://dev.mysql.com/doc/refman/5.1/en/date-and-time-functions.html#function_timestampdiff
You need to break the query down into subqueries, and then use the TIMESTAMPDIFF function.
The function takes three arguments, the units you want the result in (ex. SECOND, MINUTE, DAY, etc), and then Value2, and last Value1.
To get the maximum value for LogDate where speed is less than 10 use:
select MAX(LogDate) from <yourtable> where Speed<10
To get the minimum value for LogDate where speed is less than 10 use:
select MIN(LogDate) from <yourtable> where Speed<10
Now, combine these into a single query with the TIMESTAMPDIFF function:
select TIMESTAMPDIFF(SECOND, (select MAX(LogDate) from <yourtable> where Speed<10, (select MIN(LogDate) from <yourtable> where Speed<10)));
If LogDate is of a different type, there are other Date/Time Diff functions to handle math between any of these types. You will just need to change 'TIMESTAMPDIFF' to the correct function for your column type.
Additional ref: http://dev.mysql.com/doc/refman/5.1/en/date-and-time-functions.html

Answer 3

Try this SQL:
;with data as (
select *
from ( values
( 5, convert(datetime,'2013-04-30 10:10:09') ),
( 6, convert(datetime,'2013-04-30 10:12:15') ),
( 4, convert(datetime,'2013-04-30 10:13:44') ),
(17, convert(datetime,'2013-04-30 10:15:32') ),
(22, convert(datetime,'2013-04-30 10:18:19') ),
( 3, convert(datetime,'2013-04-30 10:22:33') ),
( 4, convert(datetime,'2013-04-30 10:24:14') ),
(15, convert(datetime,'2013-04-30 10:26:59') ),
( 2, convert(datetime,'2013-04-30 10:29:19') )
) data(speed,logDate)
)
, durations as (
select
duration = case when speed<=10
then datediff(ss, logDate, endDate)
else 0
end
from (
select
t1.speed, t1.logDate, endDate = (
select top 1 logDate
from data
where data.logDate > t1.logDate
)
from data t1
) T
where endDate is not null
)
select TotalDuration = sum(duration)
from durations
which calculates 589 seconds from the sample data provided.