Get the total time every time a truck has no speed in SQL? - sql

Question

I have the following table in SQL Server 2014:
Vehicle_Id | Speed | Event | Datetime
-----------+---------+--------------+----------------------
1 | 0 | Door-Open | 2019-05-04 15:00:00
1 | 0 | Door-Closed | 2019-05-04 15:15:00
1 | 50 | Driving | 2019-05-04 15:35:00
1 | 0 | Parked | 2019-05-04 15:50:00
1 | 0 | Door-Open | 2019-05-04 15:51:00
1 | 0 | Door-Closed | 2019-05-04 15:52:00
1 | 50 | Driving | 2019-05-04 15:57:00
I need to identify blocks within a datetime in which the truck has been on speed = 0 for more than an hour. So every time a row appears with speed 0, it should create a unique block_id until a row with speed appears. So the total time should be the first time the truck has speed 0 until the next row it finds with speed > 0.
Expected Output:
Vehicle_Id | Speed | Event | Datetime | Block | Total_State_Time_Block(Minutes)
-----------+---------+--------------+------------------------+-------------+---------------------------------
1 | 0 | Door-Open | 2019-05-04 15:00:00 | 1 | 35 Minutes
1 | 0 | Door-Closed | 2019-05-04 15:15:00 | 1 | 35 Minutes
1 | 50 | Driving | 2019-05-04 15:35:00 | 2 | 15 Minutes
1 | 0 | Parked | 2019-05-04 15:50:00 | 3 | 7 Minutes
1 | 0 | Door-Open | 2019-05-04 15:51:00 | 3 | 7 Minutes
1 | 0 | Door-Closed | 2019-05-04 15:52:00 | 3 | 7 Minutes
1 | 50 | Driving | 2019-05-04 15:57:00 | 4 | ...
So, as it's ordered by datetime, the idea is to create groups of adjacent rows with speed = 0 so I can identify the times a truck hasn't moved for more than an hour.
I tried windowing functions to get the result by vehicle and day. But I can't achieve this last step.

Answer 1

You can try with lag()
select
vehicle_id,
speed,
event,
datetime,
sum(case when speed = rnk then 0 else 1 end) over (order by datetime) as block
from
(
select
*,
lag(speed) over (order by datetime) as rnk
from myTable
) val
output:
| vehicle_id | speed | event | datetime | block |
| ---------- | ----- | ----------- | ------------------------ | ----- |
| 1 | 0 | Door-Open | 2019-05-04 15:00:00 | 1 |
| 1 | 0 | Door-Closed | 2019-05-04 15:15:00 | 1 |
| 1 | 50 | Driving | 2019-05-04 15:35:00 | 2 |
| 1 | 0 | Parked | 2019-05-04 15:50:00 | 3 |
| 1 | 0 | Door-Open | 2019-05-04 15:51:00 | 3 |
| 1 | 0 | Door-Closed | 2019-05-04 15:52:00 | 3 |
| 1 | 50 | Driving | 2019-05-04 15:57:00 | 4 |

Answer 2

If you just want periods where the truck has been at speed = 0 for an hour or more, you don't need your expected output. Instead, you can look at the next value with a speed and calculate the decimal hours.
That is, you can get the blocks directly. This gets the start of the block with the duration:
select t.*,
datediff(second, datetime, coalesce(datetime, max_datetime)
) / (60.0 * 60) as decimal_hours
from (select t.*,
lag(speed) over (partition by vehicle_id order by datetime) as prev_speed
min(case when speed > 0 then datetime end) over (partition by vehicle_id order by datetime) as next_speed,
max(datetime) over (partition by vehicle_id) as max_datetime
from t
) t
where (prev_speed is null or prev_speed > 0) and
speed = 0