Creating a status log from rows of datetimes of status changes - sql

Question

I'm pulling down some data from a remote API to a local SQL Server table, which is formatted like so. (imagine it's sorted by StatusDT descending)
DriverID StatusDT Status
-------- -------- ------
b103 2019-03-05 05:42:52:000 D
b103 2019-03-03 23:45:42.000 SB
b103 2019-03-03 21:49:41.000 ON
What would be the best way to eventually get to a point where I can return a query showing the total amount of time spent in each status on each day for each driver?
Also, it's possible that there could be gaps of a whole day or more between status updates, in which case I'd need a row showing a continuation of the previous status from 00:00:00 to 23:59:59 for each skipped day. So, if I'm looping through this table to populate another with the structure below, the example above would need to wind up looking like this... (again, sorted descending by date)
DriverID StartDT EndDT Status
-------- --------------- -------------- ------
b103 2019-03-05 05:42:52 D
b103 2019-03-05 00:00:00 2019-03-05 05:42:51 SB
b103 2019-03-04 00:00:00 2019-03-04 23:59:59 SB
b103 2019-03-03 23:45:42 2019-03-03 23:59:59 SB
b103 2019-03-03 21:49:41 2019-03-03 23:45:41 ON
Does that make sense?
I wound up dumping the API data to a "work" table and running a cursor over it to add rows to another table, with the starting and ending date/time, but I'm curious if there's another way that might be more efficient.
Thanks very much.

Answer 1

I think this query is what you need. I couldn't test it, however, for syntax errors:
with x as (
select
DriverID,
StatusDT as StartDT,
lead(StatusID) over(partition by DriverID order by StatusDT) as EndDT,
Status
from my_table
)
select -- start & end on the same day
DriverID,
StartDT,
EndDT,
Status
from x
where convert(date, StartDT) = convert(date, EndDT)
or EndDT is null
union all
select -- start & end on different days; first day up to midnight
DriverID,
StartDT,
dateadd(ms, -3, convert(date, EndDT)) as EndDT,
Status
from x
where convert(date, StartDT) <> convert(date, EndDT)
and or EndDT is not null
union all
select -- start & end on different days; next day from midnight
DriverID,
convert(date, EndDT) as StartDT,
EndDT,
Status
from x
where convert(date, StartDT) <> convert(date, EndDT)
and or EndDT is not null
order by StartDT desc

Answer 2

Most of your answer is just using lead():
select driverid, status, statusdt,
lead(statusdt) over (partition by driverid order by statusdt) as enddte
from t;
This does not give the breaks by day. But you can add those. I think the easiest way is to add in the dates (using a recursive CTE) and compute the status at that time. So:
I would do the following:
use a recursive CTE to calculate the dates
"fill in" the statuses and union to the original table
use lead() to get the end date
This looks like:
with day_boundaries as (
select driverid, dateadd(day, 1, convert(min(statusdt) as date) as statusdt, max(statusdt) as finaldt
from t
group by driverid
having datediff(da, min(statusdt), max(statusdt)) > 0
union all
select driverid, dateadd(day, 1, statusdt), finaldt
from day_boundaries
where statusdt < finaldt
),
unioned as (
select driverid, status, statusdt
from t
union all
select db.driverid, s.status, db.statusdt
from day_boundaries db cross apply
(select top (1) status
from t
where t.statusdt < db.statusdt
order by t.statusdt desc
) s
)
select driverid, status, statusdt,
lead(statusdt) over (partition by driverid order by statusdt) as enddte
from unioned;
Note that this does not subtract any seconds from the end date. The end date matches the previous start date. Time is continuous. It makes no sense to have gaps for records that should snugly fit together.