Assuming a very simple table with:

One column ts (of type nanotimestamp),

One column value (of type double).

I want to sample the value once every hour.

The standard syntax should be like this：

ts = 2024.03.26T09:30:00.005 2024.03.26T09:31:00.123 2024.03.26T10:30:00.105
val = 1.5 2.5 2.0
t = table(ts, val)
select last(val)
from t
group by bar(ts, 1s)
order by bar_ts

output：

bar_ts              last_val
------------------- --------
2024.03.26T09:30:00 1.5     
2024.03.26T09:31:00 2.5     
2024.03.26T10:30:00 2 

However, the results obtained this way are not precisely 3600 records since the data points are not evenly distributed along the timeline. Sometimes, there is not a single value for one or two seconds, resulting in missing time intervals, such as the 09:32:00 to 10:29:00 gap in the example above.

Is there a way to accurately generate samples representing the true ‘last seen value’?