SQL fill gaps with hold - sql

Question

I've encountered a problem I cannot solve with my knowledge and I haven't found any solutions I understood good enough to solve my problem.
So here is what I try to achieve.
I have a database with the following structure:
node_id, source_time, value
1 , 10:13:15 , 1
2 , 10:13:15 , 1
2 , 10:13:16 , 2
1 , 10:13:19 , 2
1 , 10:13:25 , 3
2 , 10:13:28 , 3
I want to have a sql query to get the following output
time , value1, value2
10:13:15, 1 , 1
10:13:16, 1 , 2
10:13:19, 2 , 2
10:13:25, 3 , 2
10:13:28, 3 , 3
You see, the times are all times that occur from both nodes.
But the values have to be filled in the gaps since node1 has no value for the time :16 and :28.
I got it to the point where I get the 2 columns from one table. That was not the hard part.
SELECT T1.[value], T2.[value]
FROM [db1].[t_value_history] T1, [db1].[t_value_history] T2
WHERE ( T1.node_id = 1 AND T2.node_id = 2)
But the result doesn't look like the way I want it to be.
I found something with COALESCE and another table which holds the previous value. But that looked quiet complicated for such a easy thing.
I guess there is an easy sql solution but I haven't had much time to get into the materia.
I would be happy to get any idea which function to use.
Thanks so far.
Edit: Changed the database, made a mistake on the last line.
Edit2: I am using SQL Server. Sorry for not clarifying this. Also the values are not neccessarily increasing. I just used increasing numbers in this example here.

Answer 1

This works in SQL Server. If you are certain that there is a value for both nodes for the minimum time then you could change the OUTER APPLY to a CROSS APPLY, which would perform better.
WITH times
AS ( SELECT DISTINCT
source_time
FROM dbo.t_value_history
)
SELECT t.source_time ,
n1.value ,
n2.value
FROM times AS t
OUTER APPLY ( SELECT TOP 1
h.value
FROM dbo.t_value_history AS h
WHERE h.node_id = 1
AND h.source_time <= t.source_time
ORDER BY h.source_time DESC
) AS n1
OUTER APPLY ( SELECT TOP 1
h.value
FROM dbo.t_value_history AS h
WHERE h.node_id = 2
AND h.source_time <= t.source_time
ORDER BY h.source_time DESC
) AS n2;

Answer 2

You could use conditional aggregation to get the right set of rows:
select vh.source_time,
max(case when vh.node_id = 1 then value end) as value_1,
max(case when vh.node_id = 2 then value end) as value_2
from db1.t_value_history vh
group by vh.source_time;
If you want to fill in the values, then the best solution is lag() with ignore nulls. Supported by ANSI, but not by SQL Server (which I'm guessing you are using). Your values appear to be increasing. If that is the case, you can use a cumulative max:
select vh.source_time,
max(max(case when vh.node_id = 1 then value end)) over (order by vh.source_time) as value_1,
max(max(case when vh.node_id = 2 then value end) over (order by vh.source_time) as value_2
from db1.t_value_history vh
group by vh.source_time;
In your data, value is increasing, so this works for the data in your example. If that is not the case, a more complex query is needed to fill in the gaps.

Answer 3

This will do it in SQL Server. It is not 'nice' though:
SELECT DISTINCT
T1.source_time,
CASE WHEN T1.node_id = 1 THEN T1.[value] ELSE ISNULL(T2.[value], T3.[value]) END,
CASE WHEN T1.node_id = 1 THEN ISNULL(T2.[value], T3.[Value]) ELSE T1.[value] END
FROM
[db1].[t_value_history] T1
LEFT OUTER JOIN [db1].[t_value_history] T2 ON T2.source_time = T1.source_time
AND T2.node_id <> T1.node_id -- This join looks for a value for the other node at the same time.
LEFT OUTER JOIN [db1].[t_value_history] T3 ON T3.source_time < T1.source_time
AND T3.node_id <> T1.node_id -- If the previous join is empty, this looks for values for the other node at previous times
LEFT OUTER JOIN [db1].[t_value_history] T4 ON T4.source_time > T3.source_time
AND T4.source_time < T1.source_time
AND T4.node_id <> T1.node_id -- This join makes sure there aren't any more recent values
WHERE
T4.node_id IS NULL