Filling NULL values with preceding Non-NULL values using FIRST_VALUE - sql

Question

I am joining two tables.
In the first table, I have some items starting at a specific time. In the second table, I have values and timestamps for each minute in between the start and end time of each item.
First table
UniqueID Items start_time
123 one 10:00 AM
456 two 11:00 AM
789 three 11:30 AM
Second table
UniqueID Items time_hit value
123 one 10:00 AM x
123 one 10:05 AM x
123 one 10:10 AM x
123 one 10:30 AM x
456 two 11:00 AM x
456 two 11:15 AM x
789 three 11:30 AM x
So When joining the two tables I have this:
UniqueID Items start_time time_hit value
123 one 10:00 AM 10:00 AM x
123 null null 10:05 AM x
123 null null 10:10 AM x
123 null null 10:30 AM x
456 two 11:00 AM 11:00 AM x
456 null null 11:15 AM x
789 three 11:30 AM 11:30 AM x
I'd like to replace these null values with the values from the non-null precedent row...
So the expected result is
UniqueID Items start_time time_hit value
123 one 10:00 AM 10:00 AM x
123 one 10:00 AM 10:05 AM x
123 one 10:00 AM 10:10 AM x
123 one 10:00 AM 10:30 AM x
456 two 11:00 AM 11:00 AM x
456 two 11:00 AM 11:15 AM x
789 three 11:30 AM 11:30 AM x
I tried to build my join using the following function without success:
FIRST_VALUE(Items IGNORE NULLS) OVER (
PARTITION BY time_hit ORDER BY time_hit
ROWS BETWEEN CURRENT ROW AND
UNBOUNDED FOLLOWING) AS test
My question was a bit off. I found out that UniqueID were inconsistent that is why I had these null values in my output. So the validated answer is a good option to fill null-values when joining two tables and one of your tables has more unique rows than the other.

Answer 1

You could use first_value (but last_value would also work too in this scenario). The import part is to specify rows between unbounded preceding and current row to set the boundaries of the window.
Answer updated to reflect updated question, and preference for first_value
select
first_value(t1.UniqueId ignore nulls) over (partition by t2.UniqueId
order by t2.time_hit
rows between unbounded preceding and current row) as UniqueId,
first_value(t1.items ignore nulls) over (partition by t2.UniqueId
order by t2.time_hit
rows between unbounded preceding and current row) as Items,
first_value(t1.start_time ignore nulls) over (partition by t2.UniqueId
order by t2.time_hit
rows between unbounded preceding and current row) as start_time,
t2.time_hit,
t2.item_value
from table2 t2
left join table1 t1 on t1.start_time = t2.time_hit
order by t2.time_hit;
Result
| UNIQUEID | ITEMS | START_TIME | TIME_HIT | ITEM_VALUE |
|----------|-------|------------|----------|------------|
| 123 | one | 10:00:00 | 10:00:00 | x |
| 123 | one | 10:00:00 | 10:05:00 | x |
| 123 | one | 10:00:00 | 10:10:00 | x |
| 123 | one | 10:00:00 | 10:30:00 | x |
| 456 | two | 11:00:00 | 11:00:00 | x |
| 456 | two | 11:00:00 | 11:15:00 | x |
| 789 | three | 11:30:00 | 11:30:00 | x |
SQL Fiddle Example
Note: I had to use Oracle in SQL Fiddle (so I had to change the data types and a column name). But it should work for your database.

Answer 2

One alternative solution would be to use a NOT EXISTS clause as JOIN condition, with a correlated subquery that ensures that we are relating to the relevant record.
SELECT t1.items, t1.start_time, t2.time_hit, t2.value
FROM table1 t1
INNER JOIN table2 t2
ON t1.items = t2.items
AND t1.start_time <= t2.time_hit
AND NOT EXISTS (
SELECT 1 FROM table1 t10
WHERE
t10.items = t2.items
AND t10.start_time <= t2.time_hit
AND t10.start_time > t1.start_time
)
Demo on DB Fiddle:
| items | start_time | time_hit | value |
| ----- | ---------- | -------- | ----- |
| one | 10:00:00 | 10:00:00 | x |
| one | 10:00:00 | 10:05:00 | x |
| one | 10:00:00 | 10:10:00 | x |
| one | 10:00:00 | 10:30:00 | x |
| two | 11:00:00 | 11:00:00 | x |
| two | 11:00:00 | 11:15:00 | x |
| three | 11:30:00 | 11:30:00 | x |
Alternative solution to avoid using EXISTS on a JOIN condition (not allowed in Big Query): just move that condition to the WHERE clause.
SELECT t1.items, t1.start_time, t2.time_hit, t2.value
FROM table1 t1
INNER JOIN table2 t2
ON t1.items = t2.items
AND t1.start_time <= t2.time_hit
WHERE NOT EXISTS (
SELECT 1 FROM table1 t10
WHERE
t10.items = t2.items
AND t10.start_time <= t2.time_hit
AND t10.start_time > t1.start_time
)
DB Fiddle

Answer 3

I guess you are expecting an output by using INNER JOIN. But not sure why you used FIRST_VALUE.
SELECT I.Item, I.Start_Time, ID.Time_hit, ID.Value
FROM Items I
INNER JOIN ItemDetails ID
ON I.Items = ID.Items
Please explain if you are looking for any specific reasons to look over this approach.