Choosing Single Row from multiple rows - Teradata - teradata

Question

I have a requirement below to choose latest status.
Table 1:
Table2:
Results Expected:
Below is the logic we need.
SELECT
Table1.ID,
,CASE WHEN (Table1.hub=Table2.hub) THEN Table2.Status ELSE NULL END AS Original_Status
,CASE WHEN (Table1.hub<>Table2.hub AND Table2.Status like 'Found%' ) THEN Table2.hub ELSE NULL END AS Derived_Hub
,CASE WHEN (Table1.hub<>Table2.hub AND Table2.Status like 'Found%' ) THEN Table2.Status ELSE NULL END AS Derived_Status
from
Table1
Join Table2
ON (Table1.ID=Table2.ID)
With this code, I am getting 3 rows. If I put max in the above case statements, I am getting Hub3 instead of Hub2.
Could you please let me know how I can merge everything into single row. Please note that I want to show latest found Status from Table 2 even if there are 2 hubs with found status.
Let me know if you need any further information.

Answer 1

I think your best bet is to join into Table 2 twice. The first time, much like you are doing here, but using a LEFT OUTER JOIN on both id and hub. Then the second time on a derived version of table2 where you only select the top 1 records when sorted by timestamp in descending order:
SELECT
t1.id,
t1.hub as "Original Hub"
t2.status as "Original Hub Status",
t3.hub as "Found Hub",
t3.status as "Found Hub Status"
FROM
Table1 t1
LEFT OUTER JOIN Table2 t2 ON
t1.id = t2.id AND
t1.hub = t2.hub
LEFT OUTER JOIN
(
--Select a hub with the same id, that doesn't share the same hub number
--Only choose the top record when sorted by timestamp in descending order
SELECT TOP 1 id, hub, status
FROM table2
WHERE t1.hub <> table2.hub
ORDER BY TimeStamp Desc
) t3 ON
t1.id = t3.id
That subquery that has alias t3 is using a correlated subquery by referencing the t1.hub value in it's WHERE statement.

Answer 2

In TD14.10 you can utilize LAST_VALUE to access data from the "last" row. The CASEs are based on your query:
SELECT
t2.ID,
CASE WHEN t1.Hub = t2.Hub THEN t2.Hub END AS Original_Hub,
CASE WHEN t1.Hub = t2.Hub THEN t2.Status END AS Original_Status,
-- get the last Hub
LAST_VALUE(CASE WHEN t1.Hub <> t2.Hub AND Table2.Status like 'Found%'
THEN t2.Hub
END)
OVER (PARTITION BY t1.ID
ORDER BY CASE WHEN t1.Hub = t2.Hub THEN 0 ELSE 1 END, t2.TS
ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) AS Derived_Hub,
-- get the last Status
LAST_VALUE(CASE WHEN t1.Hub <> t2.Hub AND Table2.Status like 'Found%'
THEN t2.Status
END)
OVER (PARTITION BY t1.ID
ORDER BY CASE WHEN t1.Hub = t2.Hub THEN 0 ELSE 1 END, t2.TS
ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) AS Derived_Status
FROM
Table1 AS t1
JOIN Table2 AS t2
ON (t1.ID=t2.ID)
QUALIFY
ROW_NUMBER () -- return the 1st row only
OVER (PARTITION BY t1.ID
ORDER BY CASE WHEN t1.Hub = t2.Hub THEN 0 ELSE 1 END, t2.TS) = 1
EXPLAIN should combine all OVER into a single STATS-step as they use the same PARTITION BY and ORDER BY.