interesting t-sql exercise - sql

Question

I am trying to resolve t-sql exercise
I need to update first table with values from second by joining by id. If I can not join then use value from default ID (default iD is the Id that is null)
please run it to see it
declare #t as table (
[id] INT
,val int
)
insert into #t values (null, null)
insert into #t values (2, null)
insert into #t values (3, null)
insert into #t values (4, null)
declare #t2 as table (
[id] INT
,val int
)
insert into #t2 values (null, 11)
insert into #t2 values (2, 22)
insert into #t2 values (3, 33)
select * from #t
select * from #t2
update t
set t.val = t2.val
from #t as t join #t2 as t2
on t.id = t2.id
or
(
(t.id is null or t.id not in (select id from #t2))
and t2.id is null
)
select * from #t
here is result
--#t
id val
---------------
NULL NULL
2 NULL
3 NULL
4 NULL
--#t2
id val
---------------
NULL 11
2 22
3 33
--#t after update
id val
---------------
NULL 11
2 22
3 33
4 NULL
how to make val in last row equal 11?
4 11

Answer 1

This solution left joins to t2 twice and then does a coalesce.
The ON on the second join matches on records that failed on the join and then looks for the "Default" case.
UPDATE t
set t.val = COALESCE(t2.val,t3.val)
from #t as t
LEFT join #t2 as t2
on t.id = t2.id
LEFT JOIN #t2 t3
ON t2.id is null and t3.id is null
See it working here

Answer 2

try this for the update...
update t
set t.val = t2.val
from #t as t join #t2 as t2
on t.id = t2.id
or
(
(t.id is null or not exists (select * from #t2 where id = t.id))
and t2.id is null
)
Problem is with not in operator and null values. This would also work...
update t
set t.val = t2.val
from #t as t join #t2 as t2
on t.id = t2.id
or
(
(t.id is null or t.id not in (select id from #t2 where id is not null))
and t2.id is null
)

Answer 3

Here's a technique that may help.
Start with the kind of simple code you want to be writing:
MERGE INTO #t AS target
USING source
ON target.id = source.id
WHEN MATCHED THEN
UPDATE
SET val = source.val;
Then write a table expression (source) that satisfies the requirements.
Requirement 1: "joining by id"
-- simple existential quantification e.g.
SELECT id, val
FROM #t2 AS T2
WHERE id IN ( SELECT id FROM #t )
Requirement 2: "If I can not join then use value from default ID (default iD is the Id that is null)"
-- first find the id values in the target that do not exist in the source:
SELECT id
FROM #t
EXCEPT
SELECT id
FROM #t2
then cross join the result with the row from the source where Id is null:
SELECT DT1.id, T2.val
FROM ( SELECT id
FROM #t
EXCEPT
SELECT id
FROM #t2 ) AS DT1,
#t2 AS T2
WHERE T2.id IS NULL
At this point you will want to query some test data to ensure each query satisfies its respective requirement.
Union the above two results to form a single table expression:
SELECT id, val
FROM #t2 AS T2
WHERE id IN ( SELECT id
FROM #t )
UNION
SELECT DT1.id, T2.val
FROM ( SELECT id
FROM #t
EXCEPT
SELECT id
FROM #t2 ) AS DT1,
#t2 AS T2
WHERE T2.id IS NULL
Then plug the table expression into the MERGE boilerplate code:
WITH source
AS
(
SELECT id, val
FROM #t2 AS T2
WHERE id IN ( SELECT id
FROM #t )
UNION
SELECT DT1.id, T2.val
FROM ( SELECT id
FROM #t
EXCEPT
SELECT id
FROM #t2 ) AS DT1,
#t2 AS T2
WHERE T2.id IS NULL
)
MERGE INTO #t AS target
USING source
ON target.id = source.id
WHEN MATCHED THEN
UPDATE
SET val = source.val;