MS SQL Server : update with less amount of rows in source table - sql

Question

What is the best way to update Table1 with all values from Table2 if Table2 has less rows than Table1? This considering that Table2 has no key that can be joined to Table1 for update.
TABLE1 TABLE2 RESULT TABLE1
id value value id value
----------------------------------------------------
1 NULL 4 1 4
2 NULL 6 2 6
3 NULL 8 3 8
4 NULL 4 4
5 NULL 5 6
6 NULL 6 8
7 NULL 7 4
Hope I make sense.
Thanks in advance.
EDIT: Pardon, did not specify its Microsoft SQL Server 2012. :/
EXAMPLE for SOLUTION:
DECLARE #t1 TABLE(id int, avalue int)
DECLARE #t2 TABLE(id INT, avalue int)
-- Generate 20 rows in #t1 table
INSERT
INTO #t1 (id)
SELECT Number
FROM dbo.Numbers
WHERE Number BETWEEN 1 AND 20
-- Generate 5 rows and value #t2 table
INSERT
INTO #t2 (id,avalue)
SELECT Number,
Number
FROM dbo.Numbers
WHERE Number BETWEEN 1 AND 5
-- The goal is to take all rows from #t2
-- and repeatively insert them in order into #t1
UPDATE t1
SET t1.avalue = t2.avalue
FROM #t1 t1
JOIN ( SELECT t2.*, COUNT(*) OVER () AS cnt
FROM #t2 t2
) AS t2
ON (t1.id - 1) % t2.cnt = t2.id - 1;
SELECT *
FROM #t1

Answer 1

Interesting problem. This first solution is for MySQL (I originally read the question as being about that database). After this solution is the one for SQL Server.
You need to generate a join key. Let me assume that id is really sequential. Then you can use modulo arithmetic to do the match:
update table1 t1
(select (#rn := #rn + 1) as seqnum, value
from table2 cross join
(select #rn := -1) vars
) t2 cross join
(select count(*) as cnt from table2) cnt
on mod((t1.id - 1), cnt.cnt) = t2.seqnum
set t1.value = t2.value;
If the id in table1 is not sequential, you can use a variable for that as well. It just further complicates the query:
update table1 t1 join
(select #rn1 := #rn + 1) as seqnum, id
from table1 t1 cross join
(select #rn1 := 0) vars
order by id
) t1s
on t1.id = t1s.id join
(select (#rn := #rn + 1) as seqnum, value
from table2 cross join
(select #rn := -1) vars
) t2 cross join
(select count(*) as cnt from table2) cnt
on mod((t1s.seqnum - 1), cnt.cnt) = t2.seqnum
set t1.value = t2.value;
EDIT:
You can readily do the same thing in SQL Server. It is actually easier:
update table1 t1
set t1.value = t2.value;
from table1 t1 join
(select t2.*, count(*) over () as cnt
from table2 t2
) t2
on (t1.id - 1) % t2.cnt = (t2.id - 1);
This formulation depends on the ids being sequential with no gaps. It is easy enough to loosen this restriction, but the query gets a wee bit more complicated.

Answer 2

Try this query.
DECLARE #Table1 AS TABLE
(
ID INT,
Value INT
)
DECLARE #Table2 AS TABLE
(
Value INT
)
INSERT INTO #Table1
SELECT 1, NULL UNION
SELECT 2, NULL UNION
SELECT 3, NULL UNION
SELECT 4, NULL UNION
SELECT 5, NULL UNION
SELECT 6, NULL UNION
SELECT 7, NULL
INSERT INTO #Table2
SELECT 4 UNION
SELECT 6 UNION
SELECT 8
DECLARE #nCOUNT as INT
SET #nCOUNT = (SELECT COUNT(*) FROM #Table2)
UPDATE TB1 SET TB1.Value = TB2.Value FROM #Table1 AS TB1
INNER JOIN
(SELECT T1.ID, T2.Value FROM
(SELECT *, CASE WHEN (ROW_NUMBER() OVER(ORDER BY ID) % #nCOUNT) = 0
THEN #nCOUNT
ELSE (ROW_NUMBER() OVER(ORDER BY ID) % #nCOUNT)
END AS ROID
FROM #Table1) AS T1
LEFT JOIN (SELECT VALUE, ROW_NUMBER() OVER(ORDER BY Value) AS ID FROM #Table2) AS T2 ON T2.ID = T1.ROID) AS TB2
ON TB1.ID = TB2.ID
SELECT * FROM #Table1