Existing SQL Server 2008 script improvement - sql

Question

SQL Server 2008
I have two tables with OrderIds and ItemIds. I need a resulting table with each OrderId from forst table linked with OrderId from second table where the number of identical ItemIds is maximum.
I did a script that does this using two loops but if the number of OrderIds in those tables is big (~1000) it means the loop has to be run 1000x1000 times, which might be too long. Ca this be achieved in a better way?
See my below my already written script:
drop table #Match, #OrderRec, #OrderSent
create table #Match(
OrderIdRec int NULL,
OrderIdSent int NULL)
create table #OrderRec(
OrderIdRec int NOT NULL,
ItemId int NULL)
create table #OrderSent(
OrderIdSent int NOT NULL,
ItemId int NULL)
insert #OrderRec values (1, 1)
insert #OrderRec values (1, 5)
insert #OrderRec values (1, 7)
insert #OrderRec values (1, 4)
insert #OrderRec values (1, 15)
insert #OrderRec values (1, 10)
insert #OrderRec values (2, 21)
insert #OrderRec values (2, 15)
insert #OrderRec values (2, 21)
insert #OrderRec values (2, 26)
insert #OrderRec values (5, 4)
insert #OrderRec values (5, 3)
insert #OrderRec values (5, 12)
insert #OrderRec values (5, 1)
insert #OrderSent values (121, 1)
insert #OrderSent values (121, 2)
insert #OrderSent values (121, 5)
insert #OrderSent values (121, 10)
insert #OrderSent values (121, 9)
insert #OrderSent values (122, 6)
insert #OrderSent values (122, 7)
insert #OrderSent values (122, 9)
insert #OrderSent values (122, 11)
insert #OrderSent values (142, 1)
insert #OrderSent values (142, 12)
insert #OrderSent values (142, 4)
insert #OrderSent values (142, 11)
set nocount on
declare #OrderIdRec int,
#OrderIdSent int,
#cnt numeric(10),
#cnt_max numeric(10),
#OrderIdSentMax int
select #OrderIdRec = MIN(OrderIdRec)
from #OrderRec
while ISNULL(#OrderIdRec,0) > 0
begin
select #OrderIdSent = MIN(OrderIdSent)
from #OrderSent
set #cnt_max = 0
set #OrderIdSentMax = NULL
while ISNULL(#OrderIdSent,0) > 0
begin
set #cnt = 0
select #cnt = COUNT(*)
from #OrderRec r
inner join #OrderSent t
on t.ItemId = r.ItemId
where r.OrderIdRec = #OrderIdRec
and t.OrderIdSent = #OrderIdSent
if isnull(#cnt, 0) > #cnt_max
begin
set #cnt_max = #cnt
set #OrderIdSentMax = #OrderIdSent
end
select #OrderIdSent = MIN(OrderIdSent)
from #OrderSent
where OrderIdSent > #OrderIdSent
end
insert #Match(
OrderIdRec,
OrderIdSent)
values (#OrderIdRec, #OrderIdSentMax)
select #OrderIdRec = MIN(OrderIdRec)
from #OrderRec
where OrderIdRec > #OrderIdRec
end
select *
from #Match
order by OrderIdRec
The actual script starts with set nocount on, what is before is just to create a set of data to play with.
The result is:
OrderIdRec OrderIdSent
1 121
2 NULL
5 142

Answer 1

;WITH s AS
(
SELECT OrderIdRec, OrderIdSent,
rn = ROW_NUMBER() OVER (PARTITION BY OrderIdRec ORDER BY c DESC)
FROM
(
SELECT r.OrderIdRec, s.OrderIdSent,
c = COUNT(*) OVER (PARTITION BY r.OrderIdRec, s.OrderIdSent)
FROM #OrderRec AS r
INNER JOIN #OrderSent AS s
ON r.ItemId = s.ItemId
) AS s2
),
d AS (SELECT OrderIdRec FROM #OrderRec GROUP BY OrderIdRec)
SELECT d.OrderIdRec, s.OrderIdSent
FROM d LEFT OUTER JOIN s
ON d.OrderIdRec = s.OrderIdRec AND s.rn = 1
ORDER BY d.OrderIdRec;

Answer 2

The following query gets the counts for all pairs between the two tables:
select orec.OrderId, osent.OrderId, count(*) as cnt
from OrderRec orec join
OrderSent osent
on orec.itemId = osent.itemId
group by orec.OrderId, osent.OrderId;
The following gets the highest cnt value for each orec.OrderId:
select oo.*
from (select orec.OrderId, osent.OrderId, count(*) as cnt,
row_number() over (partition by orec.OrderId, osent.OrderId order by count(*) desc
) as seqnum
from OrderRec orec join
OrderSent osent
on orec.itemId = osent.itemId
group by orec.OrderId, osent.OrderId
) oo
where seqnum = 1;