dynamic operations with 2 tables and 1 insert SQL Server

Question

I have 2 tables:

table1

id someval someatt
-------------------
1 23 name1
2 56 name2
3 76 name3
4 456 name4
5 53 name5
6 67 name6
7 12 name7
8 43 name8
9 99 name9
10 567 name10

table2

 id someval someatt
 -------------------
 1 23.3 name1
 2 5.6 name2
 3 8.76 name3
 4 4.56 name4
 5 5.3 name5
 6 6.7 name6
 7 1.2 name7
 8 4.3 name8
 9 9.9 name9
 10 56.7 name10

I need to insert to a new table some operations of various fields of both tables so, for example

iteration1
x = get value from table1 where id 1 
y = get value from table1 where id 2
a = get value from table2 where id 1
b = get value from table2 where id 2
iteration2
r = get value from table1 where id 2
s = get value from table1 where id 3
u = get value from table2 where id 2
v = get value from table2 where id 3
iterationn.
q = get value from table1 where id n-1
p = get value from table1 where id n
o = get value from table2 where id n-1
e = get value from table2 where id n

Then insert into NEWTABLE

(a*b + x+y), (r*s + u*v) ...(q*p+o*e) 

sO I was thinking (IF I have to do this 100 times):

SET @counter = 1;
SET @template = '';
SET @N = 100;
WHILE(@counter < @N)
 Select @x = value from table 1 where id = @counter
 Select @y = value from table 1 where id = @counter + 1
 Select @a = value from table 2 where id = @counter
 Select @b = value from table 2 where id = @counter + 1
 @template = @template + ' (@x*@y + @a*@b), '
end

so at the end of the loop I have a full template

and then EXEC(@template) ,

• How could you optimize this?
• How to do something like I described? (table1 and table2 are samll examples of my tables)
• Is there a way The 4 select statements are avoided?

output table:

result1 result2 result3 .... result100
---------------------------------------
 float float float ... float 

where float is a result of calculating (a*b + x+y) for each value in table1 and table2

• mellamokb: Why do you want the results horizontal? That's better handled by the application than by the database layer.
• edgarmtze: well, the reason is that I will insert results the same way as described so, I will get a table with 100 columns and n rows, where a row is calculated as described
• mellamokb: The code you provided only calculates a single row. Do you mean you will repeat EXEC(@template) over and over until you are down to a single value?
• edgarmtze: Not EXEC(@template), template will changue and take other values from other tables and keep inserting in that new table

Answer 1

This will work with SQL 2005 & up, using the same data you listed.

SELECT t1x.id AS smaller_id,
 ( t1x.someval * t1y.someval ) + ( t2a.someval * t2b.someval ) AS result
FROM dbo.table1 AS t1x
 INNER JOIN dbo.table1 AS t1y ON t1x.id + 1 = t1y.id
 INNER JOIN dbo.table2 AS t2a ON t2a.id = t1x.id
 INNER JOIN dbo.table2 AS t2b ON t2b.id = t1y.id

If you want it horizontal, you'll have to PIVOT the results.

SELECT 'Calculation Result' AS CalcRes,
 [result1], [result2], [result3], [result4], [result5], [result6], [result7], [result8], [result9]
FROM (
 SELECT 'result' + cast(t1x.id AS varchar(4)) AS result_name, 
 ( t1x.someval * t1y.someval ) + ( t2a.someval * t2b.someval ) AS result
 FROM dbo.table1 AS t1x
 INNER JOIN dbo.table1 AS t1y ON t1x.id + 1 = t1y.id
 INNER JOIN dbo.table2 AS t2a ON t2a.id = t1x.id
 INNER JOIN dbo.table2 AS t2b ON t2b.id = t1y.id
 ) AS Results
PIVOT (
 min( result ) --there needs to be an aggregate function here
 FOR result_name IN ([result1], [result2], [result3], [result4], [result5], [result6], [result7], [result8], [result9])
) AS PivotedResults

You'll need to use dynamic SQL for this, because you need to know the names of your result colums before you pivot.

Answer 2

Join all the records together:

insert
 NEWTABLE (theid, newval)
select
 t1_p.id as theid
 (t2_p.someval * t2_n.someval) + (t1_p.someval * t1_n.someval) as newval
from
 Table1 t1_p
inner join
 Table1 t1_n on t1_p.id + 1 = t1_n.id
inner join
 Table2 t2_p on t1_p.id = t2_p.id
inner join
 Table2 t2_n on t1_n.id = t2_n.id

• edgarmtze: So if Now I have to do this 100 times and insert each of this results into a new table, How does the query look like?? Dynamic SQL?
• mellamokb: @darkcminor: No, you would only run this query once. It does the operation for all of the (100+) rows in a single query. Updated with an example how you would insert.
• mellamokb: @darkcminor: Can you provide an example output what NEWTABLE should look like? Then I can give a more specific answer as well.

Answer 3

Not sure I understand your question fully but....

;with cte1 as
(
 select
 T1_1.id,
 T1_1.someval as x,
 T1_2.someval as y
 from @T1 as T1_1
 inner join @T1 as T1_2
 on T1_1.id = T1_2.id-1
),
cte2 as
(
 select
 T2_1.id,
 T2_1.someval as a,
 T2_2.someval as b
 from @T2 as T2_1
 inner join @T2 as T2_2
 on T2_1.id = T2_2.id-1
)
select (a*b + x*y)
from cte1 as T1
 inner join cte2 as T2
 on T1.id = T2.id

Answer 4

Might you try something like

 INSERT INTO SomeTable
 SELECT
 (x.value * y.value) + (a.value * b.value)
 FROM
 Table1 x 
 Table1 y on y.id = x.id + 1
 Table2 a on a.id = x.id
 Table2 b on b.id = x.id + 1
 WHERE
 x.Id = @counter

• edgarmtze: One question...In your query how does @counter variable is incremented, and if lets say @counter depends on another variable???