Dude,

Do it all in a query with unions.
ie:

declare @tmp table ( id int identity(1,1), sumvalue numeric(12,2) )

insert into @tnp (sumvalue)
select sum(column_x)
from table_x
union all
select sum (column_y)
from table_y

select sum( sumvalue)
from @tmp

You can use the last select to put where ever you want.

Hope this helps.
AL


On Fri, Feb 27, 2009 at 4:21 AM, pantagruel <[email protected]>wrote:

>
> Hi,
>
> I have a situation where I have a number of queries in a stack. The
> number is generally 1, but it can theoretically be any amount.
>
> I need to execute the queries and then loop through the rows of each
> query and then combine these results, the final combination is written
> to a text file matching a particular archiving format.
> The queries do not return the same nunber of columns or rows.
> If query returns two rows 7 columns, query 2 returns 0 rows but has 4
> possible columns (if it did return anything), and query 3 returns 3
> rows 2 columns then the final output should be
>
> 3 rows 13 columns with the empty columns in the database replaced by
> particular characters.
>
> Currently I have an algorithm that does my output if I have 1 query.
>
> what I was thinking was:
>
> loop through number of queries, make new data adapter, loop through
> rows, read rows into some data structure where I can append a
> particular result by an index, for example an array and then at the
> end loop through that data structure and write it out.
>
> However I worry about that because of the potential size of what is
> returned - say a couple MB is not an unwarranted expectation that the
> choice of data structure becomes somewhat important. If I were to
> follow this method what data structure do you think I should use?
>
> The application does not need to have any particular good performance,
> but it can't crash under the strain of too much data either.
>
> Alternately if you can suggest something else that would be great, as
> the method outlined above seems inefficient. Although the best I can
> come up with under the constraints of the problem.
>
> Thanks
>
>
>

Reply via email to