Dynamic Tables Names and Temporary Tables Options

Firstly I consider myself quite an experienced SQL Server user, and
am
now using SQL Server 2005 Express for the main backend of my
software.

My problem is thus: The boss needs to run reports; I have designed
these reports as SQL procedures, to be executed through an ASP
application. Basic, and even medium sized (10,000+ records) reporting
run at an acceptable speed, but for anything larger, IIS timeouts and
query timeouts often cause problems.

I subsequently came up with the idea that I could reduce processing
times by up to two-thirds by writing information from each
calculation
stage to a number of tables as the reporting procedure runs..
ie. stage 1, write to table xxx1,
stage 2 reads table xxx1 and writes to table xxx2,
stage 3 reads table xxx2 and writes to table xxx3,
etc, etc, etc
procedure read final table, and outputs information.

This works wonderfully, EXCEPT that two people can't run the same
report at the same time, because as one procedure creates and writes
to table xxx2, the other procedure tries to drop the table, or read a
table that has already been dropped...

Does anyone have any suggestions about how to get around this
problem?
I have thought about generating the table names dynamically using
'sp_execute', but the statement I need to run is far too long
(apparently there is a maximum length you can pass to it), and even
breaking it down into sub-procedures is soooooooooooooooo time
consuming and inefficient having to format statements as strings
(replacing quotes and so on)

How can I use multiple tables, or indeed process HUGE procedures,
with
dynamic table names, or temporary tables?

All answers/suggestions/questions gratefully received.

Thanksbrstowe wrote:

Quote:

Originally Posted by

How can I use multiple tables, or indeed process HUGE procedures,
with
dynamic table names, or temporary tables?

Isn't this your answer right here? Just CREATE TABLE #foo instead
of CREATE TABLE foo, etc. and each process will get their own local
version of the #foo table.|||On Oct 5, 7:07 pm, Ed Murphy <emurph...@.socal.rr.comwrote:

Quote:

Originally Posted by

brstowe wrote:

Quote:

Originally Posted by

How can I use multiple tables, or indeed process HUGE procedures,
with
dynamic table names, or temporary tables?

>
Isn't this your answer right here? Just CREATE TABLE #foo instead
of CREATE TABLE foo, etc. and each process will get their own local
version of the #foo table.

please be honest and tell me if I have completely misunderstood...?
the temporary tables are session/process/user specific...

two users could essentially create/drop tables, and populate/delete
information from tables with exactly the name e.g. #temptable1 at
exactly the same time?

thanks

dynamic sql to loop over fiscal years

thanks for reading.

i'm interested in improving the format of this query. consider me clueless today, if you will. :) how can i fix this to make it dynamically move over the years? is there something i can do with set manipulation that is smarter than this?

the goal of this query is to return cases per year, where "year" is defined as (Oct 1, YYYY - Sep 30, YYYY+1) instead of the typical YYYY

problem is, i have to write it as some cludgy dynamic sql looping over an incremented year. i don't know of any other way.

again, thanks for reading ... and any help in advance.

SELECT count(*) as 'Data Points', '2001' as 'Experiment Year'
FROM tbl_experiment_data

WHERE start_date BETWEEN '9/30/2001' AND '10/01/2002'
and completion_date BETWEEN '9/30/2001' AND '10/01/2002'
and status = 'CaseClosed'

UNION

SELECT count(*) as 'Data Points', '2002' as 'Experiment Year'
FROM tbl_experiment_data

WHERE start_date BETWEEN '9/30/2002' AND '10/01/2003'
and completion_date BETWEEN '9/30/2002' AND '10/01/2003'
and status = 'CaseClosed'

UNION

...

expected output...

Data Points______ Experiment Year
32_____________ 2001
102____________ 2002
... ...Create a table called ExperimentYears, populate it with ExperimentYear char(4), YearStart datetime, YearEnd datetime. Then do just one SELECT similar to yours:

SELECT count(*) as [Data Points], ExperimentYear
from tbl_experiment_data ted
inner join ExperimentYears ey
on ted.start_date between ey.YearStart and ey.YearEnd
and ted.completion_date between ey.YearStart and ey.YearEnd
where ted.status = 'CaseClosed'
group by ey.ExperimentYear|||We use a separate calendar table. something like this:

create table FiscalCalendar
(FiscalYear int,
StartDate datetime,
EndDate datetime)

This should reduce your query to something like

select count(*), fc.FiscalYear
from tbl_experiment_data a, FiscalCalendar fc
where a.startdate between fc.startdate and fc.enddate
and a.enddate between fc.startdate and fc.enddate
group by fc.fiscalyear

Been a while since I messed with this, so experiment with this for a bit. As a curiosity, what happens to experiments that start in one fiscal year and end in the next?|||as i'm re-reading my post now i can see that maybe it wouldn't even work as is because i have the count(*) without a 'group by'

still, i hope these sorts of mistakes can be overlooked as i ask for help.

it also occurred to me just now that maybe i could use a user-defined function that returns the value of the year as redefined by the "year" range above.

that way i could rewrite the query like this...

==============
SELECT count(*) as 'Data Points', getFiscalYear(start_date) as 'Experiment Date'
FROM tbl_experiment_data

WHERE DATEPART(YEAR, start_date) = getFiscalYear(start_date)
and DATEPART(YEAR, completion_date) = getFiscalYear(completion_date) and status = 'CaseClosed'

GROUP BY getFiscalYear(start_date)

ORDER BY getFiscalYear(start_date)

==============

any comments? criticisms? other ideas?

thanks again for reading ... and your input

oh, great! i just noticed the responses now too. thank you. i'll try these ideas out.|||SELECT count(*) as 'Data Points',
year(dateadd(d, 92, StartDate))-1 as 'Experiment Year'
FROM tbl_experiment_data
WHERE status = 'CaseClosed'

The year(dateadd(d, 92, StartDate))-1 function returns the experiment year by addint 92 days (Oct +Nov +Dec) and then subtracting 1 from the year. Note that if you just subtracted days you would have to account for leap years.

You will need to decide what to do if an experiment starts in one year and ends in the next. Your original code would skip those instances entirely.