Temp Table Vs. Union: Which Has Better Performance?
Aug 13, 2007
Right now, a client of mine has a T-SQL statement that does the
following:
1) Create a temp table.
2) Populate temp table with data from one table using an INSERT
statement.
3) Populate temp table with data from another table using an INSERT
statement.
4) SELECT from temp table.
Would it be more efficient to simply SELECT from table1 then UNION
table 2? The simply wants to see the result set and does not need to
re-SELECT from the temp table.
This may be a dumb question, but I can't seem to get the syntax right. I have two temp tables that have the same columns, I want to do a union on them and store the results in a temp table. Any ideas?
Ie.
select * from #tmpTable1 union select * from #tmpTable2 into #tmpTable3
I'm having trouble creating a temp table out of a select statement that uses multipe union alls.
Here's what I have, I'm trying to get the results of this query into a temp table...
select parent, (select cst_id from co_customer (nolock) where cst_key = Parent) as cst_id, (select cst_name_cp from co_customer (nolock) where cst_key = Parent) as cst_name_cp, (select org_total_assets_ext from dbo.co_organization_ext where org_cst_key_ext = parent) as Parent_Total_assets, sum(own_assets) as Total_child_own_assets
from ( Select parent, Child, (select org_own_assets_ext from dbo.co_organization_ext where org_cst_key_ext = child) as Own_assets
from (Select Cst_key as Child, dbo.return_org_parent(cst_key,0,1) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,1) is not null union all
Select Cst_key as Child, dbo.return_org_parent(cst_key,0,2) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,2) is not null union all
Select Cst_key as Child, dbo.return_org_parent(cst_key,0,3) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,3) is not null union all
Select Cst_key as Child, dbo.return_org_parent(cst_key,0,4) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,4) is not null union all
Select Cst_key as Child, dbo.return_org_parent(cst_key,0,5) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,5) is not null union all
Select Cst_key as Child, dbo.return_org_parent(cst_key,0,6) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,6) is not null union all Select Cst_key as Child, dbo.return_org_parent(cst_key,0,7) as Parent from co_customer (nolock) where cst_type = 'Organization' and cst_delete_flag = 0 and dbo.return_org_parent(cst_key,0,7) is not null )as c ) as d
group by parent
having sum(own_assets) <> (select org_total_assets_ext from dbo.co_organization_ext where org_cst_key_ext = parent)
I have a query which does 3 selects and Union ALLs each to get a final result set. The performance is unacceptable - takes around a minute to run. If I remove the Union All so that the result sets are returned individually it returns all 3 from the query in around 6 seconds (acceptable performance).
Any way to join the result sets together without using Union All.
Each result set has exactly the same structure returned...
Query below [for reference]...
WITH cte AS ( SELECT A.[PoleID], ISNULL(B.[IsSpanClear], 0) AS [IsSpanClear], B.[SurveyDate], ROW_NUMBER() OVER (PARTITION BY A.[PoleID] ORDER BY B.[SurveyDate] DESC) rownum FROM[UT_Pole] A LEFT OUTER JOIN [UT_Surveyed_Pole] B ON A.[PoleID] = B.[PoleID]
The function is basically a set of Case Statements and various calculations dependant upon the Case.
Is this the best (performance wise) way to do it or should I dump the needed info in a Temp Table and do the calcs on it and then tie the select statement to the table.
I've seen both approaches done, but they both seem to be a different way of getting to the same conclusion. I'm just wondering which puts the lightest load on the server.
I have 2 temporary tables from a previous operation, Tab1 and Tab2, with the same dimensions. How do I create a third table Tab3 with the same dimensions containing the the combined rows of the 2 previous tables? TIA!
Hi, I have follwing union query. I want to put this all in a temp table.
select Store_Id,batchnumber From Adjustments where updatedDt between '10/30/2007' and '11/20/2007' and Store_id in(8637 ,8641) group by Store_Id, batchnumber Union select DestinationId,b.batchNumber from batch b inner join Carton C on C.Carton_Id = b.General_ID inner join Document d on d.Document_Id = c.Document_Id where b.BatchType = 'Warehouse' and b.TranTable = 'Carton' and (d.DestinationId in (8637 ,8641) ) and c.UpdatedDt Between '10/30/2007' and '11/20/2007' Union select d.DestinationId,b.Batchnumber From batch b inner join Document d on d.Document_Id = b.General_Id where b.BatchType = 'TransferIn' and b.TranTable = 'Document' and (d.DestinationId in (8637,8641) ) and d.UpdatedDt Between'10/30/2007' and '11/20/2007' Union select d.SourceId,b.batchNumber From batch b inner join Document d on d.Document_Id = b.General_Id where b.BatchType = 'TransferOut' and b.TranTable = 'Document' and (d.SourceId in (8637,8641) ) and d.UpdatedDt Between'10/30/2007' and '11/20/2007' order by batchnumber
I'm trying to summarize costs assigned to active jobs for a manufacturing business. I need to aggregate work in process (WIP) cost that resides in labor-transaction and part-transaction tables based on transaction types, and transaction dates. Some transactions increase the WIP cost of the job while others decrease WIP. The business needs to see how much $$ is tied up in each job as of a particular date -- the calculation is: ToDate (cost of materials and labor assigned to job) - ToInv (cost of materials returned to inventory) - ToSales (cost of materials sold).
I developed this query incrementally and, so far, the #ToDate, #ToInv, and #ToSales temp tables seem to be populating with the correct data. My thought was to combine these three tables with a UNION and then extract the grand totals and here's where I started getting the following error: ------------------------------------------ Incorrect syntax near the keyword 'UNION'. ------------------------------------------ The problem is with the UNIONs going into #myTotal.
I would appreciate any help with this. Also, please let me know if you can suggest a better design for this. Thanks!
--M&S To Date SELECT pt.jobnum, SUM(pt.extcost) AS Cost FROM parttran pt JOIN jobhead jh ON pt.jobnum=jh.jobnum WHERE trantype IN ( <valid trans types> ) AND jh.JobReleased = 1 AND pt.TranDate < '2007-9-30' GROUP BY pt.jobnum
UNION -- This one works ok.
--L&B To Date SELECT jh.JobNum, sum(l.LaborRate*l.LaborHrs) + sum(l.BurdenRate*l.BurdenHrs) AS Cost FROM LaborDtl l JOIN JobHead jh ON l.JobNum = jh.JobNum WHERE jh.JobReleased = 1 AND l.PayrollDate < '2007-9-30' GROUP BY jh.JobNum
SELECT pt.jobnum, SUM(pt.extcost) AS ToInv FROM parttran pt JOIN jobhead jh ON pt.jobnum=jh.jobnum WHERE trantype IN (<valid trans types>) AND jh.JobReleased = 1 AND pt.TranDate < '2007-9-30' GROUP BY pt.jobnum
SELECT pt.jobnum, SUM(pt.extcost) AS ToInv FROM parttran pt JOIN jobhead jh ON pt.jobnum=jh.jobnum WHERE trantype IN (<valid trans types>) AND jh.JobReleased = 1 AND pt.TranDate < '2007-9-30' GROUP BY pt.jobnum
I have a view which uses UNION of two tables. First table has a 1.5 Million records and the second one has 40,000 records. When I query the view with a column (that is indexed in both tables) in the where clause, it's taking taking 3 Minutes to give the result. The column is of DateTime data Type. Any ideas as to how to improve the query performance ???
I have been researching some performance problems in a very large application and I have a couple of questions about temp tables. (SQL 7.0 SP2)
I have one large procedure that I have been using as a test case. Originally this procedure was a cursor with lots of processing steps involving writing to, reading from and deleting in temp tables inside the cursor. I remember reading that temp tables inside a cursor were a potential performance problem, so I rewrote the procedure, replacing the cursor with a While Loop.
Doing this showed no increase in performance. Since Profiler was showing .5 second duration times on statements in the procedure accessing the temp tables I tested some more. I moved all the create statements to the top of the procedure, as I know these statements after processing steps can cause recompiles to happen. Still no performance increase.
Finally I replaced all the temp tables with actual tables, just to see what would happen. With no other changes the performance increased by more than 500%.
Can someone give me some clues as to what is happening here, because if this is a symptom of something I don't understand, the potential performance problems from other places where temp tables are similarly used in the application are enormous.
I think this is a very simple question, however, I don't know the answer. What is the difference between a regular Temp table and a Global Temp table? I need to create a temp table within an sp that all users will use. I want the table recreated each time someone accesses the sp, though, because some of the same info may need to be inserted and I don't want any PK errors.
i am inserting something into the temp table even without creating it before. But this does not give any compilation error. Only when I want to execute the stored procedure I get the error message that there is an invalid temp table. Should this not result in a compilation error rather during the execution time.?
--create the procedure and insert into the temp table without creating it. --no compilation error. CREATE PROC testTemp AS BEGIN INSERT INTO #tmp(dt) SELECT GETDATE() END
only on calling the proc does this give an execution error
Simple example: declare @tTable(col1 int) insert into @tTable(col1) values (1) select * from @tTable
Works perfectly in SQL Server Management Studio and the database connection is OK to as I may generate PP table using complex (or simple) queries without difficulty.
But when trying to get this same result in a PP table I get an error, idem when replacing table variable by a temporary table.
Message: OLE DB or ODBC error. .... The current operation was cancelled because another operation the the transaction failed.
If on the source I have a new column, the script generated by SqlPackage.exe recreates the table on the background with moving the data into a temp storage. If the table is big, such approach can cause issues.
Example of the script is below: in the source project I added columns [MyColumn_LINE_1] and [MyColumn_LINE_5].
Is there any way I can make it generating an alter statement instead?
BEGIN TRANSACTION; SET TRANSACTION ISOLATION LEVEL SERIALIZABLE; SET XACT_ABORT ON; CREATE TABLE [dbo].[tmp_ms_xx_MyTable] ( [MyColumn_TYPE_CODE] CHAR (3) NOT NULL,
[Code] ....
The same script is generated regardless the table having data or not, having a clustered or nonclustered PK.
The SP UserPersist_GetByCriteria does a "SELECT * FROM tbl_User WHERE gender = @Gender AND culture = @Culture", so why am I receiving this error when both tables have the same structure?
The error is being reported as coming from UserPersist_GetByCriteria on the "SELECT * FROM tbl_User" line.
Is there a way to attach a single record to a Select expression without another table?
Here's my case: I am developing a Select query for a report. The report has to show a row for each month, but not all months are necessarily represented in the data, so the result is a report with some of the months missing. Can't have that. To "force" the appearance of all months, I'd like to UNION my Select query to "dummy" records for each month with zeros in the data fields, or something like that. But I'd like to avoid creating a table just to hold the months for the Union. Is there a way to attach records to the result of a Select without those records coming from a table? I'd like to be able to say something like this:
"Do this SELECT. Now, append a row with the following values..."
I want to insert the data from temp table to other table. Only condition is, it needs to sorted based on tool number and tool date. For example if we have ten records for tool number 1000, it should be order by tool number and then based on tool_dt. Both tables doesn't have any primary keys. Please find below my code. I removed all the unnecessary columns for simple understanding. INSERT INTO tool_summary (tool_nbr, tool_dt) select tool_nbr, tool_dt from #tool order by tool_nbr, tool_dt...But this query is not working as expected. Data is getting shuffled.
WE have a job that loads data from an Oralce DB into our SQL Server 2000 DB twice a day. The schedule has just changed so that now there is a possibility of having my west coast users impacted when it runs at 5 PM PST and my east coast users impacted when it runs at 7 AM EST. As a workaround, I have developed a DTS package that loads the data into temp tables instead of the real tables. IE. Oracle -> XTable_temp instead of Oracle -> XTable. The load sometimes takes about an hour to an hour and a half to load, so this solution works great, but I want to then lock the table, delete it and rename the temp table to table X. The pseudo code would be:
Begin Transaction
Lock Table XTable
Drop XTable
Alter Table XTable_temp rename to XTable
Release Lock XTable
End Transaction
Create XTable_temp
I see two issues with this solution. 1) I think if I can lock XTable that the lock would be released when the table is dropped and the XTable_temp was being renamed. 2) I can't find a command to rename a table.
I want to pass the 'inserted' table from a trigger into an SP, I think I need to do this by dumping inserted table into a temporary table and passing the temp table. However, I need to do this for many tables, and don't want to list all the column names for each table/trigger (maintenance nightmare).
Can I dump the 'inserted' table to a temp table WITHOUT specifying the column names?
I have three tables "UNIONED" and I want the this inserted into a table.
INSERT INTO mytable (A, B, C, D, E) SELECT A, B, C, D, E FROM (SELECT * FROM temp_PARTS1 UNION SELECT * FROM temp_PARTS2) UNION (SELECT A, B, C, D, E FROM a_lot_of_parts) GROUP BY A,B,C,D,E
This part alone works just like I want it:
(SELECT * FROM temp_PARTS1 UNION SELECT * FROM temp_PARTS2) UNION (SELECT A, B, C, D, E FROM a_lot_of_parts)
I just want it inserted inte stated columns in my table.
I've stared so much at this I'm "homeblind", ie I can't see the forest because of all the trees...
