There are advantages and disadvantages of using either options, however in my opinion the advantages of permanent work tables outweigh the advantages of # tables for ETL but # tables win for TP systems
- permanent work tables are available for debugging
- permanent tables can be conditionally created – whereas temporary tables can’t
- # tables are automatically cleaned up whereas you have to manually either recreate or truncate work tables
- # tables always live in tempdb (although you can also create permanent work tables there as well)
- # tables are unique to the session so never any risk of contention
As an alternative to creating a view over multiple tables, there is also the option of using the queryout parameter in BCP to select the data for extracting (same example as used in the view usage)
bcp “select cust.CustomerId, cust.CustomerForename, cust.CustomerSurname, cust.Telephone, addr.HouseNo, addr.Street, addr.Town, addr.county, Addr.postcode From tblCustomer cust Join tblAddress addr ON Cust.AddressId = addr.AddressId” queryout cust.dat -S(local) -T -c
Obviously it’s a lot longer than using a view – but sometimes you don’t have permissions to create views so it’s a usefull alternative
The easiest way of doing this is to create a view over the two tables and then bcp the content of the view out in your required format.
e.g. if you have customer details in a customer table and an address table and you want to extract the contents of the combined table in a tab delimited file
create view vw_Customer_Address as select cust.CustomerId, cust.CustomerForename, cust.CustomerSurname, cust.Telephone, addr.HouseNo, addr.Street, addr.Town, addr.county, Addr.postcode
From tblCustomer cust
Join tblAddress addr
ON Cust.AddressId = addr.AddressId
Then just bcp this table out
BCP mydb.dbo.vw_Customer_address out cust.dat -S(local) -T -c
Two ways to log jobs – both equally effective
You can either pipe the output if you are running batch (.bat) files or
In the jobs steps – advanced use the output file option.
If you have multiple job steps
First step set the overwrite option
Subsequent steps set the append option
Database dumps / log dumps
Make the database dump overwrite and then for log or differential dumps append so that you have a full job history in a single file
Don’t forget to include the logs directory in the server backup schedule!
One word of warning…Permissions – the account that SQLServerAgent is running under will need permission to create files in the log directory.
All jobs for a server should be placed in the same log directory – and don’t forget to occasionally check the contents of the log files – since that’s how you will find out about problems!
If you have static data tables then consider the following option – for low volume static data.
You can set up a DTS package to export the tables to an excel workbook very easily. Just use the export wizard and schedule the DTS package to run at a convenient time.
ODBC drivers are configured on the server – 2000 installed by default but not SQL server 7
The sql server agent account has permissions to write to the folder for the spreadsheet.
That the option to delete the data first is set.
Don’t forget to set up a DTS import package and….test, retest and test again.
The easiest way of getting a column list into a variable is to assign the contents of syscolumns (or sys.columns for SQL 2005).
Create a variable to hold the column list and then reassign it in the select statement.
To get the column key use the object id of the object
Select @variable = @variable + coalesce(@variable + ‘,’,”) from sys.columns where object_id = object_id(‘tablename’)
We are using the fact that coalesce returns the first non null value here so that we get a ‘,’ for each of the subsequent field names but just the column name for the first field.
Much easier than using a cursor on sys.columns to get each field.
If you don’t use coalesce for the first item in the array, you will also need to initialise the variable before assigning it since null + anything = null
create table temptable(id int identity, scol1 char(30), scol2 char(30)
Declare @myvar varchar(max);
select @myvar = coalesce(@myvar + ‘,’,”) from sys.columns where object_id = object_id(‘temptable’) order by column_id
2000 only getting space used from tables
This isn’t a new tip – you may have seen many references to using sp_MSForEachTable and it’s partner sp_MSForEachDB – just put it into context with getting the data back out in a clean format by using insert .. exec
/* create a table to hold the results of the exec command */
create table #SpaceUsed (name nvarchar(129), rows char(11), reserved varchar(18),data varchar(18), index_Size varchar(18), unused varchar(18))
insert #SpaceUsed exec sp_msforeachTable 'sp_spaceused "?"'
--Run the script - then
select * from #SpaceUsed order by cast(rows as int) desc
--gives you the number of rows
select * from #SpaceUsed order by cast(left(reserved,len(reserved) - 3)as int) desc
–Gives you the data based on space used (including indexes)
Note that in SQL 2008 and subsequent versions – you can simply use the standard reports in the database e.g. right click on the database -> Reports ->Standard Reports->Disk Usage by Tables which can then be exported e.g. in excel by right clicking on the report -> Export -> Excel
I haven’t added that in as a SQL 2008 tip since most people already know about the custom reports – however, worth bearing in mind that compatability must be set to at least 90 for the custom reports to work!
Format files give a lot of the flexibility of DTS Transformations – you can skip columns, reorder the columns and do most of what you want.
They can also be used in conjunction with Bulk Insert to get the speed efficiency from Bulk Insert.
But if possible – order the table and the source data in the same way and use a sensible delimiter which isn’t in the data – if the data is from a PC based system then try to use tab delimiters – if it’s sourced from a mainframe look at ‘%’ ‘#’ or ‘$’ since they tend not to be extensively used in mainframes.
If the source file is generated by COBOL then it’s likely to be fixed length – so use that instead to create your format files – saves you 1 byte per column – if you’ve got a lot of fields e.g. y/n indicators then you can halve the data by using fixed format. FTP’s will be faster
the easiest way of checking for the existence of a # table is to check it’s object id
if exists(select 1 from tempdb.dbo.sysobjects where object_id = object_id(‘tempdb.dbo.#worktable’))
drop table #worktable
Also applies to sql 2005 and 2008
Welcome to the new updated SQL Server Tips site from SVEK Computing Ltd, we will try to give you a variety of interesting and usefull tips for SQL Server – from 2000 onwards.
We hope you like the updated version of this site
We are currently updating the content and all the original content should be available by 21st October 2011