Generating a list of numbers in T-SQL

I wanted to create a list of numbers so I could cross join them to another set of data to generate some test data quickly. There are ways to generate numbers using a CTE (Common Table Expression) or if you’re in a real hurry you can make us of a fairly undocumented table called master..spt_values. Seeing as this table isn’t well documented don’t go using it for production code since it might change but for quick one-off tasks it seems fine.

The following SQL creates a list of numbers from 0 to 2047 :

select [number]
from master..spt_values
where type = 'P'

Importing from Excel into SQL Server

I was going through a mountain of pain trying to get an import working from an Excel spreadsheet working into SQL Server 2005 but I finally managed it and here’s how…

Here’s the SQL I’m using (slightly abridged!) :

set transaction isolation level read committed

declare @sql nvarchar(1024)
set @sql = 'select * into ##Temp from opendatasource(''Microsoft.Jet.OLEDB.4.0'', ''Extended Properties=Excel 8.0;Data Source=' + @ExcelFileToImport + ''')...' + @ExcelWorkSheet
exec sp_executesql @sql

Points to note :

• I’m using dynamic sql because the opendatasource() doesn’t like being passed variables… so be sure to validate the contents of @ExcelFileToImport and @ExcelWorkSheet.
• Because I was forced to use dynamic sql I had to use a global temporary table. I wanted to load the data into a temporary table so that the columns were generated on-the-fly and I could check the correct columns were supplied by examining tempdb.sys.columns. It therefore had to be a global temporary because a standard one would have been thrown away after sp_executesql() had finished! For my situation the global temporary table is ok… but you’ll have to consider your own situation and whether concurrency allows.
• I kept getting the error below on the subsequent re-tries of uploading but worked around this by adding ‘set transaction isolation level read committed’ in the sproc.

  The requested operation could not be performed because OLE DB provider "Microsoft.Jet.OLEDB.4.0" for linked server "(null)" does not support the required transaction interface.

I was having trouble executing within a transaction but with the changes made above the sql could be called with the following :

using (TransactionScope scope = new TransactionScope())
{
    UploadExcelIntoDatabase(importTicket, importFilename);
    scope.Complete();
}

Conditional Aggregate Totals in SQL

I’ve been working on a SQL Server Reporting Services report which had to display several counts on each row. Each row represented a project and each project could have several tasks that were either ordinary tasks or milestones and could also be escalated. The report need to show something like this :

Project Name	Milestone Counts					Task Count					Escalated Task Count
	1st Week	2nd Week	3rd Week	>=4th Week	Total	1st Week	2nd Week	3rd Week	>=4th Week	Total

As you can see a task initially has to break down into two major colours of milestone and ordinary task counts but then is broken down into counts in the 1st, 2nd, 3rd and greater than 4th week totals including the total itself.

Initially this seemed tricky when thinking in terms of using count() but there’s an incredibly simple technique that solves this but which isn’t immediately obvious. The Sum() aggregate function can contain an expression and in this example that expression could compare the task date against late week dates and return either a 1 or 0 :

select ProjectName,
   sum(case when IsMilestone = 1
      and TaskDate < @LateDate
      and Taskdate >@OneWeekLateDate
      then 1 else 0 end) as Week1LateMilestoneCount
from Project
group by ProjectName

I’ve only included one example there but I think it’s enough to get the idea.

Selecting a result set from a stored procedure in SQL Server

Selecting data from a stored procedure in SQL Server is already a documented feature and here’s an example:

insert into #systables exec sp_executesql N'select * from Northwind.sys.tables'

…problem is this example doesn’t run without first creating the temp table and therefore knowing all the column definitions. When I’m running quick queries this isn’t exactly convenient. I’ve seen blogs posts using a linked server but there’s another way :

select * into #systables
from openrowset(
   'sqlncli',
   'server=.;trusted_connection=Yes',
   'sp_executesql N''select * from Northwind.sys.tables'''
)

I wouldn’t necessarily use this as a day-to-day process on a production environment but for administration or scripting installations I think it fits the bill.

String Concatenation in T-SQL and Importing from Excel

I keep forgetting how simple this little trick is so it’s time to blog it.

I wanted to write some T-SQL that would confirm that the columns just created by an import from an Excel spreadsheet into a temporary table contained the expected columns irrespective of order or other columns that exist in the file. I also wanted to be able to inform the user of the problem instead of just fail with a horrible error. I chose a fairly simple method, I would concatenate the columns from the temporary table (#tempImport in this case) and compare that with an expected string. If they don’t match I can report back to the user that the expected columns do not exist (in this implementation I’m not interested reporting which columns are missing.) The implementation I chose was this :

select * into #tempImport
from openrowset('Microsoft.Jet.OLEDB.4.0',
   'Excel 8.0;Database=e:\temp\aSpreadsheet.xls', Sheet1$)

declare @columns varchar(512)
set @columns = ''

select @columns = @columns + ',' + [name]
from tempdb.sys.columns
where [object_id] = object_id('tempdb..#tempImport')
    and [name] in ('Store','Store Name','PSA1','PSA2','Mtrg')
order by [name]

if (@columns <> ',Mtrg,PSA1,PSA2,Store,Store Name')
    print 'The expected columns in the spreadsheet do not exist'

This is a brute-force type of implementation but I like it because it’s very simple and keeps everything in one place. There will be better ways to implement this but for the moment this fits the bill.

Update:

http://support.microsoft.com/kb/326548 is a useful guide to using the ISAM provider.

Generating random numbers in a SQL Update

I wanted to fill a column in a table with random numbers so off the top of my head wrote the following sql :

update code set serialNumber = floor(rand() * 30000000000)

Unfortunately this set the serialNumber to exactly the same value since RAND() isn’t being re-executed for each row. I was in a bit of a hurry so instead found the following on google :

update code set serialNumber = abs(cast(cast(newid() as binary(4)) as int)) + 30000000000

Newid() is executed for each row and in this example it’s being used to created (aka ‘bodge’) an int. An interesting solution and did the trick for me.