# Thursday, 22 November 2007

About half a year ago I attended a webcast about ICE, today I can update that post with a link to the white paper. Maybe more on it later... when I found the time to actually read it.

Technorati tags: ,
Thursday, 22 November 2007 23:46:47 (W. Europe Standard Time, UTC+01:00)
# Tuesday, 13 November 2007

Bah, for the forth time in a week I encounter a star in the select-list of a view. So if no-one told you yet; when creating views, don’t do a SELECT * FROM source, it can (and in time will) hurt you badly. Let’s show by creating a table and two views based on that table;

CREATE TABLE tblSource(
    Col1 nchar(10),
    Col2 nchar(10))
GO
CREATE VIEW vwBad
AS
SELECT * FROM tblSource
GO
CREATE VIEW vwGood
AS
SELECT Col1, Col2 FROM tblSource
GO
INSERT tblSource (Col1, Col2) VALUES ('Column 1', 'Column 2')
GO

So far, so good. Selecting from both views shows you the expected results:

SELECT 'vwGood' view_name, * FROM vwGood
GO
SELECT 'vwBad' view_name, * FROM vwBad
GO

Now add a column to the table, and add some data for that column;

ALTER TABLE tblSource
    ADD Col3 tinyint
GO
UPDATE tblSource
SET Col3 = 3
GO

SELECT 'vwGood' view_name, * FROM vwGood
GO
SELECT 'vwBad' view_name, * FROM vwBad
GO

vwGood still behaves as expected, but vwBad doesn’t return ALL columns as you may have expected based on the *. Now let's see what happens if a column is removed from the table;

ALTER TABLE tblSource
    DROP COLUMN Col2
GO

SELECT 'vwGood' view_name, * FROM vwGood
GO
SELECT 'vwBad' view_name, * FROM vwBad
GO

You will notice that vwGood throws an error (rightfully so, it is instructed to select from a non-existing column), but vwBad doesn’t… it pretends Col3 is Col2. Based on the data, we know better. However this might not be so obvious if based on this change the purchase price is shown as sales price due to changes in the data model and your company starts losing money.

I bet now you want to know which views contain * so you can code them properly, try this query if you're on SQL Server 2005:

SELECT v.name view_name
    , s.name schema_name
    , u.name user_name
    , v.create_date
    , v.modify_date
    , sc.text
FROM sys.views v
    INNER JOIN sys.syscomments sc ON v.object_id = sc.id
    INNER JOIN sys.schemas s ON v.schema_id = s.schema_id
    INNER JOIN sys.sysusers u ON s.principal_id = u.uid
WHERE sc.text LIKE '%*%'

And this one for SQL Server 2000:

SELECT so.name view_name
    , u.name user_name
    , so.crdate create_date
    , sc.text
FROM dbo.sysobjects so
    INNER JOIN dbo.syscomments sc ON so.uid = sc.uid
    INNER JOIN dbo.sysusers u ON so.uid = u.uid
WHERE sc.text LIKE '%*%' AND v.type = 'V'

And don't forget to clean up the example:

DROP VIEW vwGood
GO
DROP VIEW vwBad
GO
DROP TABLE tblSource
GO

Technorati tags: ,
Tuesday, 13 November 2007 12:35:23 (W. Europe Standard Time, UTC+01:00)
# Friday, 09 November 2007

As I wrote before, I passed my beta-exam and am well underway. Because of my experience preparing for the exam, the folks from NewLevel asked if I could do a presentation for them on Windows Server 2008... for marketing sake.

I said yes, so if you're interested, available on November 27th (13:30 - 16:00) and can be in Amersfoort:

  • Overview of product features and why you might want to use them.
  • How these features map to the new generation certifications and what this new generation certification could mean to you.
  • Training options to prepare for the Windows Server 2008 certifications and job-roles.
  • Besides all the talking, a demonstration deploying Network Access Protection in Windows Server 2008.

If you're interested, contact NewLevel by phone +31 73 599 0 150 or mail to Rein Floris at NewLevel. The presentation will be in Dutch an a small fee is charged.

Friday, 09 November 2007 20:05:57 (W. Europe Standard Time, UTC+01:00)
# Monday, 05 November 2007

Just visited the prometric-website, I've passed my 71-649. Also did a quick check on my MCP-transcript, but it's not showing there, yet! The preparation paid off.

Technorati tags: , , ,
Monday, 05 November 2007 23:06:34 (W. Europe Standard Time, UTC+01:00)

If an SQL Server 2005 database has multiple snapshots, do the first changes get written to all snapshots or only to the most recent? Good question, let's make the answer:

The following scenario is based on this sql-script. It assumes you have the AdventureWorks and your database engine uses C:\Program Files\Microsoft SQL Server\MSSQL.1\MSSQL\Data. Be careful, the script changes some data in AdventureWorks, so some examples from books-online may behave unexpected after this demo.

The scenario is in four stages;

  1. The first snapshot is created and a query displays the sizes of the snapshot file on disk (BytesOnDisk) and the internal size (BytesWritten).
  2. Person.Contact is updated, the Firstname and LastName for all records are swapped, practically resulting in the whole table being written to the snapshot before the update. Right after this update, the second snapshot is created and the query again shows the information about the (now 2) snapshots. The first snapshot holding the Person.Contact table before the update, the second being still empty.
  3. Person.Address is updated, AddressLine1 and AddressLine2 are swapped (and NULLs for AddressLine2 are set to empty strings). The query following this update shows both snapshots have grown, proving all snapshots get written (and we have the answer!!!).
  4. To show a snapshot only processes the changes once since it was created, the FirstName and LastName columns in Person.Contact are swapped again. This time, only the second snapshot gets written to (you may notice one or two data pages (8192 bytes) being added to the first snapshot, an insignificant change compared to previous changes).
Monday, 05 November 2007 22:36:03 (W. Europe Standard Time, UTC+01:00)