SQL Server 2005 Book Reviews

Now that my SQL Server 2005 presentation is over, I have time to write reviews of the books I used to gather my material! My two primary resources were:

• A First Look At SQL Server 2005 for Developers from Addison Wesley
• Introducing Microsoft SQL Server 2005 for Developers from MS Press

As you have probably noticed, both books have a developer focus. The reason for this is twofold. First, the presentation was for the Fort Wayne .NET user group, a group consisting primarily of developers. The other reason is that they were all I could find on the subject!

Both of these books were fabulous, particularly considering the unpredictable nature of beta software. The books provide excellent preliminary documentation for the new features of SQL Server 2005 including the .NET Runtime Host extensions and Broker Services.

In addition to being well written and full of examples, these books are a great compliment of each other. The Addison-Wesley book shines in the area of defining the syntax of the new T-SQL extensions and providing examples. The MS Press book does not lay out the syntax but provides better examples and often includes a diagram to accompany a discussion.

As is expected with any literature based on a beta product, there are a few issues with the books. Foremost, the books make reference of the System.Data.SqlServer namespace. As of December of 2004, this namespace has been removed and its functionality has been moved back into System.Data.SqlClient. Additionally, The MS Press book makes use of a ":" syntax for creating an object based off of a member of an assembly. For consistency, the ":" syntax has been replaced with a "." syntax.

If I had to suggest one of these books, my vote would immediately go to the MS Press book. The Addison-Wesley book tends to be wordy and drawn out. The MS press book is half the length of the Addison-Wesley book and covers the same information almost in the same level of detail. The MS Press book is also an easy read and its use of diagrams makes scanning for information simple. For anyone serious about getting involved with SQL Server 2005 development though, I would recommend reading these books together to get a better view of the overall picture.

<SUBJECT>At a Loss</SUBJECT>

As I indirectly mentioned in my previous post, I will be giving a presentation on SQL Server 2005 at the March .NUFW meeting. I have been spending a great deal of time preparing for this presentation and have most recently been studying the new XML capabilities including extensions to the FOR XML clause, the inclusion of the XML data type, and the native support for XQuery.

All of these enhancements seem exciting but I am at a loss as to why much of this is being natively supported in an RDBMS, especially one that allows for embedded CLR assemblies. Don't get me wrong, I think that XML is a great technology and have used it many times but like everything else, it is a tool and needs to be used appropriately.

I fully understand wanting to manipulate XML as close to the database as possible to reduce the amount of network traffic typically required to build a dynamic XML document. As I mentioned above, SQL Server 2005 supports embedded CLR assemblies. One of the new features includes in version 2.0 of the .NET Framework is the SqlServer namespace that provides classes for querying and manipulating data against the context connection. This provides an easy and familiar mechanism for building and parsing XML documents inside of the database because the classes defined in the System.Xml namespace are accessible!

I even understand the need for an XML data type so that XML documents may be passed into or returned by stored procedures or user defined functions for parsing or building, respectively. What I do not understand is storing XML in a relational database!

All of the arguments I have read or heard for storing XML in a relational database fall into a couple of categories:

• XML is universal because it uses UNICODE.
• XML allows for native mapping of properties to objects, thus, it provides a simple mechanism for persisting and recovering the data related to an object.

These arguments have not convinced me that storing XML in a relational database is a good practice. The UNICODE argument falls apart immediately in that if UNICODE is needed, values can [typically] be stored natively in the database through the use of an NCHAR or NVARCHAR data type (SQL Server and Oracle support this). If natively storing the value as UNICODE is not an option, it can be converted later.

As for the second argument, I will agree that XML does allow for an easy mechanism for persisting and recovering object related data. This is also known as laziness! I would even argue that this approach is lazy even when the data is "guaranteed" (is anything ever certain?) not to change.

I have never seen an XML document that wasn't able to be easily represented in a relational format. It is true that many documents require many tables but the point is that it can still be done. In many aspects, it is also more efficient to store data values individually spread across multiple tables than it is to store a single XML document (or fragment) in a single column.

Parsing the data values into individual columns across an appropriate number of tables provides many advantages to storing an entire XML document in a single column.

• When storing raw XML, a great deal of space is wasted just to store tags and attribute names. Depending on the structure of the XML, this can quickly add up since it will be extended across all of the rows in the table storing the XML.
• What if a single value needs to be changed? How about adding a new Item? If the XML document is large, this can be an expensive operation since the entire column value will need to be replaced. (It should be noted here that SQL Server 2005 does support a proprietary XML Data Manipulation Language (DML) but the XQuery specification does not currently support this functionality so it is likely to change in future versions.)
• What if a different view of the data is needed? Suppose the XML represents an article and only the title is needed? The entire document would need to be read and parsed to extract a single value. The title itself could be broken out into its own column but then other problems may exist:
  1. Multiple copies of the same data if the title is left in the XML. This is wasting space and adding complexity because the values need to be kept in sync.
  2. The title could be removed from the XML and stored in its own column. This is eliminating the purpose of storing the XML in the database in the first place because the title will either need to be added back into the XML for delivery to the client or two queries will need to be run!
• Another gripe about storing raw XML in a database is that it is difficult to query against. An example is if the XML column stores documents that describe books and a user wants to find all of the books written by "Bob", either every column needs to be returned and searched using XPath or XQuery (SQL Server 2005 provides native support for XQuery but the syntax is an awkward combination of T-SQL and FLWOR)
• The XML specification does not require values to be strongly typed. Strongly typed data ensures consistency. Storing the values in individual columns that are appropriately typed will add an extra layer of protection to the data.
• Finally, a "typical" XML document does not represent a scalar value but rather, a complex type.

It is for these reasons that I do not support storing XML in a relational database even for document-centric type applications. As I mentioned earlier, I have no issue with XML itself. I believe that it is a great tool and deserves a place in every developer's toolkit. Like all tools though, it needs to be used appropriately in order to be effective. If my logic is flawed or I have missed something, please feel free to try to change my mind.

.NUFW Plug

The .NET User Group of Fort Wayne (.NUFW) meets on the second Tuesday of every month. Meetings are free and open to the public so if you are in the area and wish to attend, please stop by.

Upcoming Events

	2/8/2005	3/8/2005
Topic	Message Oriented Architecture	SQL Server 2005 (Yukon) Overview
Speaker	Mike Amundsen [INETA]	Dave Fancher

Please refer to the .NUFW Web site for meeting time and location.

GMail

I have four GMail invites if anyone is interested. Offer some constructive comments to one of my technical posts and an invite is yours! [First come, first serve but the invites will only be given if a relevant, thought provoking comment is posted]

A Matter of Great Importance

It has been a while since my last post. Work and the holidays have kept me pretty busy. During the past few weeks, I have had some time do so some of the reading that I have been meaning to do. Based on the content of the various books and journals I have been reading, it seems as though there is a great, ongoing debate as to which part of an application is more important: the user interface (UI) or the back-end code (for the purposes of this entry, I am including any data storage mechanisms and logic code including databases and stored procedures).

To me, this debate seems pointless. The answer to the question of "which is more important?" is really quite simple...both parts are equally important!

The equality of importance is apparent when one examines the true purpose of a program. The true purpose of a program is to make some task easier and more efficient for someone. It is extremely important to never lose sight of this purpose since doing so will negatively impact the quality of the end product.

Both parts of a program directly impact the overall user experience. Obviously, the UI is what the end user sees and interacts with. If the UI is cumbersome and/or difficult to learn, the task has not been simplified and may in fact have actually become more difficult. User interfaces must be intuitive.

On the other side of the coin, if the back-end performs poorly (including stability issues), the user becomes upset because although the task may have been simplified, it may not be more efficient. Even the perception of poorly written code will ultimately result in lower user satisfaction. Therefore, it is important to minimize the level of poorly written code in order to minimize areas of poor performance and the perception of poor performance.

The most elegant user interface ever designed cannot save an application with a poorly written back-end. The most efficient logic code ever written will not be used or appreciated if the user cannot figure out how to exploit it because of a cumbersome UI.

It is for these [brief] reasons that I believe that equal time should be spent on the design and implementation of both the user interface and back-end portions of any program. Only when all components have been well designed and developed will user satisfaction reach its maximum potential.