Tags: , , , , , , , , | Categories: Development Posted by bsstahl on 2/11/2019 7:55 PM | Comments (0)

What is the result of converting a value that is close to, but not at, the maximum value of an Int64 from a double to a long (Int64)?  That is, what would be the result of an expression like:

(long)((double)(Int64.MaxValue – 1))

  1. 9223372036854775806 (263-2, the correct value numerically)
  2. -9223372036854775808 or another obviously incorrect value
  3. OverflowException
  4. Any of the above

Based on the framing of the question it is probably clear that the correct answer is "D". It is possible, depending on the hardware details and current state of your system, for any of the 3 possible outcomes.  Why is this and what can we do to be sure that the results of our floating-point operations are what we expect them to be?

Before we go into the ways we can modify the behavior of our operations, let's take a look at the two data types in question, Int64 and Double.

An Int64 value, also known as a long, is a fairly straightforward storage mechanism that uses 63 bits for the value and 1 bit to represent the sign.  Negative numbers are stored in twos-complement form to make mathematical operations simpler.  The result is that the Int64 type can store, with perfect fidelity, any integral value between -9223372036854775808 and 9223372036854775807.

The Double data type on the other hand is far more complex. It requires storage for continuous values, not just integers. As a result, the Double data type uses 52 bits to store the mantissa (value), 11 bits to store the exponent (order of magnitude) and the remaining bit of the 64-bit structure to store the sign. Both the exponent and mantissa are shifted by a few bits based on some fairly safe assumptions.  This gives us a range of values for the exponent of -1023 to 1024 and a little more than 52 bits of fidelity in the mantissa.

It is this difference in fidelity; 63 bits for Int64 and roughly 52 bits for Doubles, that can cause us problems when converting between the two types.  As long as the integer value can be stored in less than 52 bits (value < 4503599627370495) values can be converted back and forth between Int64 and Double without any data loss. However, as soon as the values cannot be represented completely in 52 bits, data loss is likely to occur.

To store such a value in a Double data type, the exponent is adjusted higher and the best available value for the mantissa is found.  When converted back to Int64, this value will be rounded automatically by the framework into the closest integer value. This resulting value may, or may not, be exactly the same as the original value.  To see an example of this, execute the following code in your favorite C# environment:

Console.WriteLine((long)9223372036854773765.0);

If your system is like mine, you’ll get an answer that is not the same as the original value. On my system, I get the result 9223372036854773760. It is said that this integer does not “round-trip” since it cannot be converted into a Double and then back to an integer.

To make matters worse, the rounding that is required for this conversion can be unsafe under certain conditions. On my machine, if the values get within 512 of Int64.MaxValue, even though they don’t exceed it, attempting the conversion may result in an invalid result, or an OverflowException. Even performing the operation without overflow checking using the unchecked keyword or compiler switch doesn't improve things since, if done unchecked, any overflow in the operation will result in an incorrect value rather than an exception. I prefer the exception in this kind of situation so I generally keep overflow checking on.

The key takeaway for me is that just checking to make certain that a Double value is less than Int64.MaxValue is not enough to guarantee it will convert without error, and certainly does not guarantee the accuracy of any such conversion. Only integer values below 52 bits can be accurately converted into Int64 values. 

It is always best to avoid type conversions if possible, but if you are in a situation where it is necessary to convert from large Double values into Integers, I recommend trying some experiments in your production environment to see what range of values will convert accurately. I also highly recommend including very large integers, approaching or at Int64.MaxValue as test data against any method that accepts Int64 values.  Values that are very large in the negative direction (nearing Int64.MinValue) are also good candidates to be used as test data in these methods.

I’ve attached a number of resources below that I used in my research to produce this article, and to fix the bug I caused doing this kind of conversion.  If you have run into this situation and come up with an interesting way of handling it, or if the results of your conversions are different than mine, please let me know about it on Twitter

Resources


      Tags: , , , | Categories: Development Posted by bsstahl on 6/16/2008 8:17 AM | Comments (0)

      Over the last few years I've heard a number of public statements from developers about the lack of need for multiple implementation inheritance in .NET and other modern development platforms. Their logic often seems to imply that if you need multiple implementation inheritance, you are not designing your applications properly.  While admittedly, there are usually work-arounds (such as interface inheritance) that allow us to simulate this feature, they usually require that portions of our code are duplicated, violating the Agile requirement "Don't Repeat Yourself".

      One commonly seen example of where multiple implementation inheritance would be very valuable is in multi-tiered, domain specific applications, especially in the data-tier where we may wish to have more-than-one implementation to support multiple data-stores.  Think about the typical data-tier scenario. In this scenario we have a set of domain objects, based on an inherited set of entities with common properties and methods that represent a physical object in the problem domain. These objects also have a commonality in that they are implementations of an object-type common to that data store and may have properties and methods relating specifically to the storage of data.  So, an object whose responsibility it is to persist an Employee entity to a SQL Server data store, could inherit from both our domain Employee entity, and our SQL Data Storage object.  If we also had an implementation that stored data in XML format, we might have an object that inherits both from the same Employee entity as well as from the XMLNode object. If multiple implementation inheritance were supported in our framework, we could avoid the common work-around of repeating our entity implementation by using an interface to simulate that inheritance, or by simply repeating our data persistence logic in each object.

      I certainly understand the need to ship a product.  Since I am also well aware of the added complexity that multiple implementation inheritance creates in compilers and frameworks, it is easy for me to imagine why this feature did not make it into either of the first two major revs of Microsoft's Common Language Runtime.  It is my opinion however that, with the third major release of the CLR forthcoming (Rev 3s being where Microsoft traditionally "nails it") they should strongly consider adding support for multiple implementation inheritance.

      Tags: , , , , , | Categories: Event Posted by bsstahl on 6/3/2008 6:59 PM | Comments (0)

      I would once again like to thank the organizers, presenters and sponsors of Desert Code Camp for giving of their time, effort and funding to support such an outstanding community resource.  Desert Code Camp IV, held yesterday, May 31st at the University of Advancing Technology in Tempe, AZ featured well over 40 sessions on topics such as Agile & TDD, ASP.NET, Flash, Silverlight, iPhone SDK, XAML, Apache, Ruby and much more.  I was fortunate enough to attend 4 of these sessions, all of which were well worth my time in attending.

      The first session I attended was "Scrum 101" presented by Dan Weinmann (who I think works for Desert Code Camp sponsor Neudesic but I am not sure because it is not listed in Dan's extremely minimalist bio on the Code Camp website).  Dan spent the first part of the session explaining general Agile concepts which is appropriate for a "101" class, and his explanation was quite solid.  What I was looking for however came in the remainder of the discussion where Dan gave specific examples of how his organization has utilized Scrum as an effective project management methodology.  I found especially interesting the parts where Dan explained how they use Scrum "under the covers" when working with a client who, for whatever reason, will not use Scrum.  I found this similar in a number of ways to how my team currently functions and came-away with several ideas of how we might be able to improve on our current processes.

      The next session I attended was "Silverlight Zero to Hero" given by Simon Allardice of Interface Technical Training and gets my vote for the mythical "session of the day" award. This session gets my vote not for the abundant humor, which had the room in stitches and led me to refer to Simon on Twitter as "..the Eddie Izzard of the tech world, without the dress...".  No, this was the best session I attended because of the unique perspective he gave to the topic.  That is, he didn't waste our time by giving us the same overview of Silverlight that we could get in any 10k foot video from the Mix conference.  Instead, we were taken step-by-step through Simon's unique metaphors detailing how we can use the generalized feature-set of Silverlight as well as how it could be used to create Rich Internet Applications that are truly effective in communicating  with the user.  The next time I am looking to take a class, I hope Simon is teaching it. In addition, if anyone is looking to become an instructor, I hope they sit-in on some of Simon's sessions which couldn't help but improve anyone's teaching technique.

      My third session was "integrating Data with Silverlight 2.0 Applications" presented by Simon's colleague at Interface, Dan Wahlin. Dan, who described himself as "...not funny like Simon...", certainly had his moments, especially when he (unintentionally?) made a comment about some of his demo data to the effect of "...there are Johns in the room...nothing wrong with Johns."  Dan's wife was videotaping at the time, I hope that clip ends up on YouTube.    Regardless of the humor factor, this presentation as well was full of useful specifics on binding Silverlight 2.0 apps to data services including both SOAP and RESTful services.

      Finally, I attended a preview by Scott Cate of MyKB of his TechEd presentation next month called "C# 3.5 Compiler Tricks".  This session provided me with some fascinating insights into the workings of the C# compiler, including several situations where the compiler uses "syntactical sugar" to provide constructs that compile to .NET 2.0 IL code and have no dependencies on .NET 3.0 or 3.5 libraries.  In these cases, it is possible to use these constructs in Visual Studio 2008 (or more specifically, when using the C# 3.5 compiler) even when targeting the .NET 2.0 framework.  I was also fortunate enough to be able to spend some very enjoyable time with Scott after the session, discussing his most recent project, EasyDB.com.  Scott set me up with access to the service beta and I spent the rest of the afternoon working with this fantastic "SQL in the Cloud".  I will be blogging about this application and my experiences with it much more in the near future.

      Again thanks to everyone who helped to make this event happen.  Sponsors that I haven't mentioned yet who also deserve props for their support include Infusionsoft and JumpBox.

      Tags: , , , | Categories: Development Posted by bsstahl on 7/1/2007 9:51 AM | Comments (0)

      I recently stumbled across an interesting item in a back-issue of MSDN Magazine. The article, "Improving Application Startup Time" by Claudio Caldato, appeared in the CLR Inside Out segment in February 2006. While discussing strong-named assemblies, Claudio recommended adding them to the GAC for performance.

      If an assembly is not installed in the Global Assembly Cache (GAC), you will pay the cost of hash verification of strong-named assemblies along with native code generation (NGEN) image validation if a native image for that assembly is available in the machine. In other words, if an assembly is strong named, the CLR will ensure the integrity of the assembly binary by verifying that the cryptographic hash of the assembly matches the one in the assembly manifest. But if the assembly is in the GAC, this verification can be skipped because the verification is performed as part of installation into the GAC and any update requires administrative permissions. So the CLR is basically assured that changes have not occurred.

      The hash verification process is expensive because it involves touching every page in the assembly, which can be bad for cold startup. Also, the hash computation is CPU-intensive and thus impacts warm startup, too. The extent of the impact depends on the size of the assembly being verified.

      If an assembly has been precompiled using NGEN but it is not installed in the GAC, then during binding, fusion needs to verify that the native image and the MSIL assembly are the same version (to avoid cases where a newer version of the assembly is deployed on the machine but a newer version of the native image is not generated). In order to accomplish that, the CLR needs to access pages in the MSIL assembly, which can hurt cold startup time.

      I found this particularly interesting because I generally do not recommend putting assemblies into the GAC unless there is a particular need. The GAC is a very useful and powerful tool, but it does add complexity to the deployment of applications, occasionally limiting the frequency with which applications can be deployed, and often increasing the testing requirements for deployment of applications that use shared assemblies. As a result, I usually avoid putting assemblies in the GAC unless they truly need to be there (such as shared .dlls in applications that require that they be using the same version of the assembly). I have also heard of people pulling assemblies that were installed in the GAC, back out into bin-folder type deployments in order to simplify the deployment process.

      The information from this article adds a wrinkle to the process of removing assemblies from the GAC because it makes the best-practice for doing so include the removal of the strong-name (which was required for inclusion in the GAC). As a result, there may be a performance penalty incurred at each application startup for these apps if the strong-name is left in place. Since removal of the strong-name will not always be possible, this is certainly something to consider. While I doubt that this could cause enough of a performance decrease by itself to make it worth keeping assemblies in the GAC that would otherwise be removed, it is a fact worth knowing, and more importantly, worth testing when considering such a move.

      Tags: , , | Categories: Development Posted by bsstahl on 4/28/2006 9:47 PM | Comments (0)
      I am seeing some things in .NET 2.0 that concern me. Much of it has to do with Microsoft putting in features that have obviously been demanded by many developers, but were not included in earlier versions of the framework because, for the most part, they are the wrong thing to do the majority of the time. For example, Microsoft has included the ability to have inline code as well as the standard code-behind model in ASP.NET 2.0 pages. While this seems like a nice feature, I can't come up with a good reason to ever mix my object code and HTML code. Perhaps someone else can. If you do, please let me know.
      Tags: , , , , , | Categories: Event Posted by bsstahl on 10/30/2003 3:24 PM | Comments (0)
      The morning sessions of Day 2 were highlighted by drill-downs into Yukon and WinFS. The most impressive demo of the conference so far was done during the WinFS drill-down by Gord Mangione and Tom Rizzo. They used the Information Agents of WinFS to configure their voicemail application so that when a call came in from a client matching specified custom criteria, and the calendar showed that the user was busy, it would respond to the caller with the time the user's calendar next showed him free.

      WinFS may finally make good on the decade-old promise of turning the file-system into a relational database. Its metadata features, including extensible schema, appear poised to make the file-system as programmatically accessible as a database server, with many of the same query capabilities including natural language or SQL style queries.

      Yukon also appears to be a major improvement in development technology. This next generation of SQL Server provides CLR (Common Language Runtime - AKA, the .NET Framework) in-process to the SQL Server. This will allow developers to separate the application (or system) tiers physically as well as logically, improving performance, scalability, security, maintainability and extensibility. It will also allow queries to be written in any CLR language, provides structured exception handling for those queries (including in T-SQL) and will allow us to build queries that easily integrate data from various sources (including Web Services).

      Needless to say, I am rather excited about many of these developments and am looking forward to installing Longhorn and Yukon on development servers when I return to the real world.