J2i.Net

Download Code(312 Kb)

Introduction

I had the need for a dynamically growing sparsely populated array. "Sparse" implies that there will be a lot of elements that contain empty values between the ones that contain non-empty values. The .Net collections namespace doesn't contain anything that meets this need. The ArrayList class dynamically grows, but it would have elements allocated for the empty and non-empty values alike. I plan to use this code on devices that have limited memory so this wouldn't do. I ended up making my own class to satisfy this need. While my initial need for this code is for byte arrays I made the code generic so that it can be used with other data types.

How Big is this Class?

Since I will be talking about memory allocation I'll need to also talk about how to get a sense of how much memory that an instance of something costs. This won't account for 100% of the memory that is consumed by by allocation of the object. But it will be close enough for you to start making judgements about if one object cost more memory or less memory than another.

There are a number of predefined value types for which the size is well known. Here are some of the most common ones.

If you build a struct it is a value type too. The size of a struct will be about the sum of the size of its members. Here is an example.

struct Location
{
    public int LocationID;   // 4 bytes
    public double Latitude;  // 4 bytes
    public double Longitude; // 4 bytes
    public double Elevation; // 4 bytes
}

The size of this structure is 16 bytes. Now what happens if you add a reference type (something that is defined as a class instead of a struct)? How big will it be? I'll add a string to the previous example.

struct Location
{
    public int LocationID;      // 4 bytes
    public string LocationName; // ?
    public double Latitude;     // 4 bytes
    public double Longitude;    // 4 bytes
    public double Elevation;    // 4 bytes
}

There are two elements of memory to consider for the reference field. There is the size of the reference and the size of the object to which it is pointing. Think of the LocationName field in the example above as holding a memory address to the area where the string is being held. The size of this memory address will depend on what type of processor architecture that the code is running against. If the code is JIT compiled for a 32-bit system then the refernece will be 32-bits (4 bytes) in size. If it is JIT compiled for a 64-bit system then the refenrece will by 64-bits (8 bytes) in size. When I am working with just a desktop then I do my calculations based on 64-bit systems. But the code on this article will run on Windows Phone so I will be taking both into consideration. There is also other elements within an objects header that is around 8 bytes. The other element of memory to take into the consideration is the size of the string itself. If the string has not yet been assigned and the element is null then the second element of size to consider will have a size of zero. If the string is assigned a value then the second element will be what ever memory is consumed by the string.

It is possible for multiple structs to refer to the same instance of a refernece type. When this occurs you'll want to make sure that you don't count the memory taken up by the instances of the reference type multiple times.

Now let's add an array and see what that does for our memory calculations. The array itself is a reference type. so the amount of memory the reference to it will consume is dependent on the processor architecture.

struct Location
{
    public int LocationID;         // 4 bytes
    public string LocationName;    // 4/8 bytes + string memory
    public double Latitude;        // 4 bytes
    public double Longitude;       // 4 bytes
    public double Elevation;       // 4 bytes
    public int[] SomeRandomArray;  // 4/8 bytes + array memory
}

The size of the memory associated with the array will be the sum of the size of the elements that it contains. If the array contains value types (the above contains a value type of int) then the memory allocated when the array is initialized is sizeof(int) (4 bytes) multiplied by the number of elements in the array. If the array contains reference types then the memory consumed by the array will be the size of the reference (4+8 bytes on a 32-bit system, 8+8 bytes on a 64-bit system) multiplied by the number of elements that the array can hold. This doesn't include the size of the individual instantiated instances of elements it contains.

What happens if I change any of the above from a struct to a class? Once a type is made into a reference type it is going to also have an object header. For a struct when you declare a variable all of the memory that is going to be used by the immediate members is going to be allocated. With a class only the memory for the reference (4 bytes on 32-bit, 8 bytes on 64-bit) is allocated. The memory needed for the members of an reference type is not allocated until there is a call to new. Also note that the memory for the instances of a reference type are allocated on the heap while for a value type they are allocated on the stack.

A Look Sparse and Contiguous Memory Allocation

If we take a look at how memory is allocated for a sparse array and a contiguous (normal) array the reason I needed this will be more clear. As an oversimplified example let's say that that an array of 40 elements must be allocated. Regular arrays will allocate the memory contiguously; the bytes associated with the array will be in the same block of memory. Lets say our sparse array is allocating blocks of memory for five elements at a time. Below the blocks that are in blue represent areas in which there is non-empty data.

At a glance the sparse array is taking up less memory than the conventional array. There are still some empty elements within the structure. This is because it's allocating memory for a range of index. If there's even a single non-empty element within a 5 block range then there's an allocation for all 5 blocks. Whether or not this results in less memory being allocated overall is going to depend on the usage patterns for the sparse array. It will work best when the populated elements are clustered closely to each other.  Can we get rid of the empty elements all together? We could by reducing the size of the chunks allocated. The smaller the chunks the lower the opportunity for empty positions within chunks to exists. If memory was allocated for single elements (a chunk that only holds one element) then we would only have populated elements in the list. But is this better?

It may be. But to get a better answer to this question there's a cost that as of yet has remained invisible for the sparse array. There's a structure for holding  each chunk that looks like the following in my implementation.

public class ArrayChunk<T>
{
    public int StartIndex { get; set;  } //4 bytes needed to contain an integer value

    public int EndIndex
    {
        get
        {
            if (Data == null)
                return -1;
            return StartIndex +  Data.Length-1;
        }
    }
    public T[] Data { get; set;  }
}

In addition to the array that holds the elements of the chunk itself there is also a field to hold the start index for the array. The start index consumes 4 bytes of memory. So there is an overhead of no less than 4 bytes for each chunk allocated.  Consider a conventional and a sparse array both of which are fully populated with 100 bytes of data. Also assume that the sparse array allocates memory for 10 bytes at a time. The memory consumed by the conventional array will be about 100 bytes. The memory consumed by the sparse array will be around 140  bytes. If I reduced the size of the chunks to only having data for 2 bytes of memory then we end up needing at least 6 bytes for each chunk. For the fully populated 100 element collection this would translated to no less than 600 bytes. With results like that one may wonder why even bother with the sparse array.  But consider another scenario. Lets say that only 20 elements of the array are populated with 10 elements at the begining of the array and 10 at the end. For the conventional array the memory consumed will still be at least 100 bytes. For the sparse array it is around 28 bytes. Something that may become apparent is that the best case scenarios for this sparse array will occur when it partially populated and the populated data items are clustered close to each other.

Growing

The conventional array cannot grow. Once allocated its size is fixed. There are other collection classes within .Net that can grow such as the List<T> derived classes or the MemoryStream class. My understanding is that once the memory buffer for any of these classes is consumed it will allocated a new memory buffer of twice the size as the what it had, copy all of it's data to the new buffer, and then discard the old buffer to be reclaimed by garbage collection later. In trying to confirm this I found the source code for the MemoryStream class. The code of interest is below

            //From https://singularity.svn.codeplex.com/svn/base/Libraries/System.IO/MemoryStream.cs

            private bool EnsureCapacity(int value) {
            // Check for overflow
            if (value < 0)
                throw new IOException("IO.IO_StreamTooLong");
            if (value > _capacity) {
                int newCapacity = value;
                if (newCapacity < 256)
                    newCapacity = 256;
                if (newCapacity < _capacity * 2)
                    newCapacity = _capacity * 2;
                Capacity = newCapacity;
                return true;
            }
            return false;
        }

        // Gets and sets the capacity (number of bytes allocated) for this stream.
        // The capacity cannot be set to a value less than the current length
        // of the stream.
        //
        //| <include file='doc\MemoryStream.uex' path='docs/doc[@for="MemoryStream.Capacity"]/*' />
        public virtual int Capacity {
            get {
                if (!_isOpen) __Error.StreamIsClosed();
                return _capacity - _origin;
            }
            set {
                if (!_isOpen) __Error.StreamIsClosed();
                if (value != _capacity) {
                    if (!_expandable) __Error.MemoryStreamNotExpandable();
                    if (value < _length) throw new ArgumentOutOfRangeException("value", "ArgumentOutOfRange_SmallCapacity");
                    if (value > 0) {
                        byte[] newBuffer = new byte[value];
                        if (_length > 0) Buffer.BlockCopy(_buffer, 0, newBuffer, 0, _length);
                        _buffer = newBuffer;
                    }
                    else {
                        _buffer = null;
                    }
                    _capacity = value;
                }
            }
        }

This behaviour is just fine for typical scenarios, but I will be working with what are relatively large buffers (in comparison to the memory available on the devices on which I will be running my code). So I'd prefer to keep the ceiling for the maximum amount of memory allocation that occurs within the programs that I have in mind. It's also worth mentioning that the .Net runtime treats "large" objects differently than it does small ones. For more information take a look at The Dangers of the Large Object Heap. Large objects (around 85,000 bytes or larger) or allocated on a separate heap than small objects (under 85,000 bytes). During garbage collection the .Net garbage collector will try to condense the objects in the smaller heaps to being in contiguous memory. Objects in the LOH (Large Object Heap) are not as easily addressed. From the referenced article:

Large objects pose a special problem for the runtime: they can’t be reliably moved by copying as they would require twice as much memory for garbage collection. Additionally, moving multi-megabyte objects around would cause the garbage collector to take an unreasonably long time to complete.


.NET solves these problems by simply never moving large objects around. After large objects are removed by the garbage collector, they leave behind holes in the large object heap, thereby causing the free space to become fragmented. When there’s no space at the end of the large object heap for an allocation, .NET searches these holes for a space, and expands the heap if none of the holes are large enough. This can become a problem. As a program allocates and releases blocks from the large object heap, the free blocks between the longer-lived blocks can become smaller. Over time, even a program that does not leak memory, and which never requires more than a fixed amount of memory to perform an operation, can fail with an OutOfMemoryException at the point that the largest free block shrinks to a point where it is too small for the program to use.

There are improvements in the LOH in .Net 4.5. Also note that on devices with more constrained memory (such as Windows Phone) there is no LOH. But there are still advantages to avoiding fragmented memory conditions.

Collection of Chunks

While the code I've written is mean to be a type of collection class the code is still dependent on a collection class for holding onto the chunks that it has. It is possible to use one of the List<T> classes for this or a Dictionary<T1,T2> for this. I've decided to go with the List<T>. Now doesn't it look dubious that I talked about the memory usage patterns of the List<T> class and now I'm using it in my underlying implementation! Isn't that just going to cause the same problem that I described above with memory fragmentation? Well, no, at least not as severe. The List<T> contains references to ArrayChunk instances. So these will either be 4 bytes or 8 bytes. Let's assume the worst case which is 8 bytes. To grace the large object boundaries the array list would need to grow to more than 10,000 elements ( (85,000 / 8)=num of items needed to make the ArrayList large. 85,000 is the large object size and 8 is the amount of bytes needed to store a reference). The number of elements needed to make the array list this size is going to depend on how many elements you allow to be stored in each ArrayChunk. When the array does become large enough to occupy the LOH area the scenario is still better than what would happen with a conventional array since the block of memory occupying the LOH is smaller than the block of memory that would have been occupied by a contiguous array of the elements.

For what the sparse array is capable of doing in the code presented with this writeup the LinkedList<T> would have been suitable (actually more suitable). The List<T> that I use is primarily stepping forward and backwards in the list (I wrote the code making the assumption that read and writes to the list will tend to be clustered close to each other). I used the List<T> because it seems to be a better fit for some modifications I plan to do to this code in the future. I won't detail the details of those plans now since they could change. Consequently of those plans do change then I may swap out the List<T> with the LinkedList<T>. When I do make changes I want to ensure that I don't break any of the existing behaviour of the class. The project for this code also contains a few unit test to validate that the behaviour of the ArrayList<T> doesn't vary from expectations.

Testing

There's a small test project included with the code. It will become more important as time goes on because as I make additions to this code I want to ensure that I don't break any of the behaviours that are already present. Right now the test are just checking against the virtual size of the array, ensuring the memory chunks are allocated or deallocated as expected, and that data is preserved.

What is EmptyValue For?

In the sparse array cells that have not been written to are to be treated as though they exists and they contain some default value. That default value is stored in EmptyValue. There are multiple uses for this member. If there is an attempted read from an unallocated cell EmptyValue is returned. When the sparse array is searching for chunks to deallocate it will check the contents of that chunk to see if all of its elements are equal to EmptyValue. When a new SparseArray is instantiated the EmptyValue field is initialized by calling the default constructor/initializer for the type that it hosts. For the numeric types this will end up being a zero value. If this isn't appropriate for the data type that you are using there is a delegate named IsEmptyValue to which you can assign a method that returns true to indicate that a value or instance should be considered empty and false otherwise.

Using the Code

Use of this code is not much unlike how you might use a strongly typed fixed size array. The main difference is the upper bound on the SparseArray<T> grows as needed to accomodate writes to positions within the least. Reading from locations beyond the upper bound will not result in an exception. If some one reads from an indix that is above the size of the array it remains unchanged. But if someone writes to a location above the upper limit then the array will update its reported size and allocate an ArrayChunk if needed.

Class Interface

Usage Example

 //Initialize an instance of the array
SparseArray<int> myArray = new SparseArray<int>();

//Populate the array with a few elements
myArray[15] = 23;
myArray[1024] = 2;

//Print the size of the array. Will be 1025 (remember this is a zero based array)
Console.Out.WriteLine("Virtual size of the array: {0}", myArray.Length);

//Read from a position beyond the limit
Console.Out.WriteLine("Value in position 2048 : {0}", myArray[2048]);

//Print the size of the array. Will be 1025 (remember this is a zero based array)
Console.Out.WriteLine("Virtual size of the array: {0}", myArray.Length);

//Show how many chunks the sparse array has. 
Console.Out.WriteLine("Chunk count: {0}", myArray.ChunkCount);

//Set the only populated element in the second chunk to zero and then condense the array
myArray[1024] = 0;
myArray.Condense();

//Check the chunk size again. It should be reduced. 
Console.Out.WriteLine("Chunk count: {0}", myArray.ChunkCount);

CodeProject

Tags: Algorithms

It was only a few weeks ago that Windows 8 was unveiled at the Build conference. At first glance it looks a lot like Windows Phone with heavy use of the Metro visual language. The only part of the system that hasn't had the Metro touch is the desktop (unlike previous versions of Windows the Desktop is not something that is always running). The Start menu looks like the Windows Phone start screen only it scrolls horizontally instead of vertically. The only mention of Silverlight is that you could still use it in Desktop mode for backwards compatibility, but the Metro [default] instance of IE would run no plug-ins, including Silverlight. The programming model also is not based off of Silverlight or the Desktop .Net runtime. Its based on something new called WinRT (Windows Runtime). 

At first glance this is something that has concerned Silverlight and Windows Phone developers. At first glance some one might come to the conclusion that the skill in which he or she has invested has become second class in Windows 8. Is Silverlight really getting killed off? What's going to happen for the Silverlight based Windows Phone?

I don't know the future any more than the next person, but what I saw at Build isn't something that raised concern. I found it to be rather reassuring. Before I explain why let me grant the elephant in the room, the rumor that Silverlight is going to be dead. I don't believe this rumour. For years people have predicted that certain Microsoft Technologies were dead (DirectX, .Net, and many other technologies that we still use today). But I'll grant it anyway so that we can explore what seems to be a popular concern. 

Let's assume that next year Microsoft announces that it is going to sunset Silverlight and toss out the Windows Phone programming model in favour of the Windows 8 programming model. What does this mean for the skills that you have developed? Are they now useless? You've been developing skills in C#/VB, XAML, asynchronous programming, and some APIs that were specific to Silverlight and Windows Phone. Let's look at how each one of these will contribute to your Windows 8 development. 

Languages: C# and VB

C# and VB are still being used on Windows 8. If you've been using these languages you can continue to use them. Additionally if you know C++ or have algorithms that had been written in C++ you'll be able to port them over to Windows 8. Windows 8 also supports JavaScript as a programming language too. 

XAML

XAML is still used on Windows 8 for building your UI. Many of the elements you've become familiar with are present in addition to some new ones. No huge changes there. 

Asynchronous Programming

One of the challenges for developers that were new to Silverlight was that task that one may have been used to doing synchronously are only available as asynchronous calls. On Windows 8 you'll find that many of the tasks that were asynchronous in Silverlight and Windows Phone are still asynchronous. Additionally other APIs have been made asynchronous, including File IO. 

The Familiar and What This All Means

You'll come across APIs that look similar of not identical to what you've seen in Windows Phone and Silverlight. Windows 8 has the concept of an application getting tombstoned, specifying the permissions it needs, and so on. Windows 8 Metro applications will only be distributed through the Marketplace. Doesn't all of this sound familiar. If you are a Windows Phone developer it should. You've already got a head start on Windows 8 development. This is far from the doom and gloom picture that some stories would have some one believe. 

If you want to dive into Windows 8 programming the 64-bit development images are available for download. I suggest running them on real hardware. I tried them in the emulator VirtualBox and it's just not the same experience there. 

Tags: Windows Phone 7, Windows Phone, Windows, Windows 8

I've uploaded some code I was working on to animate sprites in XNA.

Animating a sprite isn't difficult, but I wanted some way to animate them but reduce the coupling between code and the animation. The Content Pipeline is perfect for this. So I created a component that will handle the animation scenarios that I need along with a content extension so that I could load these animations as content. Right now the animation information is in an XML file. This is a stepping point towards having a graphical tool for handling this.

You can read about the code here or see a brief description of it in the video below

Tags: XNA, Windows Phone, Xbox, Game

Download the code (2.90 KB)

I was working on an educational project using Windows Phone and XNA on the desktop (the phone will be controlling the desktop). The project involves demonstrating the laws of celestial mechanics. So the modeled entities have attributes with large values (ex: the mass of a planet in kilograms, the distance the planet is from its sun in meters, so on). In the Xna portion of the program the Vector3 class appeared to be a good fit for modeling a planet's position, velocity, acceleration, and so on. My only concern with Vector3 is the precision of the numeric type that it uses. It uses a single precision number. When I started my initial test I was seeing how small objects reacted on the earth's surface just to make sure that my gravity calculations were correct. Everything worked fine; the object would fall toward the earth with the expected acceleration. So I went straight from modeling a small object on the earth's surface to modeling the Earth's moon. That's when things broke down. The numbers coming back from the gravity calculations were all wrong. 

It turns out I was running into a precision problem. Floating point types have enough precision to hold the final results of my calculations, but not enough prevision to hold the large intermediate values during the calculations. There are three ways to solve this:

1. Use a Vector class with a higher precision numeric type
2. Copy values from Vector3 to doubles, perform calculations, copy results back to Vector3
3. Perform my calculations using Astronomical Units (AU) instead of meters. 

Tags: XNA, Physics

I've been carrying this equation around forever and a day and thought I would share it. With this equation you can calculate the distance between GPS coordinates. I tend to use SI units, but you should be able to easily adjust it for units of your choosing.

Tags: Windows Phone 7, Development

And now for some off topic comments!

I've been using IE9 Beta since it was released to public beta and I have to say it rocks! I've installed it on most of my machines and have to say that it looks clean and runs fast. The interface has as many buttons as needed and no more. And the things you can do with HTML 5 are awesome!

Tags: Off Topic

As a developer if you are not using some type of code generation tools then you are not achieving your true potential. In addition to plugins being available such as ReSharper or CodeRush Visual Studio has had the ability to automate some code generation builtin. One of the facility provided is code snippts. A snippet is a code template. You can define your own for patterns that you find yourself implementing over and over.

If you've never used snippets and want to get started then start off by downloading this file. It's a template for making your own snippets.  Then watch this video demonstrating how to use that file to create and use your own snippets.

Tags: Visual Studio, Productivity

Now that Visual Studio has gone RTM you may find yourself needing to use both Visual Studio 2008 for maintaining your existing Windows Mobile projects while using Visual Studio 2010 for other things. You may have also noted that Visual Studio 2008 does work with TFS 2010. To resolve this problem there's a forward compatibility update for Visual Studio 2008. The update weighs in at 11 megabytes and will allow Visual Studio 2008 to work with TFS 2010.

You can download it from here.

Tags: Visual Studio

The other day I was asked "What is cloud computing." I looked at it as being a concept that you just know like "time." How many of you can give a complete definition of time without referencing a dictionary? A short version of what I told him is that if some one uses cloud services it means they use computers that are some where else owned by some one else.

But I heard some one give a nice analogy explaining cloud computing. So I am taking it as my own.

I live in a house. Imagine for a moment that I decided to power the house by means of my own. I'd have to have a generator of some type constructed (solar batteries, hydroelectric generator, windmill, nuclear, what ever). That generator would require some type of maintenance. If I needed more power than the generator could produced then I would either need to suffer through having sufficient power or I would need to add more generators. If I add more generators and then find that many times I don't need them then I have the extra expense of maintaining these extra generators. I don't know anything about maintaining such equipment myself. So I'd either have to be trained in doing so in addition to the other tasks I do around the house or would need to pay some one else to take care of this.

Thankfully that's not how things work at my house. My neighbors and I get our electricity from our power company. I have no idea where it is generated or how it is generated. I honestly don't care as long as it makes it's way to me. If I need to use more electricity than usual in a month it's available. If I use less than I can do so and pay less. I don't have to worry about the space constraints for a generator or hiring staff to take care of it. I just pay for the electricity I need and go on about my business.

That's pretty much how Cloud Computing works in a nutshell. Only instead of purchasing electricity you are purchasing computing and/or storage. You may not know where the computer is actually located. On some plans if more computational power is needed the number of machines allocated can automatically increase for as long as needed and then decrease when no longer needed. There's no need to worry about maintaining the machines as the company that sold them will take care of that.

Tags: Cloud, Azure

Now is a perfect opportunity to get your feet wet with Microsoft Cloud Computing. You can try SQL Server Azue and Windows Azure free for one week. You can get a Windows Azure bootcamp pass to learn Azure at your own pace between May 3 and May 10. You can find information on the bootcam pass and the learning material  here.

Tags: Azure

I've been waiting for this for a while but now the Silverlight 4 themes are now available for download. If you've not seen them before they are some pretty cool looking clean themes that you can use in your XAML applications. If you are looking for the Metro look go for the theme called Cosmopolitan.

You can grab them from here.

Tags: Silverlight

Tags: MIX, Expressions, Silverlight

XNA Game Studio will be made public within the next one to two months. Here's a few of the updates

• You'll be able to use the Mouse APIs on Windows Phones. The Mouse API will actually use the Multitouch functionality and return the position of rhte first point.
• There are two XNA Profiles, Reach and HighDef
  • The Reach profile prvide maximum compatibility across the three screens
  • The HighDef profile is for Xbox/PC. While it lacks WP7S Compatibility it lets you take advantage of the device's capabiities
• The Multitouch APIs are now available to Windows targets
• The ZuneHD won't be getting XNA 4.0
• XNA 4.0 isn't designed to be backwards compatibile with XNA 3.1. Compatibility was sacrificed in the interest of progress.
• Windows Phones will have the Same aspect ratio but different screen resolutions
  • To the developer both devices have the same resolution. WP7S contains hardware scaling
  • WP7S also contains hardware functionality for scree rotation
• Custom effects are not available on Windows Phone 7 Series
• Emulator will be available. You don't need to own a Windows 7 Phone.

Tags: Windows Phone, Xna, Xbox, WP7S

Microsoft is making some adjustments to the Visual Studio and MSDN licensing with the release of Visual Studio 2010. For customers that purchase Visual Studio via retail the MSDN Essentials Subscription will be available. MSDN Essentials Subscribers will have access wo Windows 7 Ultimate Edition, Windows Server 2008 R2, SQL Server Datacenter R2, and 20 hours of Windows Azure usage. Presales of Visual Studio will open on 9 March 2010 at the Microsoft store and other select stores.

Found via Somasegar's Weblog

Tags: Visual Studio

There are times when you will want to share the same source code among several projects. A common way to do so is with a shared assembly; you put common functionality in one project and then share the output among several other projects. But at times this solution isn't suitable such as when you have functionality that you plan to share across more than one .Net runtime (ex: Desktop Framework, Compact Framework, and Silverlight Runtime). For these cases you can copy your source code to the projects for all three run times. But then you end up with three branches of code and may need to make sure thay are synced up with each other.

It is possible to use the same source file in different projects by adding a link for the file from another project so that each project is using the same runtime. Since it is the same physical files changes to the file done from one project are visible to all the projects using the same linked file. Adding a linked file is easy. To link to one file's project from another right-click on the project, select Add->Existing Item and navigate to the file. Once you've found the file click on it and then click on the down triangle on the Add button and select "Add as Link."

A potential problem from using this solution is you may have items in a class that you don't want to be visible in another class. You can selectivly hide sections of code using a few preprocessor directives. As a simple example let's say I made a Windows Form application and I have all of the files from it linked to a second project. I have code in a method that is setting the text on a label. But I want the text to be set differently depending on the project in which it is run. The preprocessor directives I will use are #if, #else, and #endif.

#if App1
            txtMyMessage.Text="Hello from App1";
#else
            txtMyMessage.Text = "Hello from App2";
#endif

In the above code only the C# code in the #else, block will be compiled. The code in the #if block will be ignore. For my first project I want to code in the #if block to be used. To accomplish this I need to add a Conditional Compilation Symbol. I right-click on the project and select Properties. Under the Build tab I can add conditional compilation symbols. I've done this for the first project and have added a symbol named App1. So now the first block of code will get compiled and the second block ignored.

While this solution has it's advantages it is not the end-all solution for sharing functionality across projects. If you find yourself using excessive conditional compilation blocks in your code then you may have reached a point at which it is better off having two seperate source files.

CodeProject

Tags: Visual Studio