Tamir Khason – Just code

INotifyPropertyChanged auto wiring or how to get rid of redundant code

For the last week most of WPF disciples are discussing how to get rid of hardcoded property name string inside INotifyPropertyChanged implementation and how to keep using automatic properties implementation but keep WPF binding working. The thread was started by Karl Shifflett, who proposed interesting method of using StackFrame for this task. During this thread other methods were proposed including code snippets, R#, Observer Pattern, Cinch framework, Static Reflection, Weak References and others. I also proposed the method we’re using for our classes and promised to blog about it. So the topic today is how to use PostSharp to wire automatic implementation of INotifyPropertyChanged interface based on automatic setters only.

So, I want my code to looks like this:

public class AutoWiredSource {
   public double MyProperty { get; set; }
   public double MyOtherProperty { get; set; }
}

while be fully noticeable about any change in any property and makes me able to bind to those properties.

<StackPanel DataContext="{Binding Source={StaticResource source}}">
    <Slider Value="{Binding Path=MyProperty}" />
    <Slider Value="{Binding Path=MyProperty}" />
</StackPanel>

How to achieve it? How to make compiler to replace my code with following?:

private double _MyProperty;
public double MyProperty {
   get { return _MyProperty; }
   set {
      if (value != _MyProperty) {
         _MyProperty = value; OnPropertyChanged("MyProperty");
      }
   }
}
public event PropertyChangedEventHandler PropertyChanged;
internal void OnPropertyChanged(string propertyName) {
   if (string.IsNullOrEmpty(propertyName)) throw new ArgumentNullException("propertyName");

   var handler = PropertyChanged as PropertyChangedEventHandler;
   if (handler != null) handler(this, new PropertyChangedEventArgs(propertyName));
}

Simple: to use aspect oriented programming to inject set of instructions into pre-compiled source.

First of all we have to build some attribute will be used for marking classes requires change tracking. This attribute should be combined (compound) aspect to include all aspects used for change tracking. All we’re doing here is to get all set methods to add composition aspect to

[Serializable, DebuggerNonUserCode, AttributeUsage(AttributeTargets.Assembly | AttributeTargets.Class, AllowMultiple = false, Inherited = false),
MulticastAttributeUsage(MulticastTargets.Class, AllowMultiple = false, Inheritance = MulticastInheritance.None, AllowExternalAssemblies = true)]
public sealed class NotifyPropertyChangedAttribute : CompoundAspect {
   public int AspectPriority { get; set; }

   public override void ProvideAspects(object element, LaosReflectionAspectCollection collection) {
      Type targetType = (Type)element;
      collection.AddAspect(targetType, new PropertyChangedAspect { AspectPriority = AspectPriority });
      foreach (var info in targetType.GetProperties(BindingFlags.Public | BindingFlags.Instance).Where(pi => pi.GetSetMethod() != null)) {
         collection.AddAspect(info.GetSetMethod(), new NotifyPropertyChangedAspect(info.Name) { AspectPriority = AspectPriority });
      }
   }
}

Next aspect is change tracking composition aspect. Which is used for marking only

[Serializable]
internal sealed class PropertyChangedAspect : CompositionAspect {
   public override object CreateImplementationObject(InstanceBoundLaosEventArgs eventArgs) {
      return new PropertyChangedImpl(eventArgs.Instance);
   }

   public override Type GetPublicInterface(Type containerType) {
      return typeof(INotifyPropertyChanged);
   }

   public override CompositionAspectOptions GetOptions() {
      return CompositionAspectOptions.GenerateImplementationAccessor;
   }
}

And the next which is most interesting one, we will put onto method boundary for tracking. There are some highlights here. First we do not want to fire PropertyChanged event if the actual value did not changed, thus we’ll handle the method on it entry and on it exit for check.

[Serializable]
internal sealed class NotifyPropertyChangedAspect : OnMethodBoundaryAspect {
   private readonly string _propertyName;

   public NotifyPropertyChangedAspect(string propertyName) {
      if (string.IsNullOrEmpty(propertyName)) throw new ArgumentNullException("propertyName");
      _propertyName = propertyName;
   }

   public override void OnEntry(MethodExecutionEventArgs eventArgs) {
      var targetType = eventArgs.Instance.GetType();
      var setSetMethod = targetType.GetProperty(_propertyName);
      if (setSetMethod == null) throw new AccessViolationException();
      var oldValue = setSetMethod.GetValue(eventArgs.Instance,null);
      var newValue = eventArgs.GetReadOnlyArgumentArray()[0];
      if (oldValue == newValue) eventArgs.FlowBehavior = FlowBehavior.Return;
   }

   public override void OnSuccess(MethodExecutionEventArgs eventArgs) {
      var instance = eventArgs.Instance as IComposed<INotifyPropertyChanged>;
      var imp = instance.GetImplementation(eventArgs.InstanceCredentials) as PropertyChangedImpl;
      imp.OnPropertyChanged(_propertyName);
   }
}

We almost done, all we have to do is to create class which implements INotifyPropertyChanged with internal method to useful call

[Serializable]
internal sealed class PropertyChangedImpl : INotifyPropertyChanged {
   private readonly object _instance;

   public PropertyChangedImpl(object instance) {
      if (instance == null) throw new ArgumentNullException("instance");
      _instance = instance;
   }

   public event PropertyChangedEventHandler PropertyChanged;

   internal void OnPropertyChanged(string propertyName) {
      if (string.IsNullOrEmpty(propertyName)) throw new ArgumentNullException("propertyName");

      var handler = PropertyChanged as PropertyChangedEventHandler;
      if (handler != null) handler(_instance, new PropertyChangedEventArgs(propertyName));
   }
}

We done. The last thing is to reference to PostSharp Laos and Public assemblies and mark compiler to use Postsharp targets (inside your project file (*.csproj)

<Import Project="$(MSBuildToolsPath)\Microsoft.CSharp.targets" />
<PropertyGroup>
  <DontImportPostSharp>True</DontImportPostSharp>
</PropertyGroup>
<Import Project="PostSharp\PostSharp-1.5.targets" />

Now we done. We can use clear syntax like following to make all our properties has public setter to be traceable. The only disadvantage is that you’ll have to drag two Post Sharp files with your project. But after all it much more convenience than manual notify change tracking all over your project.

[NotifyPropertyChanged]
public class AutoWiredSource {
   public double MyProperty { get; set; }
}

Have a nice day and be good people. Also try to thing what other extremely useful things can be done with PostSharp (or any other aspect oriented engine)

Source code for this article (1,225 KB)>>

August 9th, 2009 · Comments (6)

“Zone of Pain vs. Zone of Uselessness” or code analysis with NDepend

As for me, there are only two kinds of projects – hobbyist’s nifty tools and systems (scale may wary). The main difference between those is the easiness of making changes and refactoring. Other words, how many other developers I should persuade to do it just because “In the final analysis, it’s their war” – JFK. But what can be the good reason for such fast talk? – “You code sucks or at least it ought to”.

“Every generation needs a new revolution” – Thomas Jefferson

So, in order to win such revolution for “systems” you absolutely need static analysis tools like NDepend. Those tools are not intended for being your advocates, but those intended to help you to understand all risks and approximate the amount of work should be done to fulfill another revolution.

Unfortunately, you cannot use such tools for fair measuring of code quality because of Computer Science rules of thumb. How to decide whether “methods is too big” or “method is too complex”? However you can (and should) use it for dependencies risk detection. For example, in following illustration you can clear understand that any change inside BetterPlace.Core or BetterPlace.Model assemblies (and namespaces) can be painful.

Now the only question is who is responsible of modules, using it and how long will it take to convince them to make a revolution.

From here you can start using CQL (Code Query Language) which is SQL like language invented by Patrick, the author of NDepend, for querying code elements and sections. By using it, you can define what “method is too big” means in terms of your project.

SELECT METHODS WHERE NbLinesOfCode > 300 ORDER BY NbLinesOfCode DESC

Or see where you should replace method overloads by default arguments, introduced in .NET 4.0

SELECT METHODS WHERE NbOverloads > 1 ORDER BY NbOverloads DESC

Then, when you marked all places and human targets for revolution, you can start it. After you done, you can even compare builds and measure the amount of work and quality of results achieved.

To finalize, I just touched the tip of what good static analysis tool can be used for. So get it, learn it and use it not only when you need to make a revolutions, but also during your application design and build process to be aware about how the new monster created will looks like.

Download NDepend >>

Proper disclosure: Apr 15, Patrick, the author of NDepend, asked me to review his tool and offer one license for evaluation. I told him, that do not need to evaluate it because I’m using it for a while (also I had the license of my own) and I’ll be happy to write a review once I’ll have a bit time for it. Now it happened. Thank you, Patrick, for such good work!

July 27th, 2009 · Comments (0)

How to pass technical interview in Better Place

If you following me on Twitter, you probably know that we’re looking for developers. However those who was in interview here become a little bit shocked. It is not very standard interview for software houses. So how does it works?

Step zero – requirements and open positions

Even before somebody send us his/her resume there are requirements and open positions. Those requirements are very straight forward: “We need a developer for servers” or “we need somebody to build UI” or “we need GIS guy” etc. All open positions (at least for Israel) you can always find in our local website. Those days we have two open positions for our team: super-fantastic developer and chronic headache reliever.

Now a bit about candidates. I had some developers hired during my career and it always worked following way: there are only 10 kinds of developers – developers and not. It never worked for me to hire somebody just because I have to hire somebody. Also it did not worked to hire mythological “average” or “rather good” developer. So for me the segmentation here is very binary- can write code or can not.

Step one – CV

Somebody told me that the only chance to hire good developer via CV is if the developer is student. However, and in spite of this the very first station is HR department, they get a ton of CVs each day and filter it out by using strange and mysterious heuristics. In most of cases it works like in bad search engine – by using keywords. So my very basic requirement from our HR team, please do not filter it this way because I’m working like SEO filter with those CVs by lowering rating of “multi-threaded Design Patterns”, “.NET 1,1.1,1.2,1.3,2.1.2.2.3.4.3.4.5.6.5.4.3 and dot.Net Remounting for Agile Ajax(tm) Web Services”, “deep knowledge and application of MFC, GUI, WinForms, C++, C#, HASP, SQL, ADO, OOP, ODP, IRC, TCP, HDTV, AX, DX, COM, VC++, API” and “Windows 3.11, 95, ME, 98, 98 SE, 2000, NT 3, NT 4.2, 2001, XP, XP SP1, SP2, SP3, 2008, 2010, Vista SP3 and DOS”.

Also candidate which “management responsibility for the 26 software engineers (6 teams) located at India (Bangalore). Resources allocation for $1M budget and restructuring planning based on defined priorities” probably won’t be developer, but want to be finance controller, because we was able to make 26 developers (even in India) to work for $38K a year including materials, living, facilities, etc.

So a small tip for your CV, please write what you did (yes, it is not shame if you never heard about OOP/ODP, but knows how and when it should be virtual and when abstract).This way you’ll not waste your and your future employer time.

Step two – Phone call

If CV is readable and clear for me, I’m calling the person. Usually it takes under 5 minutes to invite him/her or not for frontal interview. However sometimes it can take even half an hour. I am not asking the same questions, this because it is impossible to ask the same somebody who write smart client applications and those from web planet. I almost never ask a candidate to repeat what he/she just told. The only exception is when I am not understand the answer. However when the candidate asks relevant questions it is big plus for him.

The general set of questions contains of three main sections: what exactly you did (even if it clear from CV), what are you looking for and whether you want to work in Better Place.

The phone call ends by taking one of four direction: invite for frontal interview (jump to step 3.1), invite for technical exam (jump to step 3.2), send home work to prepare (non-blocking wait for a year or two) or “we’ll be in touch” by transferring the candidate with mark “negative” to HR (end of the process). They know how to tell them.

Step three point one – Frontal interview

Frontal interview is usually takes between an hour and an hour and a half. It contains of general CS questions like those. Also question to understand whether the candidate is developer or “code-monkey” – one who does not know what the difference between scalar and vector or how to convert float to int faster then using “corporate frameworks developed by other team”. The thumb rule is simple – if each question takes more then 10 seconds to answer the candidate does not really knows the topic. In spite of this, I allow about a minute for each answer.

Next section is related to specific domain of the candidate. It might be architectural overview of one of his/her solutions, all he/she knows about inheritance (why, how, multiple, virtual, public, etc). Constructors/destructors/templates/generics/allocations/virtual inheritance is a next part. What is “new” and why we need it. How to cleanup remindings and how it works in real life (even if you thought that it should be done automagically). Special bonus for those who knows what interrupts and registers are and how it really works under the hood.

If the candidate good enough we keep going toward data structures. This includes linear/random access, hash, trees (with different colors), search vs seek vs lookup with yield return bonus. What is special about “string”. Lazy invocation vs evaluation, alloc/malloc (if relevant), lifetime.

If the candidate is especially good we can start speaking the target language (sometimes it’s assembler and sometimes is C# or even Python [I can speak some ]).

Step three point two – Technical exam

Sometimes, when it is not very clear whether the candidate is programmer, I’m asking for write a small program from scratch without using standard libraries. For less developer position I’m asking to find a number of bugs in real code (which includes required bugs ). Code is different for different levels of programmers. Also in most cases a candidate can use any computer language he/she wants.

To prepare such exam we need to work very hard. First of all (and before it comes to candidates) we’re sending it to all developers in similar positions and asking them to do it. Then we are measuring time and multiply it by 2. Also we’re writing down all question, asked by developers and trying to fix description to answer all commons.

Candidate got new clean machine, empty silent room with air conditioner and can use any information sources he/she wants in order to do the exam. It can be internet, friends, books, etc. However it is not so simple as it looks. Each candidate has final time to write the program. It takes longer to find the “similar” solution and adopt, rather then do it yourself.

We are require three hours to finish current exam (it took about an hour to finish it for me, and hour and a half in average for the company). If the candidate finished it before the time he/she got a bonus point. If he/she wrote “almost the same code” – failed, “almost finished” or “almost works” – failed, “has not enough time” – failed, two significant bugs – failed, “the question was not clear” – failed, “did not know who to ask” – failed, etc.

It’s amazing how hard for some code-monkeys not to use standard libraries. Sometimes it is too hard for bad programmers to understand that not all computer languages has arrays and collections, sometimes, it’s too hard to realize that if you not really understand what inheritance is you have to write the same procedure more then once.

But not only the exam determinates the future of the candidate. Sometimes (in very rare cases) we’re wrong with this test and do not filter out bad programmer. It is very hard to lay out him later, but because of such person you will not finish project at time, or will do it overbudget. Thus we prefer not to hire, then to fire. This why we have additional step.

Step three point three – Another interview

It is very important to emphasize that we have about 0.05% of those who where at stage three point one. This interview can take upto full day. You should come ready for this. Plan and watch each of your steps. Also you should be ready to speak with number of people near flip charts, whiteboards, calculators, computers and other hardware things. Also you should be ready to assert your opinion.

Most of people you’ll speak with will be very smart (this is not restricted to your direct manager or future colleague only). Sometimes you’ll speak with two or three people at simultaneously. Each person will have his own very special opinion. Also the questions may wary from base calculations to mechanical and electronic schematics or references of your vehicle.

You can take timeout, however take into account that people speaking with your sacrificing their time and projects, so you should consider it.

Step four – final

If you passed all three previous steps – you can consider yourself as our future employee. There were only once, when we had not hire developer because of financial/welfare issue. This because of our stances of good developers. If you did not hire them – you and all company will loose, not the developer. Each of those will find his/her place very soon. Even if the good developer has no CV at all or he/she is a student or foreign resident. Because the only chance for this company to produce good results is to have good human resources.

Send me your CV or contact me tamir [AT] khason.biz to tell that you are the one I’m looking for >>

Please do not send me your CV if it passes the regex of the first step – just send a message with list of what you did for last 7 years.

July 23rd, 2009 · Comments (7)

How to calculate CRC in C#?

First of all, I want to beg your pardon about the frequency of posts last time. I’m completely understaffed and have a ton of things to do for my job. This why, today I’ll just write a quick post about checksum calculation in C#. It might be very useful for any of you, working with devices or external systems.

CRC – Cyclic Redundancy Check is an algorithm, which is widely used in different communication protocols, packing and packaging algorithms for assure robustness of data. The idea behind it is simple – calculate unique checksum (frame check sequence) for each data frame, based on it’s content and stick it at the end of each meaningful message. Once data received it’s possible to perform the same calculating and compare results – if results are similar, message is ok.

There are two kinds of CRC – 16 and 32 bit. There are also less used checksums for 8 and 64 bits. All this is about appending a string of zeros to the frame equal in number of frames and modulo two device by using generator polynomial containing one or more bits then checksum to be generated. This is very similar to performing a bit-wise XOR operation in the frame, while the reminder is actually our CRC.

In many industries first polynomial is in use to create CRC tables and then apply it for performance purposes. The default polynomial, defined by IEEE 802.3 which is 0xA001 for 16 bit and 0×04C11DB7 for 32 bit. We’re in C#, thus we should use it inversed version which is 0×8408 for 16 bit and 0xEDB88320 for 32 bit. Those polynomials we’re going to use also in our sample.

So let’s start. Because CRC is HashAlgorithm after all, we can derive our classes from System.Security.Cryptography.HashAlgorithm class.

public class CRC16 : HashAlgorithm {
public class CRC32 : HashAlgorithm {

Then, upon first creation we’ll generate hashtables with CRC values to enhance future performance. It’s all about values table for bytes from 0 to 255 , so we should calculate it only once and then we can use it statically.

[CLSCompliant(false)]
public CRC16(ushort polynomial) {
HashSizeValue = 16;
_crc16Table = (ushort[])_crc16TablesCache[polynomial];
if (_crc16Table == null) {
_crc16Table = CRC16._buildCRC16Table(polynomial);
_crc16TablesCache.Add(polynomial, _crc16Table);
}
Initialize();
}

[CLSCompliant(false)]
public CRC32(uint polynomial) {
HashSizeValue = 32;
_crc32Table = (uint[])_crc32TablesCache[polynomial];
if (_crc32Table == null) {
_crc32Table = CRC32._buildCRC32Table(polynomial);
_crc32TablesCache.Add(polynomial, _crc32Table);
}
Initialize();
}

Then let’s calculate it

private static ushort[] _buildCRC16Table(ushort polynomial) {
// 256 values representing ASCII character codes.
ushort[] table = new ushort[256];
for (ushort i = 0; i < table.Length; i++) {
ushort value = 0;
ushort temp = i;
for (byte j = 0; j < 8; j++) {
if (((value ^ temp) & 0×0001) != 0) {
value = (ushort)((value >> 1) ^ polynomial);
} else {
value >>= 1;
}
temp >>= 1;
}
table[i] = value;
}
return table;
}

private static uint[] _buildCRC32Table(uint polynomial) {
uint crc;
uint[] table = new uint[256];

// 256 values representing ASCII character codes.
for (int i = 0; i < 256; i++) {
crc = (uint)i;
for (int j = 8; j > 0; j–) {
if ((crc & 1) == 1)
crc = (crc >> 1) ^ polynomial;
else
crc >>= 1;
}
table[i] = crc;
}

return table;
}

The result will looks like this for 32 bits

        0x00, 0x31, 0x62, 0x53, 0xC4, 0xF5, 0xA6, 0x97,
        0xB9, 0x88, 0xDB, 0xEA, 0x7D, 0x4C, 0x1F, 0x2E,
        0x43, 0x72, 0x21, 0x10, 0x87, 0xB6, 0xE5, 0xD4,
        0xFA, 0xCB, 0x98, 0xA9, 0x3E, 0x0F, 0x5C, 0x6D,
        0x86, 0xB7, 0xE4, 0xD5, 0x42, 0x73, 0x20, 0x11,
        0x3F, 0x0E, 0x5D, 0x6C, 0xFB, 0xCA, 0x99, 0xA8,
        0xC5, 0xF4, 0xA7, 0x96, 0x01, 0x30, 0x63, 0x52,
        0x7C, 0x4D, 0x1E, 0x2F, 0xB8, 0x89, 0xDA, 0xEB,
        0x3D, 0x0C, 0x5F, 0x6E, 0xF9, 0xC8, 0x9B, 0xAA,
        0x84, 0xB5, 0xE6, 0xD7, 0x40, 0x71, 0x22, 0x13,
        0x7E, 0x4F, 0x1C, 0x2D, 0xBA, 0x8B, 0xD8, 0xE9,
        0xC7, 0xF6, 0xA5, 0x94, 0x03, 0x32, 0x61, 0x50,
        0xBB, 0x8A, 0xD9, 0xE8, 0x7F, 0x4E, 0x1D, 0x2C,
        0x02, 0x33, 0x60, 0x51, 0xC6, 0xF7, 0xA4, 0x95,
        0xF8, 0xC9, 0x9A, 0xAB, 0x3C, 0x0D, 0x5E, 0x6F,
        0x41, 0x70, 0x23, 0x12, 0x85, 0xB4, 0xE7, 0xD6,
        0x7A, 0x4B, 0x18, 0x29, 0xBE, 0x8F, 0xDC, 0xED,
        0xC3, 0xF2, 0xA1, 0x90, 0x07, 0x36, 0x65, 0x54,
        0x39, 0x08, 0x5B, 0x6A, 0xFD, 0xCC, 0x9F, 0xAE,
        0x80, 0xB1, 0xE2, 0xD3, 0x44, 0x75, 0x26, 0x17,
        0xFC, 0xCD, 0x9E, 0xAF, 0x38, 0x09, 0x5A, 0x6B,
        0x45, 0x74, 0x27, 0x16, 0x81, 0xB0, 0xE3, 0xD2,
        0xBF, 0x8E, 0xDD, 0xEC, 0x7B, 0x4A, 0x19, 0x28,
        0x06, 0x37, 0x64, 0x55, 0xC2, 0xF3, 0xA0, 0x91,
        0x47, 0x76, 0x25, 0x14, 0x83, 0xB2, 0xE1, 0xD0,
        0xFE, 0xCF, 0x9C, 0xAD, 0x3A, 0x0B, 0x58, 0x69,
        0x04, 0x35, 0x66, 0x57, 0xC0, 0xF1, 0xA2, 0x93,
        0xBD, 0x8C, 0xDF, 0xEE, 0x79, 0x48, 0x1B, 0x2A,
        0xC1, 0xF0, 0xA3, 0x92, 0x05, 0x34, 0x67, 0x56,
        0x78, 0x49, 0x1A, 0x2B, 0xBC, 0x8D, 0xDE, 0xEF,
        0x82, 0xB3, 0xE0, 0xD1, 0x46, 0x77, 0x24, 0x15,
        0x3B, 0x0A, 0x59, 0x68, 0xFF, 0xCE, 0x9D, 0xAC

Now, all we have to do is to upon request to lookup into this hash table for related value and XOR it

protected override void HashCore(byte[] buffer, int offset, int count) {

for (int i = offset; i < count; i++) {

ulong ptr = (_crc & 0xFF) ^ buffer[i];

_crc >>= 8;

_crc ^= _crc32Table[ptr];

}

}

new public byte[] ComputeHash(Stream inputStream) {

byte[] buffer = new byte[4096];

int bytesRead;

while ((bytesRead = inputStream.Read(buffer, 0, 4096)) > 0) {

HashCore(buffer, 0, bytesRead);

}

return HashFinal();

}

protected override byte[] HashFinal() {

byte[] finalHash = new byte[4];

ulong finalCRC = _crc ^ _allOnes;

finalHash[0] = (byte)((finalCRC >> 0) & 0xFF);

finalHash[1] = (byte)((finalCRC >> & 0xFF);

finalHash[2] = (byte)((finalCRC >> 16) & 0xFF);

finalHash[3] = (byte)((finalCRC >> 24) & 0xFF);

return finalHash;

}

We done. Have a good time and be good people. Also, I want to thank Boris for helping me with this article. He promised to write here some day…

Source code for this article

April 6th, 2009 · Comments (1)

Book review: C# 2008 and 2005 Threaded Programming

A couple of weeks ago, Packt publishing asked me to review Gastón C. Hillar book “C# 2008 and 2005 Threaded Programming: Beginner’s Guide”. They sent me a copy of this book and today, I’m ready to write a review for it. But before I’ll start reviewing it, I want to apologize to the publisher and author for the impartial review.

First of all, you should understand, that this book is about how it possible (for this book author) to write four programs (with awful user interface) using different classes from System.Threading namespace to perform tasks, rather then what is multithreaded programming and how to achieve best performance by utilizing multiple CPU power. Your own programs will not run faster after reading this book, but you’ll probably know (if you did not know before) how to use BackgroundWorker, Thread, ThreadPool, AutoResetEvent and WaitHandle classes. Also, there is a small chapter about thread context switching for UI thread delegates invocation and parallel extensions.

There are some technical misconceptions and errors in this book. But it is not the major problem of it. The problem is that while reading this book I question myself whom this book aimed at? Language style is somewhere between blog chatting (better then mine) and MSDN style documentation. I admit I don’t know quite how to categorize this, the author writes in a style that is just bizarre (even more bizarre then mine in this blog ) Overall, it sounds like I’m reading a conversation between two beginner-level programmers trying to explain one each other why they are using certain coding convention in C#.

Another half of this 395 pages book is just copy-paste stuff from Visual Studio (including it default tabulations and indentations). Here one of representative examples of such copy/paste

// Disable the Start button
butStart.Enabled = false;
// Enable the Start button
butStart.Enabled = true;

…

// Some very useful property, which used as private member for another public property
private int priVeryUserfulProperty;

public int VeryUserfulProperty
{
   get
   {
      return priVeryUserfulProperty;
   }
   set
   {
      priVeryUserfulProperty = value;
   }
}

Verdict: Not very exemplary introduction to some classes inside System.Threading namespace for fellow students who like to read blogs, rather then books and documentation and do not want to understand how it works under the hoods, but write something and forget it.

3- of 5 on my scale. This book is not all bad, though, but apparently suitable for very specific audience, which definitely excludes me.

March 20th, 2009 · Comments (7)

Some new in-mix downloads

There are some very cool downloads suddenly appear on MSDN download site due to all new technologies, presented at Mix ‘09. So let’s start

• Silverlight 3 SDK beta 1
• If you do not want to install full SDK, you can install only runtime for Windows or Mac. Then, you can read documentation online. You do not need it, if you’re going to install
• Silverlight 3 tools beta 1 for VS2008 SP1. After you have all this, go to the official Silverlight web site and start working.
• If you re “in” .NET RIA Services, you can download March ‘09 preview of it also to use with new Silverlight. It also makes sense to read about what is it in Brad’s blog.
• Also new Silverlight toolkit was released with SL3 support and a bunch of new up/down controls, LayoutTransformer, Accordion and TransitioningContentControl.
• Microsoft Expression Blend 3 Preview. It includes SL3 and WPF3.5 SP1 support, but excludes SketchFlow by now.

To learn more about Silverlight 3.0 and Blend 3.0, you can see first day keynotes at mix 09, Rollup of what’s new in Silverlight 3 by Joe Stegman. This includes offline mode support by Mike Harsh. I’ll write another separate post for this topic, due to the fact, that I’m a desktop guy, so wary about the future of WPF.

To learn more about how to use new Expression Blend, it worth to see this session by Pete Blois. Another good sessions are also wrapped for you by Scott Hanselman.

After we done with all web stuff, let’s speak about a client

• Microsoft MultiPoint SDK. Do you want to use multitouch in your application? This SDK provides you with ability to use up to 250 individual mouse devices simultaneously. And yes, it works with Windows XP SP2 too
• Internet Explorer 8 for Windows XP x32, XP x64, Vista x32, Vista x64
• In case, that you do no have Windows Vista or Windows Server 2008, you can download 30-day evaluation virtual hard disk of Windows Vista or Windows Server 2008 Enterprise and see how it works .
• Also a small present for my old friends (from my military consulting era) – WPF and Silverlight APIs for GIS engine of ESRI. Have a fun!

That’s all by now, going to write a review for new book and will publish it soon (probably even before, you’ll finish with all those downloads and readings). So, stay tuned and be good people.

March 20th, 2009 · Comments (1)

WPF Line-Of-Business labs and Silverlight vs. Flash

Small update today (mostly interesting links)… During my last “Smart Client” session I was asked about WPF LOB application development labs. So, there are two full labs, I noticed about:

• Southridge, which comes from Redmond team
• Order Manager, which comes from Swiss DPE team

Both labs include WPF ribbon and DataGrid, Southridge also come with M-VV-M design sample and some other interesting features. As for me, it seemed, like some parts of those labs can be easily used “as-is” for production level applications, like it was done with SCE starter, which turned into TimesReader (by the way, it has free version again).

For those, who still trying to consider what to use for their next killer app, I propose to read following article from Jordan, which compares between Silverlight and Flash. And then see composite application guidance to use Prism for Silverlight development. Here the video of it usage by Adam Kinney from Channel 9

Have a nice day and be good people

February 18th, 2009 · Comments (5)

Quick how to: Reduce number of colors programmatically

My colleague just asked me about how to reduce a number of colors in image programmatically. This is very simple task and contains of 43 steps:

First of all, you have to read a source image

using (var img = Image.FromFile(name)) {
var bmpEncoder = ImageCodecInfo.GetImageDecoders().FirstOrDefault(e => e.FormatID == ImageFormat.Bmp.Guid);

Then create your own encoder with certain color depth (32 bits in this case)

var myEncoder = System.Drawing.Imaging.Encoder.ColorDepth;
var myEncoderParameter = new EncoderParameter(myEncoder, 32L);
var myEncoderParameters = new EncoderParameters(1) { Param = new EncoderParameter[] { myEncoderParameter } };

Then save it

img.Save(name.Replace(“.png”, “.bmp”), bmpEncoder, myEncoderParameters);

It it enough? Not really, because if you’re going to loose colors (by reducing color depth), it makes sense to avoid letting default WIX decoder to do this, thus you have to find nearest base colors manually. How to do this? By using simple math

Color GetNearestBaseColor(Color exactColor) {
Color nearestColor = Colors.Black;
int cnt = baseColors.Count;
for (int i = 0; i < cnt; i++) {
int rRed = baseColors[i].R – exactColor.R;
int rGreen = baseColors[i].G – exactColor.G;
int rBlue = baseColors[i].B – exactColor.B;

int rDistance =
(rRed * rRed) +
(rGreen * rGreen) +
(rBlue * rBlue);
if (rDistance == 0.0) {
return baseColors[i];
} else if (rDistance < maxDistance) {
maxDistance = rDistance;
nearestColor = baseColors[i];
}
}
return nearestColor;
}

Now, you can either change colors on base image directly

unsafe {
uint* pBuffer = (uint*)hMap;
for (int iy = 0; iy < (int)ColorMapSource.PixelHeight; ++iy)
{
for (int ix = 0; ix < nWidth; ++ix)
{
Color nc = GetNearestBaseColor(pBuffer[0].FromOle());

pBuffer[0] &= (uint)((uint)nc.A << 24) | //A
(uint)(nc.R << 16 ) | //R
(uint)(nc.G << 8 ) | //G
(uint)(nc.B ); //B
++pBuffer;
}
pBuffer += nOffset;
}
}

Or, if you’re in WPF and .NET 3.5 create simple pixel shader effect to do it for you in hardware. Now, my colleague can do it himself in about 5 minutes . Have a nice day and be good people.

February 9th, 2009 · Comments (2)

Bootstrapper for .NET framework version detector

You wrote your .NET program, that can be used as stand alone portable application (such as it should be for Smart Client Apps), however you have to be sure, that necessary prerequisites (such as .NET framework) are installed on client’s machine. What to do? How to detect .NET framework version installed on target machine before running .NET application. The answer is – to use unmanaged C++ bootstrapper, that invoke your application if correct version of framework is installed.

Until now there are 15 possible .NET frameworks can be installed on client’s machine. Here the table of possible and official supported versions, as appears in Q318785

All of those versions are detectible by queering specific registry keys. However, in some cases, you need to load mscoree.dll and call “GETCOREVERSION” API to determine whether specific version of .NET is installed. You can read more about it in MSDN.

So it’s really simple to write small C++ application (or PowerShell applet), that queries registry and invoke your managed application. How to do this? You can either read about it in outstanding blog of Aaron Stebner, who is Project Manager in XNA platform deployment team or attend my session next week to learn do it yourself. We’ll speak about nifty ways to do it also.

Anyway, by now, you can use small stand alone program, I wrote a while ago, that will tell you all versions of .NET frameworks installed in target machine without any prerequisites. It can be run even from shared network location

Download whoooot.exe (13K) >>

See you next week.

PS: Do not forget to download and install the new version of Visual Studio Snippet Designer, which is extremely useful tool by MVP Bill McCarthy, you’ll need it later next week…

Have a nice day and be good people.

February 4th, 2009 · Comments (6)

Making TFS better or what is TITS?

Those days me and my team work very hard toward new version of “The System”. This includes massive refactoring of all solutions, hard work with TFS (which not restricted to only adding files, but also deleting, moving, etc. other words, all stuff, which TFS is not really love). Because of this, we need a bunch of handy tools to make our dreams come true and to decrease unnecessary number of clicks inside Team System Explorer and Visual Studio. You do not really think, that we have no tools to make our everyday job easier. We have. However, we never package and release it. Let me introduce “TITS” – Tools, Invaluable for Team System. This suite I’m planning to release as another open source project within couple of months.

What “TITS” includes? First of all –

“QOF” – Quick Open File

This tools is absolutely invaluable if you have big solutions. While all it knows to do is to search. But, wait, what’s wrong with build-in search of Visual Studio? First of all, it does not search Solution items and files, are in solution directory, but not in project. Also it cannot fix your typos and errors. Also it does not know to move you quickly to found solution item in Solution Explorer or in Source Editor.

Basic set of QOF features:

• No mouse – open any file
• No mouse – locate any file in solution explorer
• Highlighting found items
• Multiple files open
• Filter by source files only, resources, owner or any other kind of filters
• Search inside TFS, including history, changesets, shelves (either private and public)
• …and much much more

Next tool is:

“WIBREW” – Who Is Breaking What

Absolutely invaluable tool to know who actually breaking what file inside TFS. For example, I do not want to lock files, while I still want to know who holds what file. TFS provides such feature out-of-the-box, however from command prompt only. You can add it even as macro. Like this:

However it not user friendly and impossible for use, ‘cos it looks as following:

You do not know what actually developer doing, where and why. With “WIBREW”, you can know:

• When developer started to break files
• What exactly he’s doing
• Is the breaking file locked or now
• Where the developer breaks it (workspace and computer name of the user)
• …and much much more

Another tool is:

“WITCH” – What I have To Check-in

If you ever worked with Team Force, you know what this tool is doing. It shows you a preview of all changed files, you’ll check-in. For some reason, TFS has no such feature. Let’s imagine, that your work method is to check out everything, change something and check-in only changed files. Until here TFS does everything, however if you want to preview changeset (for example in order to compare with “WIBREW” output), you can not. Here “WITCH” comes to help.

[Here should be a screenshot of “WITCH”, but it looks exactly the same as “WIBREW” with shameless blurring]

Another invaluable tool is:

“VOCUS” – VOid CUstom Settings for check in

This tool is absolutely UI-less. It allows developers to work with their own custom settings in Visual Studio, while for check-in and check-out it format all documents, according predefined custom settings (for example indentation). How many times, you tried to merge files, when all the difference is indentation it tab size? Well, this tool solves this problem.

It stores custom settings for each user (BTW, it also makes able for each developer to restore his settings fluently in any computer) and reformat documents on check-in action toward corporate settings, when on check-out toward custom developer’s setting.

“SHMOC” – SHow MOre Code

This is not actually tool, works with TFS. It rather works with your Visual Studio Development Environment. It’s UI-less as well and makes able to hide and restore all docking windows in VS. It makes you able to write in “Dark Room” mode (which is full screen, distraction free environment) and return to Visual Studio within one button press. It can also change VS color scheme, if required.

There are some other tools should be inside this suite, however, I still have no names for them Also, if you have something interesting, and you want to contribute it to this suite, you’re highly welcome.

PS: This blog is about code, but this post is 6th in row without even one line of code, so I have to fix it as soon as possible. Thus, I’ll example how WIBREW works under the hood. Other words, small example of how to work with TFS API from Visual Studio plugin.

First of all, as in any VS plugin, you need to acquire DTE2 application object:

_applicationObject = (DTE2)application;
_addInInstance = (AddIn)addInInst;

When you have it, you need to detect what TFS server you’re working with and what are user credentials for this session. The common problem of WIBREW for poor men, was how to work with this tool over VPN (when your connected session is only inside VS). So each time, you tried to run it, you had to enter your domain credentials – very inconvenience way of work.

In order to prevent it, let’s ask your environment about Team Foundation information:

private TeamFoundationServerExt _tfsExt;
…
_tfsExt = (TeamFoundationServerExt)_applicationObject.GetObject("Microsoft.VisualStudio.TeamFoundation.TeamFoundationServerExt");

Also, you can be notified when your work project context was changed. To do this, just subscribe to ProjectContextChanged event and handle it inside:

_tfsExt.ProjectContextChanged += OnProjectContextChanged;
…
public void OnProjectContextChanged(object sender, EventArgs e) {
         if (!string.IsNullOrEmpty(_tfsExt.ActiveProjectContext.ProjectName)) {

Now when we know, that we have out active project context, all we have to do is to ask about changes

private VersionControlExt _vcExt;
…
_vcExt = (VersionControlExt)_applicationObject.GetObject("Microsoft.VisualStudio.TeamFoundation.VersionControl.VersionControlExt");

Inside VersionControlExt object you have following self-descriptive properties and methods: FindChangeSet, History, PendingChanges, SolutionWorkspace etc. however it works only with TFS solution explorer. To handle pending changes for the project without tickling TFS, we can use it internal methods. All the difference is with references. To work with Visual Studio TFS explorer methods, you should reference:
Microsoft.VisualStudio.TeamFoundation.dll, Microsoft.VisualStudio.TeamFoundation.Client.dll and Microsoft.VisualStudio.TeamFoundation.VersionControl.dll, while working with TFS API directly, use Microsoft.TeamFoundation.dll, Microsoft.TeamFoundation.Client.dll and Microsoft.TeamFoundation.VersionControl.dll from [PROGRAM FILES]\Microsoft Visual Studio 9.0\Common7\IDE\PrivateAssemblies\. Just like this:

VersionControlServer _vcs…_vcs = (VersionControlServer)_server.GetService(typeof(VersionControlServer));…var _sets = _vcs.QueryPendingSets( new[] { new ItemSpec(serverPath, RecursionType.Full) }, null, null);…foreach (PendingSet set in sets) {…
//Get everything you need here

We done. It’s very easy to work with Team System from inside Visual Studio. Also it’s very easy to build useful tools, not built by Microsoft for some reason

Have a nice day, be good people and wait for me to beatify sources before releasing as another Open Source application.

January 27th, 2009 · Comments (9)