To prove this point, I set out to find if there is a relation between the length of each post and the comments it receives. Though, the number of comments is not the best yardstick to measure the popularity of the post, it certainly is the most concrete one. The best place to try this out seemed to be my organization’s internal blogosphere, which had a lot of sample data (3500+ posts). Since its impossible to manually collect data for that many posts, I wrote a program to write it to a text file from where any correlation could be identified. Its a mixture of retrieving the data through HTML and RSS and parsing it. Though it should be compatible with any WordPress version, the comment counting function might need some tweaking to make it work on later versions. Once the code finishes executing, you would have a text file with the name of each post, the length and the number of comments. The data in this file can be imported into Excel for further manipulation.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67

using System;
using System.Xml;
using System.Net;
using System.IO;
using System.Text.RegularExpressions;
using System.Windows.Forms;
 
public class TestRSS
{
    StreamWriter _swObj = new StreamWriter("Results.txt", true);
    public const string _feedURL = "http://blog.ganeshzone.net";
 
    public static void Main()
    {
        new  TestRSS().RunTests();
    }
 
    public void RunTests()
    {
        for (int i = 1; ; i++) //This is an infinite loop only to be broken when there are no more posts
        {
            XmlDocument _xdoc = new XmlDocument(); //a New XmlDocument object
            //Lets load the XML from the url
            _xdoc.LoadXml(GetXmlData(_feedURL + (i>1?"?paged=" + i+"&":"?" )+"feed=rss2"));
            XmlNodeList _xList = _xdoc.GetElementsByTagName("item");
            if (_xList.Count == 0)
                break; //This means there are no more blog posts
            foreach (XmlNode _tempNode in _xList)
                //This method writes the name of post, length and the number of comments delimited by a colon
                //So we need to remove the colon in case of any being present/
                //Example Data: - ASP.NET Evolution WebForms v/s MVC: 6060:0
                WriteToFile(_tempNode.FirstChild.InnerText.Replace(":", "") + ": " + _tempNode.SelectSingleNode("description").NextSibling.InnerText.Length + ":" + GetComments(_tempNode.FirstChild.NextSibling.InnerText));
        }
    }
 
    public string GetXmlData(string _url)
    {
        //An ordinary retrieval of data from the url using the HttpWebRequest
        //This is a workaround for sites with badly formed RSS feeds due to script tags.
        //Once we get the data, we can load it into an XmlDocument class
        HttpWebRequest _blogReq = (HttpWebRequest)WebRequest.Create(_url);
        HttpWebResponse _blogResp = (HttpWebResponse)_blogReq.GetResponse();
        StreamReader _respStream = new StreamReader(_blogResp.GetResponseStream());
        //Replace Script tags
        return Regex.Replace(_respStream.ReadToEnd(), @"<script[^>]*?>[\s\S]*?<\/script>", "");
    }
 
    public int GetComments(string _url)
    {
        //Retrieving the Comments from the HTML and not the RSS feed. I wasnt able
        //to find a workaround for 10 comment limit in the RSS feed in wordpress 2.3,
        //Hence retriving the HTML and matching the comment divs. Keep in mind this wont work
        //for higher Wordpress versions. Just take the comment tag and substitute accordingly
        Match _match = Regex.Match(GetXmlData(_url), "<div class=\"mycomment\"[^>]*?>[\\s\\S]*?<\\/div>");
        int _counter=0;
        for (; _match.Success; _counter++)
            _match = _match.NextMatch();
        return _counter;
    }
 
    public void WriteToFile(string _tobeWritten)
    {
        //Just write it to the file
        _swObj.WriteLine(_tobeWritten);
        _swObj.Flush();
    }
}

First, we need to set up Visual studio  to use the Sons of Strike. A few simple steps mentioned below are needed to accomplish this:-

• Enable Unmanaged Debugging in the Project Properties

• Set the symbol server to  *C:\localcache*http://msdl.microsoft.com/download/symbols. The symbol server can be set by going to Tools -> Options-> Debugging. This step allows Visual studio to download the symbols needed for debugging from the Microsoft server.

• Set a breakpoint in the code and press F5 to start the program in debug mode. Once the application hits the breakpoint, load the SOS.dll using the command .load SOS in the Immediate Window.

Now we are ready to use the debugging extension. And SOS is also compatible with debugging tools like WinDbg in case you are comfortable with using that. Click here for a list of SOS commands.

Now, lets see the code that we would use for seeing how generics work. It consists of a Generic Type Customer which contains a list of the generic type and a public integer property.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

class Employee
    {
 
        public int EmployeeID;
        public static int age;
        public static string Name;
    }
 
    class Customer<T>
    {
        public int CustomerID;
        List<T> _innerList = new List<T>();
 
        public List<T> ListItems { get { return _innerList; } }
 
        public void AddItems(T _addObject)
        {
            _innerList.Add(_addObject);
        }
    }

The main method to call this code is below. We create the Customer class in three types – a string, an integer and a custom Employee type. Of these the string and Employee are reference types while integer is a primitive value type.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

static void Main(string[] args)
        {
            Customer<string> _cxStringObj = new Customer<string>();
            Customer<int> _cxIntObj = new Customer<int>();
            Customer<Employee> _cxEmpObj = new Customer<Employee>();
 
            _cxStringObj.CustomerID = 34;
            _cxStringObj.AddItems("Ganesh");
 
            _cxIntObj.CustomerID = 23;
            _cxIntObj.AddItems(43);
 
            _cxEmpObj.CustomerID = 34;
            _cxEmpObj.AddItems(new Employee());
        }

Every .NET class has two main data structures it uses to maintain type identity – the MethodTable and EEClass. Every object has a pointer to its MethodTable, and this methodtable contains a pointer to the EEClass. In .NET 1.1, the EE class was the strongest identifier of Type Identity of an object, however with .NET 2.0 and the implementation of Generics, Generic Types can share the same EEClass, and the MethodTable takes on the mantle of being the unique type identifier.

There are two main ways of implementing Generics which were proposed by the researchers who were working at Microsoft (This paper provides details of the implementation).

• Expansion: In this way, whenever the runtime encounters a generic type, the code could be generated on the fly for the type being called. For e.g. in our code, when the object Customer<string> was encountered, it could have expanded into the string representation of the object. When the type Customer<int> was encountered, a similar expansion can be done for it.

• Code sharing: In this approach only a single instantiation is done for the generic type, and it is reused for all the calls.

The method chosen in the end was a mixture of both these approaches. Generic types share as much code as possible They contain the same EEClass as mentioned before, but also have distinct MethodTables. Even though they have distinct MethodTables, code is shared between compatible types (all reference types).

Fire up the debugger and load the SOS.dll. To see the particular instances of the type on the heap give the Command DumpHeap -type . Once all the three objects have been created, lets see the result of this command.

!DumpHeap -type Customer
 Address       MT     Size
01ed35cc 002331d8       16
01ed3604 00233284       16
01ed3638 0023336c       16
total 3 objects
Statistics:
      MT    Count    TotalSize Class Name
0023336c        1           16 GenericsTry.Customer`1[[GenericsTry.Employee, GenericsTry]]
00233284        1           16 GenericsTry.Customer`1[[System.Int32, mscorlib]]
002331d8        1           16 GenericsTry.Customer`1[[System.String, mscorlib]]
Total 3 objects

As you can see from the MT column, all the three objects have different MethodTables. Lets see the MethodTables for each of the three type instantiations.

The command is DumpMT -MD . First the Employee type. Think of the MethodTable as an extension to a conventional vtable which simply contains the method pointers. The MethodTable’s vtable points to 7 slots, 4 of which are inherited from the parent Object class and three which are defined by the object itself. If you see the JIT status, there is a YES and NONE. The NONE means the method has never been called by the JIT Compiler and it points to a temporary stub. The first time the method is called, it is compiled on the fly and the temporary stub is replaced with the actual method pointer. To optimize performance, JIT defers execution and compilation till as late as possible.

!DumpMT -MD 0023336c
EEClass: 00231420
Module: 00232c5c
Name: GenericsTry.Customer`1[[GenericsTry.Employee, GenericsTry]]
mdToken: 02000004  (Projects\GenericsTry\GenericsTry\bin\Debug\GenericsTry.exe)
BaseSize: 0x10
ComponentSize: 0x0
Number of IFaces in IFaceMap: 0
Slots in VTable: 7
--------------------------------------
MethodDesc Table
   Entry MethodDesc      JIT Name
70766aa0   705e4a34   PreJIT System.Object.ToString()
70766ac0   705e4a3c   PreJIT System.Object.Equals(System.Object)
70766b30   705e4a6c   PreJIT System.Object.GetHashCode()
707d7550   705e4a90   PreJIT System.Object.Finalize()
0023c058   0023313c     NONE GenericsTry.Customer`1[[System.__Canon, mscorlib]].get_ListItems()
0023c060   00233148     NONE GenericsTry.Customer`1[[System.__Canon, mscorlib]].AddItems(System.__Canon)
0023c068   00233154      JIT GenericsTry.Customer`1[[System.__Canon, mscorlib]]..ctor()

Lets see the MethodTable for the integer type. As you can see the EEClass for the Integer and Employee types are not same, this is because they are not compatible types and hence no code sharing can take place

!DumpMT -MD 00233284
EEClass: 002314b4
Module: 00232c5c
Name: GenericsTry.Customer`1[[System.Int32, mscorlib]]
mdToken: 02000004  (\GenericsTry\GenericsTry\bin\Debug\GenericsTry.exe)
BaseSize: 0x10
ComponentSize: 0x0
Number of IFaces in IFaceMap: 0
Slots in VTable: 7
--------------------------------------
MethodDesc Table
   Entry MethodDesc      JIT Name
70766aa0   705e4a34   PreJIT System.Object.ToString()
70766ac0   705e4a3c   PreJIT System.Object.Equals(System.Object)
70766b30   705e4a6c   PreJIT System.Object.GetHashCode()
707d7550   705e4a90   PreJIT System.Object.Finalize()
0023c080   00233254     NONE GenericsTry.Customer`1[[System.Int32, mscorlib]].get_ListItems()
0023c088   00233260     NONE GenericsTry.Customer`1[[System.Int32, mscorlib]].AddItems(Int32)
0023c090   0023326c      JIT GenericsTry.Customer`1[[System.Int32, mscorlib]]..ctor()

The last one is the string type. This brings us to interesting conclusions. Compare the first MethodTable with the last one. Both are reference types and hence compatible with other. It can be observed that the string and Employee tables have the same EEClass 00231420. The integer type however has the EEClass 002314b4. There is no way for primitive value types to share code with each other since they differ in size. For e.g. If we created a bool type for Customer class, it would still have a different EEClass and MethodTables.

!DumpMT -MD 002331d8
EEClass: 00231420
Module: 00232c5c
Name: GenericsTry.Customer`1[[System.String, mscorlib]]
mdToken: 02000004  (\GenericsTry\GenericsTry\bin\Debug\GenericsTry.exe)
BaseSize: 0x10
ComponentSize: 0x0
Number of IFaces in IFaceMap: 0
Slots in VTable: 7
--------------------------------------
MethodDesc Table
   Entry MethodDesc      JIT Name
70766aa0   705e4a34   PreJIT System.Object.ToString()
70766ac0   705e4a3c   PreJIT System.Object.Equals(System.Object)
70766b30   705e4a6c   PreJIT System.Object.GetHashCode()
707d7550   705e4a90   PreJIT System.Object.Finalize()
0023c058   0023313c     NONE GenericsTry.Customer`1[[System.__Canon, mscorlib]].get_ListItems()
0023c060   00233148     NONE GenericsTry.Customer`1[[System.__Canon, mscorlib]].AddItems(System.__Canon)
0023c068   00233154      JIT GenericsTry.Customer`1[[System.__Canon, mscorlib]]..ctor()

One more thing of notice is the Method Descriptors for the first and third types. The AddItems method in Customer class has the same Method descriptor for both the string and Employee implementation. So even after a MethodTable is duplicated, both those entries still point to the same Method Description. Even the name has the type System.__Canon, which is dynamically replaced by the type on which the method is called. This is an illustration of the second approach of code sharing given by the researchers from Microsoft.

Once the code for AddItems method is called for the first time, it is JIT Compiled and assigned a code address. The Method Descriptor holds this code address which is assembly code and can be viewed by the SOS command !u.  Till the method is Jitted, there is junk value in this field on the MethodDescriptor.

!DumpMD 00233148
Method Name: GenericsTry.Customer`1[[System.__Canon, mscorlib]].AddItems(System.__Canon)
Class: 00231420
MethodTable: 0023316c
mdToken: 06000005
Module: 00232c5c
IsJitted: yes
CodeAddr: 00aa0298

Running the CodeAddress would give you a bunch of assembly code that is executed when the method is called.

!u 00aa0298
Normal JIT generated code
GenericsTry.Customer`1[[System.__Canon, mscorlib]].AddItems(System.__Canon)
Begin 00aa0298, size 42
>>> 00AA0298 55               push        ebp
00AA0299 8BEC             mov         ebp,esp
00AA029B 57               push        edi
00AA029C 56               push        esi
00AA029D 53               push        ebx
00AA029E 83EC34           sub         esp,34h
00AA02A1 33C0             xor         eax,eax
00AA02A3 8945F0           mov         dword ptr [ebp-10h],eax
00AA02A6 33C0             xor         eax,eax
00AA02A8 8945E4           mov         dword ptr [ebp-1Ch],eax
00AA02AB 894DC4           mov         dword ptr [ebp-3Ch],ecx
00AA02AE 8955C0           mov         dword ptr [ebp-40h],edx
00AA02B1 833D142E230000   cmp         dword ptr ds:[00232E14h],0
00AA02B8 7405             je          00AA02BF
00AA02BA E8BABC3571       call        71DFBF79 (JitHelp: CORINFO_HELP_DBG_IS_JUST_MY_CODE)
00AA02BF 90               nop
00AA02C0 8B45C4           mov         eax,dword ptr [ebp-3Ch]
00AA02C3 8B4804           mov         ecx,dword ptr [eax+4]
00AA02C6 8B55C0           mov         edx,dword ptr [ebp-40h]
00AA02C9 3909             cmp         dword ptr [ecx],ecx
00AA02CB E8E0F1D16F       call        707BF4B0 (System.Collections.Generic.List`1[[System.__Canon, mscorlib]].Add(System.__Canon), mdToken: 060019f1)
00AA02D0 90               nop
00AA02D1 90               nop
00AA02D2 8D65F4           lea         esp,[ebp-0Ch]
00AA02D5 5B               pop         ebx
00AA02D6 5E               pop         esi
00AA02D7 5F               pop         edi
00AA02D8 5D               pop         ebp
00AA02D9 C3               ret

In this post, we have looked at how the CLR treats Generics and how it takes care of performance implications of Generics code by sharing as much code as possible without having to needlessly box and unbox the types being passed to the Generic classes.

So I decided to poke around the API and write a simple extension. Though it doesnt compare to FireFox yet with much of the browser still not accessible, but I am sure future versions will bring much better customization capability. The extension I wrote is very simple and just displays the latest trending topics from twitter. The source can be downloaded here and the packaged version here.

Lets see the anatomy of a Chrome Extension. An Extension consists of three main parts

• Manifest
• UI
• All your other files like scripts, stylesheets, images, etc

The manifest is a json file which contains the metadata about your extension. Details like Title, Author etc

{
  "name": "Twitter Trends",
  "version": "1.0",
  "description": "Displays Current Trending Topics",
  "icons": { "128": "icon.png" },
  "browser_action": {
    "default_title": "Twitter Trends",
    "default_icon": "icon.png",
    "popup": "Background.html"
  },
  "permissions": [
    "tabs",
    "http://twitter.trends.free.fr/*"
  ]
}

As you can see, it contains basic details like name, version, icon etc. The popup attribute tells us the page to open when the browser button is clicked. As you can see I have specified a normal HTML page here. A popup extension is nothing but a simple HTML page opened when the toolbar button is clicked. Another interesting part is the permissions section. Here we can specify what all websites the extension can make AJAX calls to retrieve data. Since it matches based on wildcards, a * denotes all possible subdirectories under the website.

Now the HTML part. Since its a normal HTML page, you can include references to scripts, images, style sheets etc. The extension directory works as the root and all paths must be relative to it. Since its a simple page, I included all the scripts and styling information in a single page.

First the styling information. Always helps to keep this separate.

  body {
  font-family: helvetica, arial, sans-serif;
  font-size: 12px;
  overflow: hidden;
  background-color:#ddeef6;
  }
 
  #title
  {
      font-size:14px;
      overflow:hidden;
  }
 
  #mainDiv
  {
      min-width:175px;
      margin-top:5px;
  }
 
  .topicDisplayClass
  {
      display:block;
     min-height:2.2em;
     cursor:pointer;
  }
 
  .linkClass
  {
      color:#2276bb;
      text-decoration:none;
      vertical-align:middle;
      padding-left:2px;
  }
 
  .topicDisplayClass:Hover
  {
      background-color:#edfeff;
  }

Next the javascript. This is the most important part of the extension. What we do here is retrieve the RSS feed for the trending topics and parse it as an xmlDocument. After iterating thorough this xml, a span element is dynamically created for each item and it is appended into our main content holder. This main function call is delayed using the setTimeout function. We do this in order to let the user see the popup and see some progress which data is being fetched. Writing this in the onload event would make the user wait for the whole process to complete before seeing the popup.

/* Developer  : Ganesh Ranganathan
* Date       : 02/12/2010
* Description: This is a Google Chrome Extension created to display
            latest trending topics on twitter. */
 
//Global variables: the xmlHttpRequest and Feed URL
var _req;
var _feedURL = 'http://twitter.trends.free.fr/feed/';
 
function main() {
    //Call the loading method after a slight delay to allow
    //Chrome to load the extension body. Else it looks ugly with
    //the pressed button waiting for the feed to load.
    setTimeout("delayedCall()", 20);
 
}
 
function delayedCall() {
    //No browser compatibility code Yay!!!
    _req = new XMLHttpRequest();
    //Hooking up the readystate changed event handler.
    _req.onreadystatechange = function() {
        if (_req.readyState == 4 && _req.status == 200) {
            //Come here only if all data is loaded.
            var doc = _req.responseText;
            var parser = new DOMParser();
            //Parsing the Data into the XML dom for manipulation
            var xmlDoc = parser.parseFromString(doc, "text/xml");
            var mainDiv = document.getElementById('mainDiv');
            mainDiv.innerHTML = "";
            //Getting all the items
            var items = xmlDoc.getElementsByTagName('item');
 
            for (var i = 0; i < items.length; i++) {
                //Iterating through all the items and getting the title
                //and link. The description isnt a big deal in this feed.
                var title = items[i].getElementsByTagName('title');
                var link = items[i].getElementsByTagName('link');
                //Create an anchor and a span
                var spanNode = document.createElement('span');
                var linkNode = document.createElement('a');
                //Set Link to span
                linkNode.setAttribute('href', link);
                linkNode.innerText = title[0].childNodes[0].nodeValue;
                linkNode.className = 'linkClass';
                spanNode.appendChild(linkNode);
                spanNode.className = 'topicDisplayClass';
                var str = link[0].childNodes[0].nodeValue
                //Hooking the click event handler to the span
                //It calls the chrome tabs API to open the clikced
                //URL in a new tab
                spanNode.addEventListener("click",new Function("clickHandler('"+str+"')"),false);
                mainDiv.appendChild(spanNode);
            }
        }
    }
 
    //Without these two the AJAX request would never get sent.
    _req.open("GET", _feedURL, true);
    _req.send(null);
}
function clickHandler(str) {
    chrome.tabs.create({ url: str });
}

The markup is largely simple. Just the main content holder which contains a progress bar which is displayed to user while we load the data. Once that is done, the progress bar is replaced with the span nodes containing the trending topics.

<span id="title">Trending Topics</span>
<div id="mainDiv">
<!-- A Progress Bar to show the user something is being done -->
<img src="ajax-loader.gif" alt="" width="175px" /></div>

Now the manifest, html and image files need to be in a single folder. After that go to Chrome and click the customize dropdown button, and select Extensions. In the extensions page, select Load Unpacked Extension and select the folder where the extension files reside.

Once the extension is loaded, it shows as unpacked and you can do all the testing here by reloading it everytime any changes are done.

After testing, we can choose the pack option to package the extension for download. A .crx file would be generated which can be distributed for download. Below are some screenshots of the extension we just created.

Data is loading

I implemented a basic Binary Tree in .NET for adding custom types. The Tree is generic and can be used to add any type objects. Also the object must implement the generic versions of IComparable and IEquitable for their types and object types. This leaves the search logic’s implementation to the type rather than the data structure. The complete solution can be downloaded here.

Lets see the declaration of the Binary Tree along with the private members.

class BinaryTree<T> where
    T:IComparable<T>,
    IEquatable<T>,
    IEquatable<object>,
    IComparable<object>
{
    BinaryTree<T> _parentNode;
    BinaryTree<T> _leftNode;
    BinaryTree<T> _rightNode;
    public T Value { get; set; }

Each BinaryTree object must be aware of its parent node and its left and right children, and the value it holds. Lets see the logic for adding the node. The adding logic is recursive. If it finds a null value, then the value is substituted. If not, then the node to be added is compared. If the value is lesser than the same method is called on the leftnode. This happens recursively till we get the node where the value is to be added.

public void AddNode(T _nodeTobeAdded)
{
    if (null==Value)
        //We found the value. Yay!!
        this.Value = _nodeTobeAdded;
    else
    {
        //Compare the node to be added. Here the
        //type's IComparable is called.
        if (_nodeTobeAdded.CompareTo(Value) < 0)
        {
            //Go down the left node. If the left node is null,
            //then create a new object, else return the existing object.
            _leftNode = null==_leftNode ?new BinaryTree<T>(this):_leftNode;
            _leftNode.AddNode(_nodeTobeAdded);
        }
        else
        {
            //Go Down the right node.
            _rightNode = null==_rightNode?new BinaryTree<T>(this):_rightNode;
            _rightNode.AddNode(_nodeTobeAdded);
        }
    }
}

The GetNode method follows a similar logic, with recursive searching of the tree till a match is found, if no match is found then a null value is returned. Note that only around half of the tree is searched, rather than all of the nodes, as would be the case with a linkedList. GetNode is implemented both for the type and an object. In both cases the Equals method’s implementation is called, giving the generic type the ability to decide on its own equality logic.

public T GetNode(T _nodeToBeGotten)
{
   if(_nodeToBeGotten.Equals(Value))
       return Value;
   else if(_nodeToBeGotten.CompareTo(Value)<0)
       return null==_leftNode? default(T): _leftNode.GetNode(_nodeToBeGotten);
   else
       return null==_rightNode?default(T): _rightNode.GetNode(_nodeToBeGotten);
}

public T GetNode(object o)
{
    if (Value.Equals(o))
        return Value;
    else if (Value.CompareTo(o)>0)
        return (null == _leftNode) ?
            default(T) : _leftNode.GetNode(o);
    else
        return null == _rightNode ?
            default(T) : _rightNode.GetNode(o);
}

Deleting is a bit tricky business. We have four scenarios here:-

1. Its a leaf node, i.e. both the left and right children are null. This is the simplest scenario. Simply set the node itself to null
2. The LeftNode is null but the right node is not, in this scenario – the node’s right node is set to the parent node’s right node, and the node is set to null.
3. The Right node is null but left node is not. It is handled similar to the previous scenario.
4. Both leftNode and rightNode is not null. In this scenarios we need to go down the leftnode, and iterate to its right most non empty node. This node is to be replaced with the node to be deleted. Just the value needs to be replaced and the right most node should be orphaned.

public void DeleteNode(object o)
{
    BinaryTree<T> _tempNode1 = this.GetTreeNode(o);
    DeleteNode(_tempNode1);
}

public void DeleteNode(T _nodeTobeDeleted)
{
    BinaryTree<T> _tempNode1 = this.GetTreeNode(_nodeTobeDeleted);
    DeleteNode(_tempNode1);
}

private void DeleteNode(BinaryTree<T> _tempNode1)
{
    BinaryTree<T> _tempNode2 = _tempNode1;
    if ((null == _tempNode1._leftNode) && (null == _tempNode1._rightNode))
        _tempNode1._parentNode = null;
    else if (null == _tempNode1._leftNode)
        _tempNode1._parentNode._rightNode = _tempNode1._rightNode;
    else if (null == _tempNode1._rightNode)
        _tempNode1._parentNode._leftNode = _tempNode1._leftNode;
    else
    {
        bool _goingRight = false;
        _tempNode1 = _tempNode1._leftNode;
        while (_tempNode1._rightNode != null)
        {
            _tempNode1 = _tempNode1._rightNode;
            _goingRight = true;
        }
        _tempNode2.Value = _tempNode1.Value;
        if (_goingRight) _tempNode1._parentNode._rightNode = null;
        else _tempNode1._parentNode._leftNode = null;
        _tempNode1 = null;
    }
}

//Private methods to get the nodes to be deleted
private BinaryTree<T> GetTreeNode(T _nodeToBeGotten)
{
    if (_nodeToBeGotten.Equals(Value))
        return this;
    else if (_nodeToBeGotten.CompareTo(Value) < 0)
        return null == _leftNode ? default(BinaryTree<T>) : _leftNode.GetTreeNode(_nodeToBeGotten);
    else
        return null == _rightNode ? default(BinaryTree<T>) : _rightNode.GetTreeNode(_nodeToBeGotten);
}

private BinaryTree<T> GetTreeNode(object o)
{
    if (Value.Equals(o))
        return this;
    else if (Value.CompareTo(o) > 0)
        return (null == _leftNode) ?
            default(BinaryTree<T>) : _leftNode.GetTreeNode(o);
    else
        return null == _rightNode ?
            default(BinaryTree<T>) : _rightNode.GetTreeNode(o);
}

The class which could be added to the tree needs to implement the four interfaces i mentioned. I implemented an Employee class which has just two fields Name and age and the comparison is done based on the age.

class Employee : IComparable<Employee>,IEquatable<Employee>,IEquatable<object>,IComparable<object>
    {
        private string _name;
        private int _age;

        public int Age{get{return _age;}}

        public Employee(string _name, int _age)
        {
            this._name = _name;
            this._age = _age;
        }

        #region IComparable<Employee> Members

        public int CompareTo(Employee other)
        {
            if (this._age < other._age)
                return -1;
            else
                return 1;
        }

        #endregion

        public bool Equals(Employee other)
        {
            if (this._age == other._age)
                return true;
            else
                return false;
        }

        #region IEquatable<object> Members

        public override bool Equals(object other)
        {
            int _age=0;
            if (int.TryParse(other.ToString(), out _age))
            {
                return this._age == _age ? true : false;
            }
            else return false;
        }

        #endregion

        #region IComparable<object> Members

        public int CompareTo(object other)
        {
            int _age = 0;
            if (int.TryParse(other.ToString(), out _age))
                return this._age < _age ? -1 : 1;
            else return 1;
        }

        #endregion

    }

This is the main method which adds a few objects to the tree collection.

static void Main(string[] args)
{
    BinaryTree<Employee> _btree = new BinaryTree<Employee>(new Employee("RootValue",40));
    _btree.AddNode(new Employee("Test1", 34));
    _btree.AddNode(new Employee("Test2", 26));
    _btree.AddNode(new Employee("Test3", 65));
}

Here is the code for the game as a silverlight application. The complete project can be downloaded here.

The XAML is quite simple. Just a grid with three rows and three columns

    <grid x:Name="gridGame" Background="White" Width="300" Height="300" HorizontalAlignment="Center" VerticalAlignment="Center">
        </grid><grid .RowDefinitions>
            <rowdefinition Height="100" />
            <rowdefinition Height="100" />
            <rowdefinition Height="100" />
        </grid>
        <grid .ColumnDefinitions>
            <columndefinition Width="100" />
            <columndefinition Width="100" />
            <columndefinition Width="100" />
        </grid>

It is then filled with the buttons in the Code Behind. I have hardcoded the grid size of three but this will work with any size.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

 public MainPage()
        {
            InitializeComponent();
            InitializeGameLogic(true);
        }
 
        private void InitializeGameLogic(bool _fromConstructor)
        {
            if (!_fromConstructor)
            {
                gridGame.Children.Clear();
            }
            _logic = new TicTacLogic();
            for (int i = 0; i < 3; i++)
                for (int j = 0; j < 3; j++)
                {
                    Button _btnGrid = new Button();
                    _btnGrid.Name = "btn" + i.ToString() + j.ToString();
                    gridGame.Children.Add(_btnGrid);
                    Grid.SetRow(_btnGrid, j);
                    Grid.SetColumn(_btnGrid, i);
                    _btnGrid.Height = 80;
                    _btnGrid.Width = 80;
                    _btnGrid.Click += new RoutedEventHandler(_btnGrid_Click);
                }
            _logic.StartGame();
        }

All the dynamically generated buttons have a single event handler. The idea is to identify the button which sent the event from the name. Once we have the position of the button, the content has to be changed with an X. After the user marks, the computer does its stuff and makes its move. After every marking, a check is done if someone has won or the board is full.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

void _btnGrid_Click(object sender, RoutedEventArgs e)
        {
            Button _userClicked = sender as Button;
            int _xPos = Convert.ToInt32(_userClicked.Name.Substring(3,1));
            int _yPos = Convert.ToInt32(_userClicked.Name.Substring(4,1));
            _userClicked.Content = GetMarking("X");
            _logic.ClickUser(_xPos, _yPos);
            if (!(_logic.CheckIfSomeoneWon() || _logic.CheckIfBoardFull()))
            {
                int _xCom = 0;
                int _yCom = 0;
                _logic.MoveComputer(ref _xCom, ref _yCom);
                foreach (UIElement _tempElem in gridGame.Children)
                {
                    Button _comBtn = _tempElem as Button;
                    if (_comBtn != null && _comBtn.Name == "btn" + _xCom + _yCom)
                    {
                        _comBtn.Content = GetMarking("O");
                        break;
                    }
                }
                if (_logic.CheckIfSomeoneWon() || _logic.CheckIfBoardFull())
                    InitializeGameLogic(false);
            }
            else
                InitializeGameLogic(false);
 
        }
 
        private TextBlock GetMarking(string _mark)
        {
            TextBlock _t = new TextBlock();
            _t.Text = _mark;
            _t.FontSize = 25;
            _t.FontWeight = FontWeights.Bold;
            return _t;
        }

The Board class maintains two important things. The position of all the squares and which player has it marked. The Board doesnt depend on the Player objects. It has an object of the WinningPositions object which calculates the winning positions which can be compared with later. This object makes it possible to play with larger grids.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88

public class Board
{
    readonly int MAXSIZE;
    List<positionmarking> _boardState = new List</positionmarking><positionmarking>();
    List<position []> _allWinningPositions;
 
    public List<positionmarking> BoardState { get { return _boardState; } }
    public List<position []> AllWinningPositions { get { return _allWinningPositions; } }
    public int MaxSize { get { return MAXSIZE; } }
 
    public Board()
    {
        this.MAXSIZE = 3;
        Initialize();
    }
 
    public Board(int _maxSize)
    {
        MAXSIZE = _maxSize;
        Initialize();
    }
 
    private void Initialize()
    {
        _allWinningPositions = (new WinningPositions(MAXSIZE)).AllWinningPositions;
        for(int _xPos=0;_xPos<maxsize ;_xPos++)
            for (int _yPos = 0; _yPos < MAXSIZE; _yPos++)
            {
                _boardState.Add(new PositionMarking(_xPos,_yPos));
            }
    }
 
    internal void MarkPosition(int _xUser, int _yUser, bool _userMarked)
    {
        foreach (PositionMarking _tempPos in _boardState)
        {
            if (_tempPos.BoardPosition.X == _xUser && _tempPos.BoardPosition.Y == _yUser)
            {
                _tempPos.MarkPosition(_userMarked);
                break;
            }
        }
#if DEBUG
        PrintBoardState();
#endif
    }
 
    internal bool? MarkedByUser(Position _tempPos)
    {
        foreach (PositionMarking _tempMark in _boardState)
        {
            if (_tempPos.X == _tempMark.BoardPosition.X && _tempPos.Y == _tempMark.BoardPosition.Y)
                return _tempMark.IsUserMarked;
 
        }
        throw new Exception("Position Not found on Board");
    }
 
    #if DEBUG
    internal void PrintBoardState()
    {
        string _userMarked = "";
        string _compMarked = "";
        foreach (PositionMarking _tempMark in _boardState)
        {
 
            if (_tempMark.IsUserMarked == true)
                _userMarked += "("+ _tempMark.BoardPosition.X + "," + _tempMark.BoardPosition.Y + ") ";
            else if (_tempMark.IsUserMarked == false)
                _compMarked += "(" + _tempMark.BoardPosition.X + "," + _tempMark.BoardPosition.Y + ") ";
        }
        Debug.WriteLine("USER: " + _userMarked);
        Debug.WriteLine("COMP: " + _compMarked);
        Debug.WriteLine("________________________________________");
    }
#endif
    internal bool IsBoardFull()
    {
        foreach (PositionMarking _tempMark in _boardState)
        {
            if (_tempMark.IsUserMarked == null)
                return false;
 
        }
        MessageBox.Show("Board Full");
        return true;
    }
}

The Winning Position class is given below and all positions are calculated at initialization. First the horizontal ones are considered and then the vertical ones. The remaining diagonal positons are taken care later. Each winning position is stored in an array of the position object which just holds the value of the Squares. These arrays are stored in a generic List.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54

public class WinningPositions
{
    List<position []> _allWinningPositions = new List</position><position []>();
    Position[] _winPos1;
    Position[] _winPos2;
 
    public List</position><position []> AllWinningPositions
    {
        get { return _allWinningPositions; }
    }
 
    public WinningPositions(int _maxSize)
    {
        _winPos1 = new Position[_maxSize];
        _winPos2 = new Position[_maxSize];
        PopulateWinningPositions();
    }
 
    private void PopulateWinningPositions()
    {
        int _iCounter=0;
        int _xPos = 0;
        int _yPos = 0;
        for (_xPos = 0; _xPos < _winPos2.Length; _xPos++)
        {
            for (_yPos = 0; _yPos < _winPos2.Length; _yPos++)
            {
                _winPos1[_iCounter] = new Position(_xPos, _yPos);
                _winPos2[_iCounter] = new Position(_yPos, _xPos);
                _iCounter++;
            }
            _allWinningPositions.Add(_winPos1);
            _allWinningPositions.Add(_winPos2);
            _winPos1 = new Position[_winPos1.Length];
            _winPos2 = new Position[_winPos2.Length];
            _iCounter = 0;
        }
        _iCounter = 0;
        for (_xPos = 0, _yPos = 0;_xPos < _winPos1.Length; _xPos++, _yPos++)
        {
            _winPos1[_iCounter] = new Position(_xPos, _yPos);
 
            _iCounter++;
        }
        _allWinningPositions.Add(_winPos1);
        _iCounter = 0;
        for (_xPos = 0, _yPos = _winPos1.Length - 1; _xPos < _winPos1.Length; _xPos++, _yPos--)
        {
            _winPos2[_iCounter] = new Position(_xPos, _yPos);
            _iCounter++;
        }
        _allWinningPositions.Add(_winPos2);
    }
}

Then there is the Player class of which we create two objects. The user takes care of his own moving logic, all we need to do is to change the state of the square for him. But the computer has its method called the SmartMove. Here we first move through the winning positions to identify any open squares that can be marked. If there are no squares, then we look to block the opponent's winning. If even this opportunity is not found, we mark an available open square.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83

public class Player
{
    Board _currentBoardPosition;
    bool _isComputer;
    List<position> _allMarkedPositions;
 
    public Player(bool _isComputer,Board _currentBoardPos)
    {
        this._isComputer = _isComputer;
        _allMarkedPositions = new List</position><position>();
        _currentBoardPosition = _currentBoardPos;
    }
 
    internal void SmartMove(ref int _xCom, ref int _yCom)
    {
        bool _markingDone = false;
        bool _blockUser = false;
        Position _userBlock = new Position();
        foreach (Position[] _winPos in _currentBoardPosition.AllWinningPositions)
        {
            int _userMarked = 0;
            int _computerMarked = 0;
            int _nullMarked=0;
 
            Position _openPos= new Position();
            foreach (Position _tempPos in _winPos)
            {
                bool? isUserMarked = _currentBoardPosition.MarkedByUser(_tempPos);
                if (isUserMarked==true)
                    _userMarked++;
                else if (isUserMarked == false)
                    _computerMarked++;
                else
                {
                    _openPos = _tempPos;
                    _nullMarked++;
                }
 
            }
            if (_computerMarked == 2 && _nullMarked == 1)
            {
                Debug.WriteLine("Wining block: " + _openPos.X + "," + _openPos.Y);
                _currentBoardPosition.MarkPosition(_openPos.X, _openPos.Y, false);
                _xCom = _openPos.X;
                _yCom = _openPos.Y;
                _markingDone = true;
                break;
            }
            else if (_userMarked == 2 && _nullMarked == 1)
            {
                _userBlock = _openPos;
                _blockUser = true;
            }
 
        }
        if (_blockUser && !_markingDone)
        {
            _currentBoardPosition.MarkPosition(_userBlock.X, _userBlock.Y, false);
            _xCom = _userBlock.X;
            _yCom = _userBlock.Y;
            _markingDone = true;
        }
 
        if (!_markingDone)
        {
 
            //No Winning Position Found or not able to block;
            foreach (PositionMarking _posMark in _currentBoardPosition.BoardState)
            {
                if (_posMark.IsUserMarked == null)
                {
                    _markingDone = true;
                    _currentBoardPosition.MarkPosition(_posMark.BoardPosition.X, _posMark.BoardPosition.Y,false);
                    _xCom = _posMark.BoardPosition.X;
                    _yCom = _posMark.BoardPosition.Y;
                    break;
                }
            }
 
        }
    }
}
</position>

Since the Board needs to maintain the state of each and every position on itself, we have another object for it, which maintains a position object and a nullable boolean. If the boolean is true, then the square is marked by the user, if false - then by the computer. If its still null - its open and can be marked. The marking is done by the object's method which makes a check to make sure no players marking is overwritten by another.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

public class PositionMarking
{
    Position _boardPos;
    bool? _isUserMarked;
 
    public Position BoardPosition { get { return _boardPos; } }
    public bool? IsUserMarked { get { return _isUserMarked; } }
 
    public PositionMarking(int _xPos,int _yPos)
    {
        _isUserMarked = null;
        _boardPos = new Position(_xPos, _yPos);
    }
 
    public void MarkPosition(bool _markedByUser)
    {
        if(_isUserMarked == null)
        _isUserMarked = _markedByUser;
        else
        throw new Exception("Position already taken");
    }
}

Then there is the rest of the code including our business logic layer.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93

public class TicTacLogic
{
    bool _userToMove;
    bool _resultHasCome;
    Player _user;
    Player _computer;
    Board _gameBoard;
 
    public TicTacLogic()
    {
        _userToMove = true;
        _resultHasCome = false;
    }
 
    internal void StartGame()
    {
        Initialize();
    }
 
    internal bool ClickUser(int _xUser, int _yUser)
    {
        return _gameBoard.MarkPosition(_xUser, _yUser, true);
    }
 
    internal void MoveComputer(ref int _xCom, ref int _yCom)
    {
        _computer.SmartMove(ref _xCom, ref _yCom);
    }
 
    internal void Initialize()
    {
        _gameBoard = new Board(3);
        _user = new Player(false,_gameBoard);
        _computer = new Player(true,_gameBoard);
 
    }
 
    public bool CheckIfSomeoneWon()
    {
        foreach (Position[] _tempWin in _gameBoard.AllWinningPositions)
        {
            int _userWon=0;
            int _compWon=0;
            foreach (Position _p in _tempWin)
            {
                bool? isUserMarked = _gameBoard.MarkedByUser(_p);
                if (isUserMarked == null)
                    break;
                else if (isUserMarked == true)
                    _userWon++;
                else
                    _compWon++;
 
            }
            if (_userWon == _gameBoard.MaxSize)
            {
                MessageBox.Show("User Won");
                return true;
            }
            else if (_compWon == _gameBoard.MaxSize)
            {
                MessageBox.Show("Computer Won");
                return true;
            }
 
        }
        return false;
    }
 
    internal bool CheckIfBoardFull()
    {
        return _gameBoard.IsBoardFull();
    }
}
 
public class Position
{
    int _xPos;
    int _yPos;
 
    public Position()
    {
    }
 
    public int X { get { return _xPos; } }
    public int Y { get { return _yPos; } }
 
    public Position(int xPos, int yPos)
    {
        this._xPos = xPos;
        this._yPos = yPos;
    }
}

]]> http://blog.ganeshzone.net/blog/index.php/2010/01/tic-tac-toe-game-in-silverlight/feed/ 0 http://blog.ganeshzone.net/blog/index.php/2010/01/dependency-injection-for-test-driven-development/ http://blog.ganeshzone.net/blog/index.php/2010/01/dependency-injection-for-test-driven-development/#comments Thu, 28 Jan 2010 19:43:48 +0000 Ganesh Ranganathan http://blog.ganeshzone.net/blog/?p=831 As applications grow more and more complex, they cost more money to develop, test and maintain. Each bug that leaves the development stage, needs the QA’s effort to identify and notify the developer who again spends time to fix it and get it signed off from the QA. Hence its imperative that good [...]]]> As applications grow more and more complex, they cost more money to develop, test and maintain. Each bug that leaves the development stage, needs the QA’s effort to identify and notify the developer who again spends time to fix it and get it signed off from the QA. Hence its imperative that good code is written the first time around. This fundamental need is what gave birth to what we call Test Drive Development (TDD), i.e. where the developer writes an automated test case first, sees it failing and then writes the code for it to verify the test passes. This technique is also called the red-green style of development.

Automated testing is an integral part of Test Driven Development. A developer simply doesn’t have the time to do functional unit testing of his application, an effort which would neither be reproducible, nor comprehensive. Hence developers use Automated testing tools which decrease the testing effort, even if a little bit more code needs to be written.

The biggest problem while automating testing for applications, especially web applications is the tight coupling between layers making it impossible to test one layer without the another. This can create a lot of complications, since you might get false positives and negatives. Imagine a scenario where you write a test case for a business layer method which retrieves some data from the database and processes it. The aim here is to only test the business layer’s functionality, but if there is a concrete dependency on the data layer, then the result of the test would inevitably depend on the retrieval of data. This is where you could get false positives or negatives.

Hence the reasoning that the test cases should be limited to only the functionality being tested. Consider the below code.

    //Your Business Class
    class BusinessObjects
    {
        //This is a concrete dependency on the dataLayer Object
        private DataObjects _dataLayerObj;
 
        public BusinessObjects()
        {
            //Initializing the DataLayer object
            _dataLayerObj = new DataObjects();
        }
 
        //Get the customer's Name in upper case
        public string GetCustomerNameInUpperCase(int _customerID)
        {
            return (string)_dataLayerObj.RunScalarQuery("Your Select Query here").ToUpper();
        }
    }

So If you write a test on the GetCustomerNameInUpperCase method, it would need to have the database up and running in order to pass. To avoid this, a technique called dependency injection is used in which the application’s layers are loosely coupled with each other enabling us to test each layer’s functionality independent of each other.

This is accomplished by introducing an interface from which the Data layer would inherit and do its work. The business layer would be aware only of these interfaces. Business layer would no longer care about the implementation and thus would allow us to define our own implementation of the interface. This implementation can be “injected” through an additional constructor. Lets see the code for the data layer now.

    //This is the Interface on which the business layer
    //would depend and not the data layer class itself
    interface IData
    {
        public object RunScalarQuery(string _selectQuery);
    }
 
    //Other code is oblivious to the implementation behind
    //the data layer as long as it returns the name
    class DataObjects: IData
    {
        //Implementation of the interface
        public object RunScalarQuery(string _selectQuery)
        {
            //Do all the database related stuff
            return "Some Customer's Name";
        }
    }

Now we would need to modify the business layer to make it depend on the interface.

   class BusinessObjects
    {
        //It now depends on the interface
        private IData _dataLayerObj;
 
        public BusinessObjects()
        {
            //Initializing the DataLayer object
            _dataLayerObj = new DataObjects();
        }
 
        //This is the additional constructor that is used
        //to inject a custom implementation
        public BusinessObjects(IData _dataObject)
        {
            this._dataLayerObj = _dataObject;
        }
 
        //Get the customer's Name in upper case
        public string GetCustomerNameInUpperCase(int _customerID)
        {
            return ((string)_dataLayerObj.RunScalarQuery("Your Select Query here")).ToUpper();
        }
    }

Now lets write the test case for this method. Instead of using the default constructor making the Business layer go to the database for the customer’s name, our very own Mock Class does the trick by supplying a dummy value.

    [TestFixture]
    public class BusinessLayerTests
    {
        BusinessObjects _bObjects;
 
        [SetUp]
        public void Init()
        {
            //Pass our own implementation of IDATA using
            //the additional constructor we defined
            _bObjects = new BusinessObjects(new MockClass());
        }
 
        [Test]
        public void CheckIfUserNameIsUpper()
        {
            //User regular Expressions to make sure all upper cases are returned
            string _returnedName = _bObjects.GetCustomerNameInUpperCase(2342);
            Console.WriteLine(_returnedName);
            Assert.IsTrue(Regex.IsMatch(_returnedName, "^[A-Z]+$"));
        }
    }
 
    //This inherirs from IData
    public class MockClass : IData
    {
        #region IData Members
 
        public object RunScalarQuery(string _selectQuery)
        {
            //Simply return a lower case name
            return "lowercaseName";
        }
 
        #endregion
    }
 
}

This does seem like too much work to test such a simple functionality, but believe me, the benefits grow exponentially with the size of the project. As you start following this approach, it would be easier to test as you go, making the chances of missing out on that tiny bug much lesser. And at the end of the day, nothing makes the developer more happy than seeing all his unit tests pass with the large green NUnit bar.

• In your Hosting provider’s IIS management panel, create a sub directory and configure it as an appplication- root, so that it is recognized as an asp.net application.

• Set your MVC dll’s Copy To Local property to true, so that they are copied to the bin directory of the project.

• Edit your web.config file and update the debug property to false. This holds true for all web applications.

• Update the connection strings to your hosting provider’s connection strings.

Thats it. Your MVC application should run fine. Mine does.

MVC seeks to give developers an alternative, where they have complete control over the application including the viewstate of the controls. My first look was at ASP.NET MVC was a training session given by an architect from Microsoft, and I absolutely didn’t like the complexity it brought to a simple demo application that was showed. I found that we were writing too much code, again bringing back spaghetti code that died with ASP. But over time, after digging into MVC, I found that it made a lot of sense. If my need is just to display a set of data, why should I sacrifice performance over all of the gridview’s bulky viewstate. Rather, it makes sense to use a HTML table and apply CSS to style it.  That is exactly the flexibility that MVC gives you, instead of relying on built in abstraction done by Webform architecture, you get to control every detail of your application.

WebForms

ASP.NET Webforms are radically different from any other server side scripting technology. For one, its event driven. Just like Windows programming, you assign an event handler to an element, like a button and it calls the handler when the button is clicked. No longer, do you have to keep on submitting the form to handle click events and write bulky code to do simple stuff. ASP.NET made the server side code clean – in fact a helluva lot cleaner than ASP.

But all this ignores the age old truth about the web/HTTP – that it is stateless. The server forgets about the page the minute its sent to the client and thats the end of it. So how is context maintained in WebForms? The answer is Viewstate. All of the control states are Base64 encoded and recreated on the server again where the request is executed. There is a lot of built-in plumbing to achieve this level of abstraction.

The main problem with webforms is that though it cleans up your code and markup and seperates them neatly in your Visual studio solution, it horribly clutters the rendered markup to the browser. Every control takes up more and more viewstate and the nested control names which are really long, make the markup bulky, and since this does a round trip on every request, it starts impacting performance and eventually scalability.

One more concern is the Testability, due to the event driven approach, the GUI is tighly coupled with your business layer and there isnt any effective way to test your web page. Automating such testing could be an tedious effort and you would most probably end up doing manual testing.

Despite all of these problems, Webforms are not a bad architecture. It is extremely mature and extensible. A lot of good documentation exists for it and it still makes sense for quite a few scenarios.

MVC

With MVC, it was love at first byte. In fact I never realized how inefficient webforms were, until i worked with MVC. Instead of a page receiving the request, its the Controller which lives at the heart of MVC. Once it gets the request, it decides which view to render and may use models.

One great thing about MVC is the clean Seperation of Concerns (SoC ) which you are able to achieve out of the box. Both the Controllers and the Views can be tested separately using testing tools like NUnit or MBUnit. So that means no longer. Manual testing doesn’t really make any sense for most scenarios, because the effort is not repeatable and when you want to try with a different input, you would need to start all over again.

One major point that is oft repeated is that we shouldn’t jump the gun to convert every existing webforms applications into MVC just because its the in thing. You would have to rewrite your application completely, and the benefit derived would often not be worth the effort. But I would certainly  recommend that  you use MVC for any new applications you create. Its certainly the smarter way.

]]> http://blog.ganeshzone.net/blog/index.php/2010/01/asp-net-evolution-webforms-vs-mvc/feed/ 3 http://blog.ganeshzone.net/blog/index.php/2010/01/thread-synchronization-techniques-part-2-the-readerwriterlock-class/ http://blog.ganeshzone.net/blog/index.php/2010/01/thread-synchronization-techniques-part-2-the-readerwriterlock-class/#comments Sun, 03 Jan 2010 19:22:32 +0000 Ganesh Ranganathan http://blog.ganeshzone.net/blog/?p=754 In continuation to my earlier post about multithreading in .NET, I am writing about the ReaderWriterLock class which is one more method of thread synchronization.

The ReaderWriterLock is based on the fact that the Monitor locking (lock keyword in C#) doesnt really make any distinction whether the thread accessing the variable is [...]]]> In continuation to my earlier post about multithreading in .NET, I am writing about the ReaderWriterLock class which is one more method of thread synchronization.

The ReaderWriterLock is based on the fact that the Monitor locking (lock keyword in C#) doesnt really make any distinction whether the thread accessing the variable is reading it or writing to it. If its only reading it, then we should not need an exclusive lock. So If we allow multiple threads to read the value, it would be much faster than having mutually exclusive locks, which means a more scalable application? On paper, yes – but not in reality.

The problem with ReaderWriterLock is with its implementation. Several experts have slammed this technique and found that outside of limited scenarios, it is actually far slower than the Monitor.Enter method used to get an exclusive lock. ReaderWriterLock gives higher priority to reader threads then writers. This makes sense if you have many readers and only a few writers. So a lot of readers are able to read the resource while the writer has to wait longer to get the lock. But what If you have equal or more writers. The process of favoring readers make writer threads queued up and take a very long time to complete.

Jeffrey Richter says that this performance loss is due to support for Recursion in the ReaderWriterLock class. Due to this, the class needs to maintain a record of the number of times each thread acquires the lock and increment and decrement the counter. When multiple reader threads acquire the same lock (remember ReaderWriterLock class allows simultaneous reads), a counter is maintained for each thread. This overhead is what causes the ReaderWriterLock to pale in comparison to the Monitor class. It is approximately 6 times slower.

Here is a code sample. The Threads 1 and 3 are reader threads while the Thread 2 is a writer one. When you take a look at the output, its clear that the readers are granted simultaneous access to the _value variable, but when the writer thread is writing to it, all the readers wait in queue patiently for it to finish executing. I have made the DoWorkRead method parametrized in order to identify the thread on which the method is being executed.

class Program
{
    int _value = 0;
    ReaderWriterLock _rwLock = new ReaderWriterLock();
 
    static void Main(string[] args)
    {
        Program _prgm = new Program();
        for (int i = 0; i < 20; i++)
        {
            Thread _t1Obj = new Thread(_prgm.DoWorkRead); //Reader Thread
            //Writer Thread
            Thread _t2Obj = new Thread(new ThreadStart(_prgm.DoWorkWrite));
            //Reader Again
            Thread _t3Obj = new Thread(_prgm.DoWorkRead); 
 
            //Start all threads
            _t1Obj.Start("Thread 1");
            _t2Obj.Start();
            _t3Obj.Start("Thread 3");
 
            //Wait for them to finish execution
            _t1Obj.Join();
            _t2Obj.Join();
            _t3Obj.Join();
        }
 
    }
 
    public void DoWorkRead(object threadName)
    {
        //Accquire Reader Lock.
        _rwLock.AcquireReaderLock(Timeout.Infinite);
        Console.WriteLine("Read start: Thread: " + threadName + " " + _value);
        if (threadName.ToString() == "Thread 1")
            //Irregular sleeps makes more chances of
            //Multiple threads trying to access it
            //at same time
            Thread.Sleep(10);
        else
            Thread.Sleep(250);
        Console.WriteLine("Read end  : Thread: " + threadName + " " + _value);
        _rwLock.ReleaseReaderLock();
        //Release Lock
    }
 
    public void DoWorkWrite()
    {
        _rwLock.AcquireWriterLock(Timeout.Infinite);
        Console.WriteLine("\nWriter start: " + _value);
        _value++; //Writing
        Console.WriteLine("Writer End: " + _value);
        _rwLock.ReleaseWriterLock();
        Console.WriteLine();
    }
 
}

A look at the output shows the once the writer acquires a lock, it is mutually exclusive, but both the reader threads are able to access the variable _value.

One more problem with the ReaderWriterLock class is that it allows Reader threads to acquire writer locks. If you set an infinite timeout, it will create a deadlock situation, where the thread just waits to get the Writer lock but cant because the very same thread holds on to the Reader lock and is yet to release it. So the application just waits and waits.

        public void DoWorkRead(object threadName)
        {
            //Accquire Reader Lock.
            _rwLock.AcquireReaderLock(Timeout.Infinite);
            Console.WriteLine("Read start: Thread: " + threadName + " " + _value);
            //NEVER EVER DO THE BELOW. IT WILL CREATE A DEADLOCK
            _rwLock.AcquireWriterLock(Timeout.Infinite);
            Thread.Sleep(10);
            Console.WriteLine("Read start: Thread: " + threadName + " " + _value);
            _rwLock.ReleaseReaderLock();
            //Release Lock
        }

Keeping all these issues in mind, Microsoft introduced a new class – the ReaderWriterLockSlim class from .NET 3.0 onwards, while leaving the old one in there as well, for backwards compatibility. This class placed Lock recursion in the hands of the developers by giving an enum that could be set during object recursion. With recursion out of the way, performance is dramatically better and comparable to the Monitor class. It also took care of the deadlock situation above. As a conclusion, it is always better to avoid the ReaderWriterLock class and use the slim implementation instead. If you are on .NET 2.0, then since that option is not open to you, its better to look at some other synchronization technique.