With effect from Aug 2009, all non nexus related blog posts can now be found at http://jefferytay.wordpress.com

Digging into the source code reveals that the ClientStateField property is not updated each time a tab new tab is selected.

To workaround this, include the following javascript into the head of the document

function ActiveTabChanged(sender, e) {
sender.get_clientStateField().value = sender.saveClientState();
}

and inside the TabContainer set the OnClientActiveTabChanged="ActiveTabChanged"

However during the course of my development in ASP.NET 3.5, there have been several occasions which seems to contradict this behaviour, perhaps it is because he is using WinForms and i’m using ASP.NET. So i took his code, did some modification for it to run several rounds and in IIS and went off from there (The source code can be found at the bottom of the page)

The test procedure is exactly the same as what he did, except since its only 2 i did 2 sets of 20, First doing SortedDictionary then Dictionary (Test result 1 to 20) and the second set doing Dictionary then SortedDictionary (Test result 21 to 40).

The averages of both set do seem to indicate negligible differences in the execution sequence.

The results are extremely interesting. Dictionary outperforms SortedDictionary on all counts! Memory usage, insertion time and search time are all much lower for Dictionary compared to SortedDictionary!

Not really sure why this is so, but if i remember correctly Dictionary uses hashes, so my best guess is it is easier to search for a hash as compared to string comparison..

Time Taken for Search

Time Taken for Insert

Memory Usage for Insert

The generated raw results from the web application

Source Code (.aspx.cs)

using System;
using System.Collections.Generic;
using System.Linq;
using System.Web;
using System.Web.UI;
using System.Web.UI.WebControls;
using System.Diagnostics;
using System.Data;

public partial class test_cittka : System.Web.UI.Page
{
protected void Page_Load(object sender, EventArgs e)
{
DataTable tbl = new DataTable();
tbl.Columns.Add(”Memory_SortedDictionary”);
tbl.Columns.Add(”Memory_Dictionary”);
tbl.Columns.Add(”Insert_SortedDictionary”);
tbl.Columns.Add(”Insert_Dictionary”);
tbl.Columns.Add(”Search_SortedDictionary”);
tbl.Columns.Add(”Search_Dictionary”);

for (int i = 0; i < 20; i++)
{
DataRow row = tbl.NewRow();
sortedDictionary = new SortedDictionary<string, string>();
dictionary = new Dictionary<string, string>();
Insert_Test(”SortedDictionary”, sortedDictionary, row, 0, 2);
Insert_Test(”Dictionary”, dictionary, row, 1,3);
Search_Test(”SortedDictionary”, sortedDictionary, row,4);
Search_Test(”Dictionary”, dictionary,row,5);
tbl.Rows.Add(row);
}
GridView gv = new GridView();
gv.DataSource = tbl;
gv.DataBind();
Page.Form.Controls.Add(gv);
}
private readonly int searchIndex = 88888;
private readonly int numberRepetition = 500000;
private SortedDictionary<string, string> sortedDictionary = new SortedDictionary<string, string>();
private Dictionary<string, string> dictionary = new Dictionary<string, string>();

private void Insert_Test(string name, IDictionary<string, string> dict, DataRow row, int idxMemory, int idxInsert)
{
Response.Write(String.Format(”<br>——–Insert {0}——–”, name));
string[] letters = { “A”, “B”, “C”, “D”, “E”, “F”, “G”, “H”, “I”, “J” };
long memoryStart = System.GC.GetTotalMemory(true);
Stopwatch watch = new Stopwatch();
watch.Start();
Random rand = new Random();
for (int i = 0; i < numberRepetition; i++)
{
string key = GetRandomLetter(letters, rand, i) + “_key” + i;
string value = “value” + i;

dict.Add(key, value);
}
long memoryEnd = System.GC.GetTotalMemory(true);
watch.Stop();
Response.Write(String.Format(”<br>Memory Allocated by {0} is: {1}bytes”,
name, memoryEnd – memoryStart));
PrintResults(watch);
row[idxMemory] = (memoryEnd – memoryStart).ToString();
row[idxInsert] = watch.ElapsedTicks.ToString();
}
private void Search_Test(string name, IDictionary<string, string> dict, DataRow row, int idx)
{
Stopwatch watch = new Stopwatch();
Response.Write(String.Format(”<br>——–Search {0}——–”, name));
watch.Start();
Response.Write(String.Format(”<br>Found:{0}”, dict["A_key" + searchIndex]));
watch.Stop();
PrintResults(watch);
row[idx] = watch.ElapsedTicks.ToString();
}
private void PrintResults(Stopwatch watch)
{
Response.Write(String.Format(”<br>Elapsed: {0}”, watch.Elapsed));
Response.Write(String.Format(”<br>In milliseconds: {0}”, watch.ElapsedMilliseconds));
Response.Write(String.Format(”<br>In timer ticks: {0}”, watch.ElapsedTicks));
}
private string GetRandomLetter(string[] letters, Random rand, int i)
{
if (i == searchIndex)
{
return “A”;
}
return letters[rand.Next(0, 10)];
}
}

What is a web garden

Simply put, a web garden is a single web server running multiple IIS processes. The good thing about this is, there are multiple process which can service your http/https calls. So if a client is holding up one process, the other processes can still continue. Also if a particular process were to hang and restart itself, your website will still continue to serve clients as the other processes are running.

Read more about Microsoft’s explanation of web garden here

The article clearly states that

all processes have their own copy of application state, in-process session state, caches, and static data

Let us examine each one of these and see what we can do to workaround it

Application state

Problem Statement

As each process has their own application state, whatever you change in one process will not be updated in another process. As this exists purely for backward compatibility with classic ASP applicationstate, Microsoft does not recommend using this alot, especially for read/write operations.

Workaround

Use Cache instead of ApplicationState if you can.
Cached Items

Problem Statement

Be it HttpRunTime.Cache or Enterprise Library Cache, the cache will not reflect latest updates inside a web garden. This is because only the worker process which updated the Cache Item will see the updated values, the other worker processes will not be notified that a change has happened and will continue using the value stored in their in memory cache.

Workaround

There are a few paid solutions, and the only free one so far which is able to address this is can be found at http://enyimmemcached.codeplex.com/Release/ProjectReleases.aspx?ReleaseId=13095 or http://www.codeproject.com/KB/aspnet/memcached_aspnet.asp

Session state

Problem Statement

Using web gardens, you will find that the session values gets rotated around. Again this is due to the fact that different worker process is handling your request, and each worker process has its own session state management.

Workaround

Simply change the inproc to either State Server or SQL Server to store the session state outside of the worker process. Performance will degrade when using either of these 2 options, but at least you can be assured of the persistancy of the data

There are 2 ways to create a custom control

1. Writing your own HTML code
2. Using the controls property

Writing your own HTML code

Using your own properties, you can easily write HTML code to generate the
output

e.g consider a simple scenario where you take in a text and color input and
render the control.

You can simply create a class and extend it from the WebControl base class to
have the functions available to your custom control.
Thereafter, just create the properties you require and use them inside the
Render function.
Do note that you can render different output when in design mode and actual
rendering, all you need is to check the DesignMode property.
If however you need input boxes, do take note that you need to handle the
postback event and put the value back into the inputs manually.

public class
SimpleControl :
WebControl

{

public
SimpleControl()

{

}

[BindableAttribute(true),
Category("UI"),
DefaultValue("")]

public
string Text

{

get

{

object obj =
this.ViewState["Text"];

if (obj ==
null)
return
“”;

else
return obj.ToString();

}

set {
this.ViewState["Text"]
= value; }

}

[BindableAttribute(true),
Category("UI"),
DefaultValue("")]

public
string Color

{

get

{

object obj =
this.ViewState["Color"];

if (obj ==
null)
return
“”;

else
return obj.ToString();

}

set {
this.ViewState["Color"]
= value; }

}

protected
override
void Render(HtmlTextWriter
writer)

{

writer.Write(“<font color=’”
+ Color + “‘>” +
Text + “</font>”);

//base.Render(writer);

}

}

Using the controls property

The good thing about using the controls property is that you do not need to
care about ViewState (it is handled by the individual control), Postback
inputs handling (handled again by the individual control).

The bad thing is

1. Have to write your own custom event handlers
2. Everything needs to be control based, you cannot emit your own html.

public class
LabelExtendedControl
: Label

{

public
SimpleControl()

{

}

[BindableAttribute(true),
Category("UI"),
DefaultValue("")]

public
string xText

{

get

{

object obj =
this.ViewState["Text"];

if (obj ==
null)
return
“”;

else
return obj.ToString();

}

set {
this.ViewState["Text"]
= value; }

}

[BindableAttribute(true),
Category("UI"),
DefaultValue("")]

public
string xColor

{

get

{

object obj =
this.ViewState["Color"];

if (obj ==
null)
return
“”;

else
return obj.ToString();

}

set {
this.ViewState["Color"]
= value; }

}

protected
override
void Render(HtmlTextWriter
writer)

{

this.Text = xText;

this.ForeColor =
System.Drawing.Color.FromKnownColor(xColor);

base.Render(writer);

}

}

Custom Events

To create your own custom events, you need
to write 3 main items

1. A readonly object containg the event
2. A function which will call the Event
3. The actual event handler itself

private
readonly
object EventCreate =
new
object();

protected
virtual
void OnCreate(EventArgs
e)

{

EventHandler
onCreateHandler = (EventHandler)Events[EventCreate];

if
(onCreateHandler != null)
onCreateHandler(this,
e);

}

[Category("Action"),Description("Raised
when a CommandEvent occurs within an item.")]

public
event
EventHandler Create

{

add

{

Events.AddHandler(EventCreate, value);

}

remove

{

Events.RemoveHandler(EventCreate, value);

}

}

Points to ensure

1. add the reference to System.DirectoryServices
2. The account running the website must have the appropriate access to the AD

Get Users in Group

List<UserID_CN>
result = new
List<UserID_CN>();

using (var
de = getEntry())
{
object
obj = de.Properties["member"].Value;

if (obj
is
string)
{

result.Add(new
UserID_CN()
{

CN = obj.ToString(),
ID = Regex.Match(obj.ToString(),
“CN=(.*?),”,
RegexOptions.IgnoreCase).Groups[1].Value

});
}
else
if (obj
is
object[])
{

foreach (object
o in (object[])obj)

{
result.Add(new
UserID_CN()
{

CN = o.ToString(),
ID = Regex.Match(o.ToString(),
“CN=(.*?),”,
RegexOptions.IgnoreCase).Groups[1].Value

});
}
}
else

{
}
}
result = result.OrderBy(p => p.ID).ToList();

return result;
}

Delete Users in Group

using (var
de = getEntry())

{

foreach (UserID_CN
u in GetUsersInGroup)

{

de.Properties["member"].Remove(u.CN);

}

de.CommitChanges();

de.Close();

}

Add Users to Group

de.Properties["member"].Add(userDN);

Get all properties of directory entry

foreach(string
key in
dirEntry.Properties.PropertyNames)

{

// Each property contains a collection
of its own

// that may contain multiple values

Label1.Text += “–”
+ key + “–<br>”;

foreach(
object propVal
in dirEntry.Properties[key]
)

{

if
(propVal.GetType() == typeof(System.Byte[]))

{

Label1.Text += “(”;

for (int
i=((byte[])propVal).GetLowerBound(0);i<((byte[])propVal).GetUpperBound(0)+1;i++)

Label1.Text += (char)
((byte[])propVal)[i];

Label1.Text += “)<br>”;

}

else
if
(propVal.GetType().ToString() ==
“System.__ComObject”)

{

try

{

Label1.Text += “(”;

long val_int64 =
GetInt64(dirEntry, key);

//DateTime dt =
DateTime.FromFileTime(val);

try

{

if (val_int64 >
0)

{

DateTime T =
DateTime.FromFileTime(val_int64);

Label1.Text += T.ToString(“dd MMM yyyy
HH:mm:ss”) + “)<br>”;

}

else

{

throw
new
Exception();

}

}

catch

{

Label1.Text += val_int64 + “)<br>”;

}

}

catch

{

Label1.Text += “(”
+ propVal + “)<br>”;

}

}

else

Label1.Text += “(”
+ propVal + “)<br>”;

}

}

Label1.Text += “—————<br>”;

ASP.NET offers several forms of caching. They are output caching, HttpRuntime.Cache and finally Enterprise Library Cache. Each of these aims to meet the requirements of different scenarios.

This article shall explore the scenarios which each of these best fit.

Output caching

There are 2 variants of output caching, you can either cache the whole page or portions of the page (these portions must be in the form of user controls)

This is commonly used when the steps to reproduce the page is huge (e.g large database calls, long file read times, huge processing requirements), but the output does not change frequently, e.g a FAQ page listing all the FAQs and employing javascript for inline fuzzy search (employing different scripts for different browsers)

When you use output caching, the output is cached for a specific duration e.g 120 seconds and can be set to vary based on different parameters e.g querystring, browser. In the previous scenario of the FAQ page, you can use the VaryByCustom=”browser” parameter to cache and vary the output based on different browsers.

MSDN on Output Caching

HttpRuntime.Cache

HttpRuntime.Cache is used to store serializable objects in a cache for future use. This cache is more powerful compared to output caching, but using it requires more work. This is frequently used to cache the return data in business layers.

If you do not wish to bother about the technicalities behind this cache, you can easily add to it via the following code

HttpRuntime.Cache["key"] = value

This will add the item to the cache with default values for all parameters. Alternatively you can look at the complete  declaration for adding an item to the cache

public Object Add(
string key,
Object value,
CacheDependency dependencies,
DateTime absoluteExpiration,
TimeSpan slidingExpiration,
CacheItemPriority priority,
CacheItemRemovedCallback onRemoveCallback
)

This allows you to add a serializable object to the cache with

• Dependencies: This indicates that this cache item is dependent on another object (usually a file), so when the dependency is changed, the cache object is automatically removed from the cache.
• Absolute Expiration: This indicates that the cache item will expire after a fixed amount of time. Once the object expires, it is removed from the cache
• Sliding Expiration: This tells the cache to extend the cache item a fixed amount of time each time the cache item is accessed. Once this amount of time is reached and the cached item is not accessed, it will expire
• Priority: This tells the cache the priority of the cache item. This comes into effect when the cache is full and items needs to be removed. Items with a lower priorities are removed first.
• CacheItemRemovedCallback: This is a delegate function which is run when the item is removed from the cache, the key, value and reason for removal is sent to this function so that you can do the appropriate processing.

MSDN on HttpRuntime.Cache

Enterprise Library Cache

There are 5 main differences between this and HttpRuntime.Cache. If your needs meet any one of these points, then it is recommended that you use Enterprise Library Cache.

1. The cache is stored independently of the web server, so even if you were to do an iisreset, the cache still remains
2. You can configure the cache to be stored in a persistent backing store e.g SQL Server
3. There can be multiple caches serving different needs in the same web application
4. You need more control over expiration policies
5. You do not need to know when the item expires, but you want to be able to reload the item if required

MSDN on Enterprise Library Caching Block

Summary

1. Output caching: Used to store the page output so that ASP.NET does not need to render the content again
2. HttpRuntime.Cache: Used by your web application to store items in a cache, the items will disappear when the website is reset
3. Enterprise Library Cache: Separate cache mechanism which has its own memory and data store, is not affected by server restarts (if using persistent backing store)

Caveats

1. All of these caches is restricted to application domain scope, i.e you cannot share them across different application domains (e.g in web garden or multi server environment)

Some nice features which are missing include

1. Advanced Search
2. Print
3. Save

I’ve written an advanced gridview class which is able to perform all these

This gridview comes with the ability to

• Display Create button which goes to another url when clicked
• Save button to save all the records in the Gridview (yes all the records)
• Print button to print all the records in the GridView (same as save button, but will pop up a new window to do the print)
• Advanced Search which gives you “google like” search e.g using “+” , “-”, quotation marks to mark words as one key term, OR. e.g aaa bbb “”ccc ddd”" +eee -fff OR ggg

Custom properties defined in this Gridview allow you to customize the top search portion

Click here to download the class and use it in your code

1. Viewstate
   Web sites are stateless by design, which means that anything you have on the page is not retained when you submit the data. What ASP.NET tries to do is to save the state on the page so that state is retained when the page submits its data.

   However this quickly becomes unmanageable when you have too many controls on the page, especially if you have large server controls e.g gridviews. As viewstate is stored on the page itself, a simple page with a few form elements may end up sending large amounts of data on postback, the majority of which is the viewstate itself.

   Other than increasing page size, there is also the performance cost incurred when serializing and deserializing the viewstate.

   Solution:
   

   • if a page is an output page and does not do any further processing, you can disable the viewstate for the page <% Page EnableViewState=”false”.
   • If the control on the page does not handle any events, is not a data-bound control or the data i constantly refreshed on postback, then you can disable the viewstate for the control
   • If all else fails, you can look into storing viewstate on the session state using SessionPageStatePersister (There is a drawback because the session state may expire before the user is done with what he or she is doing on the page – think blogs, wikis etc)
2. String concatenation & comparisonHaving experience with int i = 0; i = 1 +2, you decide to go ahead and do the same for strings. string s = “aa” + “bb”This is very bad programing, simply because strings work differently from numeric primitive data types. When you do this type of string concatenation, quite a fair bit of resources are created to hold the strings and results

   The same can be said for case insensitive string comparison, usually you would do something like string1.toLower() == string2.toLower(). Same thing occurs, temporary space needs to be created to store the lowered strings and then do a compare.

   Solution:

   use stringbuilder for string concatenation functions
   System.Text.StringBuilder sb = new System.Text.StringBuilder();
   sb.Append(”aa”);
   sb.Append(”bb”);
   sb.ToString();

   use the overloaded string compare function if you want to do case insensitive searches
   String.Compare (string strA, string strB, bool ignoreCase);

3. For loops with easily checked conditions perform far better than for each loops. However if the for loop condition requires a property call (e.g list.count()) then it will perform much slower, in this case get the count before the loop and assign it to a variable and in the for loop check against this variable.
4. If you have multiple exit conditions, put them when it is first possible to check and also put the one most likely to happen right on top. This reduces the amount of checks needed to exit, which can give you a performance improvement.