Kristan's Mental Output Dump : ASP.NET, LINQ to SQL, Community Server, Software.

Parallel.ForEach operations and the CancellationToken

2011-11-10T12:02:00.001Z

If you use the CancellationToken to enable cancellation during the parallel iteration of an enumerable, then you should note that each operation is given a chance to finish. The OperationCanceledException is only thrown after all the threads have completed their unit of work.

You could monitor the cancellation token inside the units of work as well if you liked.

Windows Azure DevFabric forcing port 80 rather port 81, 82 etc

2011-11-10T09:41:00.001Z

Hi there, it's useful to run a Windows Azure app in the dev fabric on port 80 sometimes. However, you need to be absolutely sure that port 80 is not in use beforehand.
Use TCPView from sysinternals and if you find that an item appears under "local port" called "http", then the port is in use. You cannot easily find out what process that is as it's a service, however in my case, once I'd shut down "Web Deployment Agent Service" by doing Run -> CMD -> net stop MsDepSvc this solved the issue and Azure was able to run the web role under port 80.

Also, ensure that Skype doesn't occupy port 80 too.

Distributed Lock

2011-01-25T15:22:00.001Z

I’m working on a distributed system on Azure which has a Web and Worker Roles. Plus, there’s the possibility of N number of instances per role. The Worker role is a batch processor which consists of Jobs and the Job Scheduler. My requirement is that we should only have one instance of the Job Scheduler throughout the whole system.

If this was a simple multi-threaded application with one app domain, this would be relatively simple using a mutex and the lock statement, however, given that the application could be spread across multiple virtual machines within the Azure Fabric I had to come up with another way.

I solved this problem by using a SQL Azure table, serializable transactions and a Distributed Lock Service in C#.

First create the table

/****** Object:  Table [dbo].[Lock]    Script Date: 01/25/2011 14:18:13 ******/
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE TABLE [dbo].[Lock](
    [id] [int] IDENTITY(1,1) NOT NULL,
    [LockName] [nvarchar](4000) NOT NULL,
    [AcquirerGuid] [uniqueidentifier] NULL,
    [AcquiredDate] [datetime] NULL,
 CONSTRAINT [PK_Lock] PRIMARY KEY CLUSTERED 
(
    [id] ASC
)WITH (STATISTICS_NORECOMPUTE  = OFF, IGNORE_DUP_KEY = OFF)
)
GO

Then create the Stored Procedures, “Lock_Acquire” and “Lock_Release”


IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Lock_Acquire]') AND type in (N'P', N'PC'))
    DROP PROCEDURE [dbo].[Lock_Acquire]
GO

SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO

CREATE PROCEDURE [dbo].[Lock_Acquire]
(
    @Name NVARCHAR(4000),
    @AcquirerGuid UNIQUEIDENTIFIER
)
AS
BEGIN

    SET TRANSACTION ISOLATION LEVEL SERIALIZABLE
    
    BEGIN TRANSACTION

        IF(NOT EXISTS(SELECT 1 FROM Lock WHERE LockName = @Name))
        BEGIN    
            INSERT INTO [Lock] (LockName, AcquiredDate, AcquirerGuid)
            VALUES (@Name, GETUTCDATE(), @AcquirerGuid)
        END
        ELSE
        BEGIN
            
            UPDATE L
               SET L.AcquiredDate = GETUTCDATE(),
                   L.AcquirerGuid = @AcquirerGuid
              FROM Lock L
             WHERE L.LockName = @Name 
               AND (L.AcquirerGuid = @AcquirerGuid -- owned by the current requestor
                        OR DATEDIFF(SS, L.AcquiredDate, GETUTCDATE()) > 30) -- or expired so re-assign to new requestor
        END
        
        SELECT @@ROWCOUNT; 
         
    COMMIT TRANSACTION
END

GO

IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[Lock_Release]') AND type in (N'P', N'PC'))
    DROP PROCEDURE [dbo].[Lock_Release]
GO

SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO

CREATE PROCEDURE [dbo].[Lock_Release]
(
    @Name NVARCHAR(4000),
    @AcquirerGuid UNIQUEIDENTIFIER
)
AS 
BEGIN
    
    SET TRANSACTION ISOLATION LEVEL SERIALIZABLE
    
    UPDATE Lock
    SET AcquiredDate = NULL, 
    AcquirerGuid = NULL
    WHERE AcquirerGuid = @AcquirerGuid
    
    SELECT @@ROWCOUNT;
    
END

GO

Import the Stored Procedures into Entity Framework (if you use EF)

Create the Distributed Lock Service

public class DistributedLockService : IDisposable
    {
        public Guid AcquirerId { get; set; }
        public string LockName { get; set; }
        public bool LockAcquired { get; set; }

        /// <summary>
        /// ctor
        /// </summary>
        /// <param name="lockName"></param>
        public DistributedLockService(string lockName)
        {
            LockName = lockName;
            AcquirerId = Guid.NewGuid();
        }

        /// <summary>
        /// Attempts to acquire the named lock
        /// </summary>
        /// <returns></returns>
        public bool Acquire()
        {
            LockAcquired = Facade.DataContext.Lock_Acquire(LockName, AcquirerId).FirstOrDefault() == 1;
            return LockAcquired;
        }

        /// <summary>
        /// Attempts to release the named lock
        /// </summary>
        /// <returns></returns>
        public bool Release()
        {
            if (LockAcquired) return Facade.DataContext.Lock_Release(LockName, AcquirerId) == 1;
            else return true;
        }


        #region IDisposable Members

        public void Dispose()
        {
            Release();
        }

        #endregion
    }

You may have to change Facade.DataContext to be your equivalent ORM or Sql Helper class.

Create the Unit Test

[TestClass]
    public class DistributedLockServiceTest
    {
        [TestMethod]
        public void CheckThatLockingWorks()
        {
            string fakeLockName = Guid.NewGuid().ToString();
            
            List<bool> results = new List<bool>();
            ManualResetEvent handle = new ManualResetEvent(false);
            object mutex = new object();

            Parallel.Invoke(() =>
            {
                Parallel.For(0, 100, x =>
                {
                    handle.WaitOne();
                    DistributedLockService svc = new DistributedLockService(fakeLockName);

                    lock(mutex) results.Add(svc.Acquire());
                });
            }, 
            () =>
            {

                Thread.Sleep(2000);
                handle.Set();
            });
            Assert.AreEqual(1, results.Where(x => x == true).Count(), "The number of trues should be 1 for fake lock: " + fakeLockName);

        }
    }

Use it for real…

private DistributedLockService _lockService = new DistributedLockService("JobScheduler");

if(_lockService.Acquire()) // do something, like start a Job

So the idea is that you can instantiate the Distributed Lock Service and keep it for as long as necessary, then whenever you wish to perform some action without the possibility of concurrency, do it after acquiring the lock. Please also note that the lock will last for up to 30 seconds at a time, so you must call Acquire within the 30 second period if you wish to retain non-concurrency. Otherwise the lock will be relinquished to any other requestor. If this is too short a timeframe, then you can change it inside the Lock_Acquire stored procedure.

The reason why there’s a time limit is in case the acquiring/lock-owning thread throws an exception and does not release the lock.

Please let me know should you find any problems. thanks.

Dell XPS M1330, nVidia GeForce 8400 GS overheats and the laptop cuts out

2010-09-13T15:24:00.001Z

I have had this laptop this January 2009 and it’s been fine, but recently it’s been cutting out if I use Google Earth or Bing Maps! I have measured the temperature at which it cuts out is about 110 degrees. When it turns on again, it reports either error #M1001 or #M1004. This is apparently and overheating problem with the nVidia GPU.

I am contacting Dell about this, however my motherboard was only replaced last month when my graphics card failed with what I can only describe as “on screen display corruption issues”. Weird symbols and colours appeared. Oh well, looks like I’ll need another repair, or just have to buy a new laptop as I’m a freelancer, so the loss of earnings will eventually become greater than the cost of a laptop, so I may as well get a new one… hopefully from a more reliable brand.

UPDATE: 15th September 2010: A Dell engineer changed the heat sink and now it runs perfectly!

ASP.NET MVC 2, “Request Validation has detected a potentially dangerous client input value, and processing of the request has been aborted” and the ValidateInput does not work

2010-09-13T09:03:00.001Z

If you have problems getting the [ValidateInput(false)] attribute to work in a Visual Studio 2010 ASP.NET MVC 2 solution; please remember to put requestValidationMode="2.0" on the httpRuntime element of the web.config.

Then everything will be fine. Enjoy.

Umbraco 4.x – Warning on upgrading document types

2010-08-23T14:20:00.001Z

Please be aware, though Umbraco is a great Content Management System, once you have deployed an application based on it – it’s very difficult to upgrade items in the meta-database afterwards. Umbraco Courier does a good job of transferring content, but it doesn’t take into account changes to Document, Media and Member types, nor does it take into account new macros or changes to settings on those entities. Courier will report the problems as you try to migrate content which is dependent on a latest meta database changes.

This is akin to developing your database in the development environment, then, once you’ve deployed the software, having to use the SQL Table Designer GUI make all the necessary schema changes manually.

Dynamic subdomains on Apache and PHP

2010-08-05T09:52:00.001Z

To rewrite subdomain urls internally to subdirectories under the root, ensure mod_rewrite is installed in Apache. Configure .htaccess inside the root directory as follows:-

RewriteEngine On
RewriteCond %{HTTP_HOST} !^www\.mydomain.com
RewriteCond %{HTTP_HOST} ([^.]+)\.mydomain.com
RewriteRule ^(.*)$ /home/sites/mydomain.com/public_html/%1/index.html

Obviously replace “mydomain” and change the path on the last line to where your website physically resides on the server. Use the variable %1 to use the name of the subdomain as the name of the destination subdirectory.

My iPad Review

2010-07-21T16:02:00.001Z

I got an iPad a few weeks ago and thought I’d share with you the pros and cons I have discovered:-

Pros

Great for reading e-books and PDFs
Makes surfing a web really easy as you can pinch, zoom and scroll really easily using just one or two fingers
Watching YouTube and videos is a really tactile experience, and so is viewing Photos.
If you get an iPad you can run both iPhone and iPad apps on it. This is great if you’re a developer and want to target both platforms and not have to purchase an iPhone. For me the cost of getting into iPad and iPhone development has been about £1000 (£500 for the iPad and about £500 for the Mac Mini). Although they have just put up their Mac Mini price. If I’d had to purchase an iPhone as well, the cost would be today at about £1700. I have saved £700!

Cons

The Apps in the App Store don’t have a trial period, so if you don’t like it, then you can’t get a refund. Also, sometimes the description of an App can exaggerate how good it really is
Apps can be a bit too expensive
Re iTunes; I can get the IT Crowd from channel 4 for £10. However, on a Windows PC or a Mac I can go to 4OD on Channel4.com and watch loads of stuff for free.
No support for Flash, so stuff like 4oD doesn’t work :-( Also, the BBC News website experience is degraded as they rely on flash
When I download Spotify I find I cannot play music in the background (although this is an iPhone app, maybe an iPad version will be different)
When I plug in headphones the sound doesn’t seem to come through – not sure whether this is a fault or a configuration problem
The soft-keyboard is a bit annoying (although, a keyboard or anything other than a proper physical one is generally annoying!)

All in all, if I had £500 to spend and I needed a computer, I’d probably get a netbook and pocket £100-£200. Just because a netbook supports everything and there’s an abundance of free software out there and it’s compatible with most things (e.g. Flash), and it has a regular keyboard so you can work and play on it. Sure, it won’t have integrated 3G, but get a bluetooth link to a 3G-enabled mobile phone and you’re sorted.

Steve Jobs said in his keynote speech, introducing the iPad that it occupies the same market space as netbooks. Although for me, it’s just a different type of device really – and very niche.

In summary then… an iPad is

Less functional than a netbook (you can literally do anything a PC can do, albeit slower)
More controlled and regulated by Apple HQ
However a much more tactile, novel and fun experience

My gut feeling is that, one day, a manufacturer will invent a powerful netbook with a detachable screen and all the fun of an iPad. At least I hope this will happen, alongside a de-facto open e-book format.

Get this? Windows Azure MMC, Could not load file or assembly 'Microsoft.Samples.AzureManagementTools.PowerShell, Version=1.1.3779.27370, Culture=neutral, PublicKeyToken=null' or one of its dependencies.

2010-06-30T16:58:00.001Z

You probably have a previous version of the Windows Azure SDK. Delete all but the latest folder from ‘C:\Program Files\Windows Azure SDK’. For me, I had to delete the folder ‘C:\Program Files\Windows Azure SDK\v1.0’

Google copies Bing!

2010-06-10T09:32:00.001Z

One day I woke up to find this atrocity!

And then realised they were trying to do this…

I think maybe Google use Bing.com – which is nice. Well done bing.

Escaping javascript strings in Umbraco XSLT macros

2010-04-27T15:50:00.001Z

How to escape string literals in XSLT which appear inside JavaScript function calls:-

<xsl:variable name="apos">'</xsl:variable>
<xsl:if test="data[@alias='popupMessage'] != ''">
      <xsl:attribute name="onclick">
        <xsl:text>return window.confirm('</xsl:text>
        <xsl:value-of select="umbraco.library:Replace(data[@alias='popupMessage'], $apos, concat('\',$apos))" disable-output-escaping="yes" />
        <xsl:text>');</xsl:text>
      </xsl:attribute>
</xsl:if>

ASP.NET, hMailServer, error “The SMTP server requires a secure connection or the client was not authenticated” or “SMTP authentication is required”

2010-04-25T15:52:00.001Z

If you receive the above error message when sending email via hMailServer’s SMTP Service, just go into the Admin console –> Settings/Advanced/IP Ranges/My Computer and uncheck the boxes below.

Windows Server 2008, IIS 7, Shared Configuration, System Corrupted Irrecoverably

2010-04-25T10:49:00.001Z

The story is, I tried to copy IIS7 configuration from one Windows Server 2008 server to another.

This overwrote the applicationHost.config
IIS/App Pools failed because global modules that were on the previous server were not installed on the current

Attempted to install those modules
that failed.

Thought I’d try to rollback, but it doesn’t backup the original IIS7 config
Uninstalled IIS7

Tried to reinstall IIS7
That failed

Tried to set up Windows Installer tracing with special registry entries

Tracing doesn’t work, no file is output into the %Temp% directory

Now wanting to do a system restore

There’s no system restore on Windows Server 2008

Quite a disproportional dilemma given the cause.

I like Microsoft – sometimes. But really, I have had problems with Windows Server set up before… they haven’t put the effort into making this work. And given that it’s usually professionals who are working setting up boxes, it would be a good idea to give some error messages that mean something.

Message: Type 'umbraco.DataLayer.SqlHelperException' in Assembly 'umbraco.DataLayer, Version=0.3.0.0, Culture=neutral, PublicKeyToken=null' is not marked as serializable.

2010-04-13T07:50:00.001Z

This is most likely because the database is inaccessible to Umbraco. Check the connection strings are correct and also try connecting using the same connection string credentials with SQL Management Studio.

MSMQ "The destination queue does not exist” or “BadDestinationQueue” when sending MSMQ messages or WCF net.msmq bound service calls

2010-03-26T00:06:00.001Z

If you’re sending MSMQ messages using the DNS name of the server, you may get one of the above errors… this is easily fixed by adding a new registry key which tells MSMQ to ignore the hostname validation.

Click Start, click Run, type regedit, and then click OK.
Locate and then click the following key in the registry:
HKEY_LOCAL_MACHINE\Software\Microsoft\MSMQ\Parameters
On the Edit menu, point to New, and then click DWORD Value.
Type IgnoreOSNameValidation, and then press ENTER.
On the Edit menu, click Modify.
Type 1, and then click OK.

See the Microsoft Support Article.

Buy.at product feeds gzip trailer invalid / corrupt.

2010-03-24T20:47:00.001Z

Please note that buy.at product feeds in gzip format are corrupt. There is no CRC32 or ISIZE value in the trailer of the file which is required. Please see the gzip spec. You can see this by opening a valid gzip file and their invalid gzip file in WinRAR or 7-Zip… there’s no input file size or CRC in the file.

Also see http://www.affiliates4u.com/forums/buy/88660-buy-feed-download-issue.html

GZipInputStream, GZipStream, GZipUtil, C#.NET UNIX .gz files - decompression ends prematurely - Workaround

2010-03-24T09:20:00.001Z

I have been trying for days now to decompress a 10MB gzip file using

ICSharpCode.SharpZipLib.GZip.GZipInputStream,
Ionic.Zlib.GZipStream,
and System.IO.Compression.GZipStream

Using both my own code and other people’s code from all over the Internet. I have tried analyzing the file to make sure it’s valid – and it is. The only thing I know about this file is that it’s been compressed by a UNIX based system with the “DEFLATE” method, downloaded over http and can be successfully decompressed by 7Zip, WinZip and WinRAR. But when it comes to decompressing the file within C# with various libraries, it’s just not happening. It always prematurely ends the decompression, so I get a truncated file.

In the end, I decided to wrap a call to gzip.exe which can be downloaded from http://www.gzip.org/ or the direct Windows Download is here.

Put gzip.exe inside your project folder, in Solution Explorer and ensure “Copy to output directory” is set to “Copy always”. The code for calling the exe is below.

/// <summary>
/// Uncompresses a gzip file 
/// </summary>
/// <param name="gzipFilename"></param>
/// <returns></returns>
public static string UngzipFile(string gzipFilename)
{
    const string GZIP_EXE_NAME = "gzip.exe";
    string gzipExePath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, GZIP_EXE_NAME);
    string args = string.Format("-d \"{0}\"", gzipFilename);
    var p = System.Diagnostics.Process.Start(gzipExePath, args);
    p.WaitForExit();
    return gzipFilename.Remove(gzipFilename.Length - 3, 3);
}

UPDATE: ok, the files I’m trying to decompress are corrupt – they have no ISIZE or CRC32 GZip File Trailer (well, they are zero, which can’t be good). I wish there was a way to force it to decompress as the bandwidth usage for uncompressed it massive.

How to horizontally centre align an inner div of unknown width within an outer div of known width :: CSS, JQuery, HTML.

2010-03-16T18:29:00.001Z

Situation: You have a an outer div with the id “outer” and an inner div called “inner”. The menuCont div is of fixed width; let’s say 960 for this example.

The inner div contains variable width elements and you cannot know at design time how many elements and of what width they will be. You cannot do this with CSS alone (but if you know how, please do share!)

Here’s the solution – it relies on jQuery BTW.

Set the CSS as follows:

#outer
{
    position:relative;
}
#inner
{
    float:left;
}

And HTML as follows:

<div id="outer">
  <div id="inner" style=”display:none”>
    <!-- Insert variable width elements here -->
  </div>
</div>

And then JavaScript:

$(document).ready(function() {
jQuery.fn.center = function() {
        return this.each(function() {
            var el = $(this);
            var w = el.width();
            var w_box = 960; // TODO: ADD THE CONTAINER WIDTH HERE!
            var w_total = (w_box - w) / 2;
            var css = { "position": 'absolute', "left": w_total + "px" };
            el.css(css)
       });
};
$("#inner").center();
$("#inner").fadeIn(); // optional! could use show()
});

This solution can definitely be improved – but it’s a start.

HttpModule and Application_Start

2010-03-15T20:45:00.001Z

Unfortunately there is no “Start” event to hook onto for HttpModules. Instead you should use Init() – be aware though that there may be multiple instances of your http module, serving many threads in the appdomain. So you need a bit of thread sync’ing in order to ensure that your start-up code only executes once.

    public class MyHttpModule : IHttpModule
    {
        /// <summary>
        /// Mutex object which gives the lock statement something to get an exclusive lock on
        /// </summary>
        private readonly static object _mutex = new object();
        /// <summary>
        /// Flags whether the application has started
        /// </summary>
        private static bool _isInitialised = false;
        #region IHttpModule Members
        public void Dispose()
        {
        }
        public void Init(HttpApplication context)
        {
            if (!_isInitialised) 
                lock (_mutex)  
                    if (!_isInitialised) 
                        Application_Start();
        }
        /// <summary>
        /// Application start-up code will be executed once here
        /// </summary>
        private void Application_Start()
        {
            _isInitialised = true;
            // add code
        }

ASP.NET, JQuery, Colorbox, Postback Problem

2010-03-15T15:18:00.001Z

As Colorbox appends inline divs to the body tag, this is outside the form.

See http://markive.co.uk/blog/?p=13

Change this: $('body').prepend($overlay, $cbox.append($wrap, $loadingBay))

To this: $('form').prepend($overlay, $cbox.append($wrap, $loadingBay))

or search in the minified file for “body”, and replace it with “form”.

I also recommend prepending the project name to the min file and putting a comment the top stating it’s been customised.

Umbraco, XSLT :: Reading RSS Feeds, Aggregating and Sorting

2010-03-12T15:20:00.001Z

Here's a quick summary of how to read multiple RSS feeds, merge them, sort them and then output the first 30 items. You can easily combine and concatenate nodes from external sources into an xsl:variable with xsl:copy-of and GetXmlDocumentByUrl.

Preparation

Set up a doc type named 'RssFeed' with a property called 'url' (textstring)
Create RSS Feed nodes in the Content section with your chosen feed urls
Create an XSLT macro with the code that follows.

XSLT

Just put this into an <xsl:template> and it will output the first 30 items.

<!-- Find all RSS Feed nodes, create new XML tree with all rss nodes merged into the 'container' node -->
<xsl:variable name="rssFeeds">
  <container>
    <xsl:for-each select="$currentPage/ancestor-or-self::root//node [@nodeTypeAlias='Feed Item']">
      <xsl:variable name="url" select="string(data[@alias='url'])" />
      <!-- Is the url populated? -->
      <xsl:if test="string-length($url) &gt; 0">
        <!-- Import the feed-->
        <xsl:copy-of select="umbraco.library:GetXmlDocumentByUrl( $url )"/>
      </xsl:if>
    </xsl:for-each>
  </container>
</xsl:variable>
<!-- Iterate through the container doc -->
<xsl:for-each select="msxsl:node-set($rssFeeds)/container/rss/channel/item">
  <!-- Sort the items -->
  <xsl:sort select="pubDate" order="descending"/>
  <!-- Only output the first 30 items -->
  <xsl:if test="position() &lt;= 30">
    <h1>
      <a>
        <xsl:attribute name="href">
          <xsl:value-of select="link/text()"/>
        </xsl:attribute>
        <xsl:value-of select="title/text()"/>
      </a>
    </h1>
    <p>
      <xsl:value-of select="description/text()" />
    </p>
  </xsl:if>
</xsl:for-each>

Javascript in data binding syntax

2010-02-17T14:39:00.001Z

onclick='<%# string.Format("return OpenWindow1(\"Edit/Supplier.aspx\", \"{0}\",\"WindowName\");", Eval("id")) %>'

SQL Test Data Generator

2009-09-22T12:25:00.001Z

Great script for generating test data in a SQL Server table:

declare @select varchar(max), @insert varchar(max), @column varchar(100),
@type varchar(100), @identity bit, @db nvarchar(100)

set @db = N'aqs_Lead'
set @select = 'select '
set @insert = 'insert into ' + @db + ' ('

declare crD cursor fast_forward for
select column_name, data_type,
COLUMNPROPERTY(
    OBJECT_ID(
       TABLE_SCHEMA + '.' + TABLE_NAME),
    COLUMN_NAME, 'IsIdentity') AS COLUMN_ID
from AQSProd2.INFORMATION_SCHEMA.COLUMNS
where table_name = @db

open crD
fetch crD into @column, @type, @identity

while @@fetch_status = 0
begin
if @identity = 0 or @identity is null
begin
    set @insert = @insert + @column + ', '
    set @select = @select +
        case @type
            when 'int' then '1'
            when 'varchar' then '''test'''
            when 'nvarchar' then '''test'''
            when 'smalldatetime' then 'getdate()'
            when 'bit' then '0'
            when 'decimal' then '0'
            else 'NULL'
        end + ', '
end
fetch crD into @column, @type, @identity
end

set @select = left(@select, len(@select) - 1)
set @insert = left(@insert, len(@insert) - 1) + ')'
exec(@insert + @select)

close crD
deallocate crD

Courtesy of Keith Bloom: http://stackoverflow.com/questions/1118960/create-test-data-in-sql-server

ThreadPool threads

2009-09-08T15:32:00.001Z

Note to self: When debugging inside Thread Pool threads, the debugger may miss out lines and exhibit erratic behaviour, which may drive you insane. Best thing to do is use your own Thread for all but the tiniest operations – or change back to a thread pool thread after debugging.

Polymorphic Application Integration End Points

2009-08-29T12:30:00.001Z

From now on, one of absolute “must-haves” for any software engineering effort is, to create polymorphic AI EPs. Sounds fancy, but it’s not.

All it really is, is dependency injection or IoC, not at the unit level, but at the application-integration level. At the unit level, for example, if you create a domain class which talks to a repository object, then you may swap out the repository for a mock repository during an automated unit test.

At the AI level, this is where you conceptually and logically separate out the application from other services or applications. e.g., whenever you use the SmtpClient class, you’re binding your code to the concrete implementation of an SMTP end point. SMTP is an end point, so are web services and WCF services. In fact, anything outside your domain of control. The database is also integration, hence, we often have database abstraction layers within architectures that are meant to abstract the data-storage technology from the code (ideally we could change the RDBMS from SQL Server to Oracle without changing any code)

Here are some example application integration end points AIEP:

Currency convert web service
Email/SMTP service
Database engine
Exception Reporting service
A web-browser (it’s the GUI but it helps to think of it as integration)
SMS messaging service
Logging service
MSMQ messaging

All these endpoints are about communicating beyond your application boundary. By loosely coupling your code to AIEPs, it means you have better ability to test your code in your development environment.

If you have ever accidentally sent out emails to real users from your test environment, then Polymorphic AIEPs will enable you to swap out your SmtpClient for MockSmtpClient.

For example:

public interface ISmtpClientEP{

void Send(string text);

}

public class SmtpClientEP : ISmtpClientEP{

public void Send(string text){

// instantiate the real smtpclient and send the message

}

public class TraceClientEP : ISmtpClientEP{

public void Send(string text){

// send your text to anywhere! e.g. a text file in a local directory?

}

Then if you use an IoC Container, such as Structure Map, then you can use this from the client code like this:

ObjectFactory.GetInstance<ISmtpClientEP>().Send(“hello world”);

But remember on application start up to link structure map to the concrete implementation of the class you want to use:

ObjectFactory.Initialize(x =>
{

    x.ForRequestedType<ISmtpClientEP>()
      .CacheBy(StructureMap.Attributes.InstanceScope.PerRequest)
      .TheDefaultIsConcreteType<SmtpClientEP >();

}

So what have we done here? We have utilised Structure Map to specify what the concrete implementation of ISmtpClientEP we wish to use.

Now if you have use conditional compilation arguments you can do this:

#IF(DEBUG)

ObjectFactory.Initialize(x =>
{

    x.ForRequestedType<ISmtpClientEP>()
      .CacheBy(StructureMap.Attributes.InstanceScope.PerRequest)
      .TheDefaultIsConcreteType<TraceClientEP >(); // use the trace end point for debug

}

#ELSE

ObjectFactory.Initialize(x =>
{

    x.ForRequestedType<ISmtpClientEP>()
      .CacheBy(StructureMap.Attributes.InstanceScope.PerRequest)
      .TheDefaultIsConcreteType<SmtpClientEP >(); // real one for release compilation!

}

#ENDIF

So polymorphic AIEPs utilise DI (Dependency Injection) to enable you to change (polymorph) your concrete implementations. This is just like automated-unit testing (TDD), except, in my view TDD can be a large overhead, so my hybrid solution is to use DI for Application Integration End Points and if you want, you can hard-wire you concrete implementations of other classes (classes that are internal to your app), without worrying about affecting other systems or services when you hit Debug.