Blogging is... contagious

Internal Compiler Error, create your own!

2011-04-18T09:18:00.007+01:00

You'd hope that it would be fairly difficult to make the compiler crash in Visual Studio. Sadly not. Try this one on for size:

public class CompilerFail {
  public int[] Numbers { set { ; } }
  public int FirstNumber { get { return Numbers[0]; } }
}

XmlDataSource and its wonderfully helpful cache

2010-07-08T16:52:00.003+01:00

From the looks of Google I've managed to stumble on a ~~problem~~ feature that's been causing joy for many years already, but it's a new one on me.

From the docs for XmlDataSource.EnableCaching it says the following:
The default value is true.

Which is a little strange for starters given some of its siblings (e.g. SqlDataSource and ObjectDataSource), but OK, fair enough.

It also says:
the data source will cache data until the XML file that it depends on is changed

Which would be fine, only my data source object isn't depending on a file, and I'm setting its Data property manually. Unfortunately, setting Data to what you want doesn't appear to invalidate the cache, and so you end up with what was there previously anyway.

But hang on a minute - the docs for XmlDataSource.Data specifically say:
If caching is enabled and you change the value of Data, the cache is invalidated.

I don't believe you!

Another bit of my life I'm never getting back, finding that little nugget...

jQuery Demonstration

2010-05-18T21:29:00.003+01:00

After a slight *cough* gap since my last post, I figured I should at least come up with something potentially useful to people this time.

Having just started a new role working on a bespoke ASP.NET application, I thought I'd push the case for introducing jQuery into the codebase to help with any client-side requirements. I was asked to give a presentation it to the dev team to get their views, so I spent today coming up with a self-contained demo, using jQuery itself.

I'm by no means a jQuery expert - I've essentially read half a book on the subject and slideToggle()d a couple of <div>s before - but I came up with the rest of it in less than a day, which probably says at least a little about how easy it is to get to grips with.

Anyway, without further ado... I present my short and sweet jQuery Demonstration. A couple of points to note:

You step through the demo by clicking the "jQuery" title at the top of the page;
You can click on the first grey markup block at any point to update it to show the current structure of the DOM;
When you get to the Gallery part, once the click handler has been attached to load the gallery HTML in the background, do actually click the Gallery header to load it before clicking the main jQuery title to get to the next (and final) step!

Enjoy!

Manually creating an HttpRequest for testing

2007-04-27T17:30:00.001+01:00

I'm currently writing unit tests for some code that uses the information found in the Headers property of an HttpRequest object. As such, I want to be able to create my own HttpRequest object with the appropriate headers.

This isn't as easy as I'd like it to be, as presumably the intent for the class is that you're handed an instance of it at the appropriate time, rather than putting one together yourself (which is fair enough).

Time to delve into System.Reflection for a bit of hacking:

// Whip up a fake HttpRequest
HttpRequest httpRequest = new HttpRequest("default.aspx", "http://www.diuturnal.com/default.aspx", string.Empty);

// Have to access the Headers collection first as this creates the internal collection we're about to hack
NameValueCollection headers = httpRequest.Headers;

// Accessing Headers property above will have created the private _headers member, so now we can grab it
headers = (NameValueCollection) httpRequest.GetType().GetField("_headers", BindingFlags.NonPublic | BindingFlags.Instance).GetValue(httpRequest);

PropertyInfo readOnlyInfo = headers.GetType().GetProperty("IsReadOnly", BindingFlags.NonPublic | BindingFlags.Instance);

// Hack the collection to not be read-only then add some headers
readOnlyInfo.SetValue(headers, false, null);
headers["Accept"] = "text/xml,application/xml,application/xhtml+xml,text/html;
q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5";
headers["User-Agent"] = "Mozilla/5.0 (Windows; U; Windows NT 5.1; en-GB; rv:1.8.1.2) Gecko/20070219 Firefox/2.0.0.2";

// Finished hacking, set it back to read-only
readOnlyInfo.SetValue(headers, true, null);

Job done!

UpdatePanel still performing full postback

2007-03-20T16:52:00.000Z

Had some fun tracking this one down today, fortunately someone else had already done the hard work in working out how to fix it.

Essentially I was trying out a basic AJAX page in an existing 2.0 web application project. The various additional required sections were added to the web.config and I followed the Introduction to the UpdatePanel control, where a label is updated when you click a button.

Simple. Or so you'd think. No AJAX behaviour for me - it was performing a full postback.

Fortunately I managed to find a post from "Rick" that explained what the culprit was - the <xhtmlConformance mode="Legacy"/> in the web.config. Commenting this out gives the desired behaviour.

I must admit I haven't bothered investigating any further to work out why that works - feel free to let me know.

Update:
Found another post about this that explains a bit more from Scott Guthrie.

Selecting random records from a database table

2007-03-06T10:05:00.000Z

A little nugget I picked up from my colleague Wim today, that was actually his first blog post a few years back.

Need a random sample of data from your database? Try this:

SELECT TOP 10 * FROM Orders ORDER BY NEWID()

DeflateStream and zlib headers

2007-03-02T15:39:00.000Z

The DeflateStream class in System.IO.Compression, as the docs state, represents the DEFLATE algorithm as in RFC 1951.

What it doesn't do (out of the box) is support compressed data in zlib format (see RFC 1950). When trying to decompress existing data, the easiest way I've found to get round this is simply to get rid of the 2 byte zlib header and use what's left, so something like the following works nicely if you've already got the data as a byte array in memory:

Array.Copy(zlibBytes, 2, strippedBytes, 0, zlibBytes.Length - 2);

Then just use the strippedBytes with DeflateStream instead.

Out of interest, I also tried stripping off the 4 byte Adler-32 checksum footer, and DeflateStream worked happily with that too.

Defining values in enums

2007-02-23T15:14:00.000Z

I got caught out earlier with something not working as I expected. It may well be documented behaviour somewhere, but I've not been able to find anything specifically mentioning it following a brief search, so I thought I'd document it.

In an attempt to follow the guidelines for declaring an enum, specifically, placing the values in alphabetical order and providing a default "zero value", I came up with the following:

public enum Decision {
  Accept,
  Error,
  Reject,
  Unknown = 0
}

Unfortunately this leads to the following values being assigned:
Accept = 0
Error = 1
Reject = 2
Unknown = 0

and thus Accept and Unknown become one and the same.

Make sure that any that are declared with a value go above any that aren't!

Login failures with ASP.NET 2.0 Membership

2006-09-14T11:00:00.000+01:00

I sometimes end up with a puzzled look on my face after trying to log into one of my web applications and getting a "login failed" message back despite being sure I've got the username and password right. The usual culprit is that I've forgotten to change the applicationName attribute in the config for the Membership section (explained nicely in a post by Scott Guthrie), but despite checking and re-checking when I had the same problem yesterday, it just wouldn't let me log in.

I started doing some tests, and found that if I wrote the code to do so, it would happily retrieve the password for a given username for me, but if I then used that same username and password in a call to Membership.ValidateUser() it still failed.

To cut a long story short, the problem turned out to be that the transaction log for the database was full, and the disk had run out of space. As such it's one of those rare occurrences that you may never encounter, but as it's rather obscure and there's no indication of what the problem is when the login attempt fails, maybe this post will help someone with the same issue one day.

Changing owner in SQL Server

2006-08-23T11:12:00.000+01:00

Given the number of times I've ended up having to search for this and promptly forgotten it 5 minutes later, I thought I'd put this up here so I know where to find it.

In order to change the owner of an object in SQL Server, use something like the following:

sp_changeobjectowner 'Database.dbo.Table', 'newowner'

Resolving links from the application root

2006-07-31T14:34:00.000+01:00

One of the things that I've got used to taking for granted is the ability to use the tilde (~) to specify the application root when specifying a link, in a HyperLink control, for example. Having recently ventured into the realm of writing my own controls I need to parse links like this for myself.

A quick Google later and it looks like the Control.ResolveUrl() method is the way to go, although as an aside, there appears to be a new Control.ResolveClientUrl() method in 2.0 as well. The docs don't really go out of their way to explain the difference as far as I can tell, but from my own tests it would appear that the former will provide an absolute path (using the TemplateSourceDirectory property according to the docs), whereas the latter will give you a relative path.

Quick example:
From a page at /WebApp/dir/page.aspx that wants to create a link to "~/otherdir/link.html":

ResolveUrl() gives: /WebApp/otherdir/link.html
ResolveClientUrl() gives: ../otherdir/link.html

Databinding with a LoginView inside a FormView

2006-06-08T00:52:00.000+01:00

No posts at all in April and now two within a day. Go figure.

I'm not sure if the title I've picked is the best way to describe this but it'll do. I've been using a LoginView within the ItemTemplate of a FormView to determine what the "available actions" are for a page. Here's an example:

<asp:FormView ID="fvApplication" runat="server">
  <ItemTemplate>
    <h1>Application Form</h1>
    
    <asp:LoginView ID="lvOptions" runat="server">
      <RoleGroups>
        <asp:RoleGroup Roles="Provider">
          <ContentTemplate>
            <asp:LinkButton runat="server" OnClick="Offer_Click"
              Visible='<%# ApplicationStatus.Sent == (ApplicationStatus) Eval("Status") %>'>Make Offer</asp:LinkButton>
            <asp:LinkButton runat="server" OnClick="Reject_Click"
              Visible='<%# ApplicationStatus.Sent == (ApplicationStatus) Eval("Status") %>'>Reject Application</asp:LinkButton>
          </ContentTemplate>
        </asp:RoleGroup>
        <asp:RoleGroup Roles="Student">
          <ContentTemplate>
            <asp:LinkButton runat="server" OnClick="Send_Click"
              Visible='<%# ApplicationStatus.Unsent == (ApplicationStatus) Eval("Status") %>'>Send</asp:LinkButton>
            <asp:LinkButton runat="server" OnClick="Accept_Click"
              Visible='<%# ApplicationStatus.Offered == (ApplicationStatus) Eval("Status") %>'>Accept Offer</asp:LinkButton>
            <asp:LinkButton runat="server" OnClick="Refuse_Click"
              Visible='<%# ApplicationStatus.Offered == (ApplicationStatus) Eval("Status") %>'>Refuse Offer</asp:LinkButton>
          </ContentTemplate>
        </asp:RoleGroup>
      </RoleGroups>
    </asp:LoginView>
  </ItemTemplate>
</asp:FormView>

This works fine when first coming into the page. The LoginView picks out the right template depending on the user's role, and when my Application object is databound to the FormView, the visibility of the individual buttons is set correctly. (Note that this is my own custom Application object, nothing to do with the framework.)

However, all is not well when making some changes to the application elsewhere on the page and then rebinding the FormView (i.e. this time we're dealing with a postback). This time, the LoginView fires its ViewChanging and ViewChanged events. I guess technically it's correct, as what I'm looking at on the page (i.e. my "view") may have changed if the various button visibilities have changed. However, my role hasn't changed, and hence the template the LoginView is using hasn't changed, which is kind of what I was expecting the ViewChanged event to mean.

Anyway, the problem comes in that somewhere in the middle of this "ooh, my view's changed" moment that the LoginView has, it blats its control hierarchy and recreates it, only this happens after the FormView was databound, and thus all the button visibilities are lost, resulting in them all going back to their default setting.

My workaround for now is to perform a Response.Redirect() and let everything create itself from scratch again as the way my page is setup allows this without any problems. However, if you need to remain within the context of a postback, I suggest using something like a MultiView instead of a LoginView, and manually set the correct view in there yourself at an appropriate time.

Suggestions for better solutions are, as always, welcome.

NextStepIndex in a Wizard control

2006-06-07T12:20:00.000+01:00

I've been using one of the new(ish) 2.0 Wizard controls in a project and it's been working nicely. That is, until I wanted to change it so that for one of the steps, you could leap out to a different page, update some other stuff, and then leap back in again to the step you were at before.

When coming back to the page and starting off at step number 2, I found that clicking the previous button didn't take me to step 1 as I expected (and wanted). A quick bit of breakpoint-setting and debugging later, I found that in the event handler, the NextStepIndex wasn't 0 (i.e. the first step), it was 1. Thus clicking the previous button simply left me on the second step.

This actually makes complete sense when placed in context, which a post on this thread was able to do for me. It's not previous in the sense of the last physical step as laid out in the .aspx file; it's previous in the sense of the last step in the history of your moves through the wizard. Since I was effectively starting off with a fresh copy of the wizard when coming back to the page, there was no history, and thus there was no previous step for me to go back to.

In the end, the solution is simply to specify what step you want next based on the CurrentStepIndex in the relevant event handler. Simple!

Membership configuration in ASP.NET 2.0

2006-05-15T15:30:00.000+01:00

Fine work on the documentation for this one. The documentation for Membership.MinRequiredNonAlphanumericCharacters suggests that the attribute in the <add> element when defining a provider should be minAlphanumericCharacters, or alternatively minRequiredNonAlphanumericCharacters, depending on which bit of the page you read.

In fact neither of these work, and you should use minRequiredNonalphanumericCharacters instead.

UK Postcode Implementation

2006-03-17T17:47:00.000Z

I've been meaning to put this up here ever since the day after my post about regular expressions but time has eluded me until now. I had a stab at coming up with some code for a postcode to save doing the same old checking (or alternatively none at all!) every time; let me know what you think.

Depending on how big your monitor is (and hence how wide you've likely got your browser) the ValidationString might wrap round - obviously this is all on one line.

namespace Diuturnal.Utility {
  [Serializable]
  public struct Postcode {
    private const string ValidationString =
      @"^\s*(?<out>[A-Z]{1,2}[0-9R][0-9A-Z]?) *(?<in>[0-9][A-Z-[CIKMOV]]{2})\s*$";
    private static readonly Regex _validator =
      new Regex(ValidationString);
    public static readonly Postcode Empty =
      new Postcode(string.Empty, string.Empty);
    private string _outCode;
    private string _inCode;

    private Postcode(string outCode, string inCode) {
      _outCode = outCode;
      _inCode = inCode;
    }

    public override string ToString() {
      if (0 == _outCode.Length) {
        return string.Empty;
      }
      return _outCode + " " + _inCode;
    }

    public override bool Equals(object obj) {
      if (!(obj is Postcode)) {
        return false;
      }
      return (this == (Postcode) obj);
    }

    public override int GetHashCode() {
      return this.ToString().GetHashCode();
    }

    public static bool operator ==(Postcode lhs, Postcode rhs) {
      return (lhs.ToString() == rhs.ToString());
    }

    public static bool operator !=(Postcode lhs, Postcode rhs) {
      return !(lhs == rhs);
    }

    public static Postcode Parse(string s) {
      if (null == s) {
        throw new ArgumentNullException();
      }
      string capitalised = s.ToUpper();
      if (!_validator.IsMatch(capitalised)) {
        throw new FormatException();
      }
      return new Postcode(_validator.Replace(capitalised, "${out}"),
        _validator.Replace(capitalised, "${in}"));
    }

    public static bool TryParse(string s, out Postcode result) {
      if (null == s) {
        result = Postcode.Empty;
        return false;
      }
      string capitalised = s.ToUpper();
      if (!_validator.IsMatch(capitalised)) {
        result = Postcode.Empty;
        return false;
      }
      result = new Postcode(_validator.Replace(capitalised, "${out}"),
        _validator.Replace(capitalised, "${in}"));
      return true;
    }
  }
}

Undocumented regular expression pattern

2006-02-24T17:06:00.000Z

UK Postcode validation is the theme for me today, and the standard has been very useful. Digging a little deeper even provides a couple of regular expression patterns for doing the validation with, both a simple one and a more complicated one.

Examining the patterns has led me to a construct that I can't find documented anywhere. I've looked through all the linked pages in the "Regular Expression Language Elements" section in the VS2005 help, but with no joy. It seems to work though.

[A-Z-[CIKMOV]]

Fairly self-explanatory hopefully, with this particular example matching any one of twenty capital letters. If it is actually documented somewhere, please do point me at it.

Multiple keys with KeyedCollection

2006-02-23T19:39:00.000Z

I'm currently dealing with some objects that have two different unique identifiers. One is a public, user-friendly identifier, and the other is an internal, essentially meaningless one. Meaningless, that is, other than the fact that I can ensure that it stays the same for all time, as opposed to the one that users can see, which therefore will no doubt need to be altered 30 times as random punter X changes their mind yet again.

As a result I wanted a collection where it's possible to grab an item quickly and easily using either of the identifiers. Deriving from KeyedCollection<TKey, TItem> does the job nicely for a single key, so I thought I'd extend that a bit farther and come up with my own cunningly named DoubleKeyedCollection<TKey, TSecondKey, TItem>.

To my shame, I've not even tested this code before posting it up here as it's already long past the end of the day, although it does compile, which is always encouraging if nothing else. I'm not hugely fond of the choice of GetItem() as a method name as it's a bit ambiguous, especially if your two keys are of the same type. GetItemBySecondKey() seemed a bit meaningless to an end client using a class derived from DoubleKeyedCollection though. Any thoughts on the subject are welcome (as is notification of any blatant errors that I've managed to completely overlook).

As an aside, I think I'll be examining the possibilities of changing the layout of this page. The narrow column doesn't exactly lend itself to code listings. I've done what I can to make it readable. EDIT: Template now changed - not as pretty imho, but makes the code a lot easier to read!

namespace Diuturnal.Utility {

  [Serializable]
  public abstract class DoubleKeyedCollection<TKey, TSecondKey, TItem>
    : KeyedCollection<TKey, TItem> {

    private Dictionary<TSecondKey, TKey> _secondKeyIndex =
      new Dictionary<TSecondKey, TKey>();

    public TItem GetItem(TSecondKey secondKey) {
      return this[_secondKeyIndex[secondKey]];
    }

    protected abstract TSecondKey GetSecondKeyForItem(TItem item);

    protected override void InsertItem(int index, TItem item) {
      base.InsertItem(index, item);
      AddSecondKeyIndex(this.GetSecondKeyForItem(item),
        this.GetKeyForItem(item));
    }

    protected override void RemoveItem(int index) {
      RemoveSecondKeyIndex(this.GetSecondKeyForItem(this[index]));
      base.RemoveItem(index);
    }

    protected override void SetItem(int index, TItem item) {
      RemoveSecondKeyIndex(this.GetSecondKeyForItem(this[index]));
      base.SetItem(index, item);
      AddSecondKeyIndex(this.GetSecondKeyForItem(item),
        this.GetKeyForItem(item));
    }

    protected override void ClearItems() {
      base.ClearItems();
      _secondKeyIndex.Clear();
    }

    private void AddSecondKeyIndex(TSecondKey secondKey, TKey key) {
      _secondKeyIndex.Add(secondKey, key);
    }

    private void RemoveSecondKeyIndex(TSecondKey secondKey) {
      _secondKeyIndex.Remove(secondKey);
    }
  }
}

Multiple search patterns with DirectoryInfo.GetFiles()

2006-02-09T20:19:00.000Z

Another visit to Google today showed various requests about how to specify more than one pattern to search for when grabbing the files from a directory, so for example, get all the .mp3 and .wma files in one call.

As far as I can see, you can only specify one pattern at a time (unless there's something new in 2.0 that I've missed) so I thought this was a good candidate for a simple wrapper. The code below should be fairly easy to modify if you have any specific requirements in your situation.

namespace Diuturnal.Utility {
  public class DirectoryHelper {
    public static FileInfo[] GetFiles(DirectoryInfo directory,
      params string[] searchPatterns) {

      List<FileInfo> allFiles = new List<FileInfo>();

      foreach (string pattern in searchPatterns) {
        allFiles.AddRange(directory.GetFiles(pattern));
      }

      allFiles.Sort(delegate(FileInfo x, FileInfo y)
        { return x.Name.CompareTo(y.Name); });

      return allFiles.ToArray();
    }
  }
}

FolderBrowserDialog.ShowNewFolderButton

2006-02-09T17:09:00.000Z

A quick Google will tell you that this has been discussed plenty of times before, but seeing as I stumbled across it of my own accord today, I thought I'd give it a mention.

Set the property (the title of this post) to false, then display the dialog on a Windows 2000 machine. Whoops, the "New Folder" button is still there. Oh well.

Apparently it works on Windows XP - I guess you just have to live with it if you're running Win2k.

Deploying to a Windows Mobile device

2006-01-21T21:19:00.000Z

A few months ago I got myself a Dell Axim X50v, convincing myself that it was a "good investment" as I'd be able to look into developing for the Pocket PC. In time-honoured fashion, I've never actually managed to get past a quick "Hello World" program on it, although I did think it was pretty cool the way you could just hit F5 in Visual Studio and up came the app on my PDA.

Anyway, I just decided I'd give it another bash with Visual Studio 2005 (Last time I tried I was using 2003) and upon hitting F5 I got the less than useful message "The configuration data for this product is corrupt. Contact your support personnel". Fortunately, Google is here to save the day again, and I found a post on a message board somewhere saying to close down Visual Studio, delete the directory C:\Documents and Settings\<localuser>\Local Settings\Application Data\Microsoft\CoreCon, and then restart and try again.

Did the trick for me, and I hope it works for you too if you find yourself here on a Google search of your own.

I get the impression that this folder is/was an artifact of an old beta install of VS2005 although I don't know that for certain. If anyone knows for sure where it comes from please do feel free to enlighten me!

Serialization and type versions

2006-01-18T17:12:00.000Z

After stating in my previous post that I write messages in en-GB, I go and ruin that by putting "Serialization" in the title of this one. Go figure.

Another nice addition in .NET 2.0 are five attributes in the System.Runtime.Serialization namespace. These help with telling v2 of a class how to cope when dealing with objects of type v1. The attributes are as follows:

OnDeserializedAttribute
OnDeserializingAttribute
OnSerializedAttribute
OnSerializingAttribute
OptionalFieldAttribute

Read the docs for usage, but essentially if you mark the new member fields in v2 with [OptionalField] then it doesn't fall over when attempting to deserialize an object of type v1, as it knows that the info might be missing. You can then use a method marked with [OnDeserialized] to do sensible things like set a default value for a missing field for example.

"Neutral" Resources

2006-01-11T12:03:00.000Z

The system may have a concept of a "neutral language", but in the real world I tend to write my default messages in a concrete language (which for me is en-GB). When it comes to doing localisation, there's no need to then duplicate the effort and create a separate en-GB resources file - simply mark your "neutral" version with what it actually is.

You can do this with an attribute (in AssemblyInfo.cs), like so:

[assembly: NeutralResourcesLanguage("en-GB")]

EDIT: Should have guessed this already really, but I've just realised that in VS2005 with its new GUI to edit the AssemblyInfo.cs file, this attribute is a drop down inside the Assembly Information page anyway. Simplicity itself.

Null coalescing operator

2006-01-10T15:04:00.000Z

Fancy title to get things going.

This is one of those things that'll be second nature once I start using it in anger, but until I do that, I'll consistently forget it.

C# 2.0 introduces nullable types, and the null coalescing operator (??) is a binary operator that works pretty much the same way as T-SQL's ISNULL() function.

So for:
a ?? b

think:
ISNULL(a, b)

One small caveat here is that b could of course be a nullable type as well, in which case the result of a ?? b is still a nullable type. If it's not, the result isn't either. See example below (ripped from C# 2.0 spec).

int? x = GetNullableInt();
int? y = GetNullableInt();
int? z = x ?? y;
int i = z ?? -1;

First post

2006-01-10T14:34:00.000Z

I figured it was probably fairly obligatory to have a "first post" post. So here it is. At this point I have absolutely no idea if I'll continue this blog beyond the initial novelty value length of a day or two.

Essentially my only motivation for doing this is as a repository for little tidbits of info I find out whilst wading my way through the world of .NET programming, as I've lost count of the number of times I've thought, "I remember reading about how to do that... now if only I could find it..."

With any luck if I convince myself to dump it in here then I'll be able to find it again later. So if you've got no interest in programming, look elsewhere. I don't feel the need to share with the world what I had for tea last night, who I babysat for at the weekend, or why I'll be choosing gloss rather than matt paint for decorating.

Enjoy.