Navigating in Xamarin

I was listening to a podcast (Merge Conflict) recently about the latest changes in Xamarin, and being a curious soul, decided it was time to at least understand what was going on. This time, the big new feature was the Xamarin Live Player which promises much faster debugging through your actual phone! Instead of being stuck using the creaky old Android emulator on Windows, or waiting 2 – 5 minutes on every build with your phone plugged in, you could now, with the power of the internet, debug wirelessly almost instantly. Obviously, I was suspicious, but this sounded like fun, so I jumped in.

Getting setup was a bit of a chore, I will admit. Most of the trouble was linked to being an early adopter. You see, Xamarin Live Player wasn’t truly released yet, just part of the latest preview of Visual Studio (v15.3 to be exact). So I had to get that, install almost all features (say goodbye to about 40GB on your hard drive, sorry ultrabook owners), and then I was basically ready to go. The final step was installing their app on my phone, and I was ready to go.

Within about 10 minutes I had the demo project running on my phone, pretty good I thought. I used the Master/Detail project template and that came with some basic navigation to boot. Everything looked great…but was it?

Being a big fan of MVVM, I really have 3 things I need in place before I can start building an application.

  1. A base class implementing INotifyPropertyChanged. Luckily this project came with one called ObservableObject.
  2. A class implementing ICommand. Again, Xamarin has us covered here, with their implementation imaginatively named Command.
  3. Some kind of ViewModel first navigation system, which appears to be missing.

So, before I get into what’s missing, I want to say that obviously, it’s not Xamarin’s job to create this stuff for us. I don’t need to be spoon-fed. Also, I get that they’re trying to set us up with a working system to get us rolling, but I don’t think this is it.

I’ve never liked View-first navigation in MVVM apps. I don’t really get how you’re supposed to be able to pass anything around, unless you’re constantly creating a View, then passing the newly created ViewModel into that in the constructor. Gross. Unfortunately, that’s what this demo app does, and it took me a few hours to figure out what everything was and what was doing what, before I realized I needed a better way. In the template you get, they’re basically doing navigation from events in the code-behind of the views (or pages in this case). Again, not a fan.

My first instinct was to do my old favorite, using DataTemplates, and just defining the View / ViewModel pairs explicitly in Xaml somewhere and being done with it. Unfortunately, it looked like this wasn’t going to fly in Xamarin.

So I did what I always do, started googling it. And it worked. I came across a library called VMFirstNav. It contained most of what I wanted, but had a little more in it than I wanted. Using their implementation, you have to make all Views implement an interface telling the library what ViewModel to link it to, and then all ViewModels have to implement another interface. It wasn’t a dealbreaker, but I wanted to see what he was actually doing. The nice thing of all that was that it would automagically figure out what View to load when you change (or push) to a new ViewModel.

This is where I want to thank CodeMillMatt. Without his code, I would have taken a lot longer to figure out how this all works. It’s good stuff dude.

OK, so what I did was take his library and just boil it down to the one important class, the NavigationService, which I named Navigator. I didn’t bother making a new github project, or putting it in Nuget, as I don’t think this is really finished. I could see a need for more features depending on what you need, but this is at least a solid starting point.

Here’s the code for the class –

And here’s how to use it, once you’ve got it in place. It has a static constructor, so it’s essentially a singleton, so you don’t need to go passing it in and out of every class. It’s a little more decoupled, but I prefer it that way.

There’s 2 ways of using it.

  1. Explicitly specify all the View / View Model pairs on startup. In the past I’ve liked this method as at least you know where they are and how it works. The other advantage of this is you can name these classes whatever you want.
     private void SetupAllViewToViewModelPairs()
         var nav = Navigator.Instance;
         nav.Register(typeof(LoggingViewModel), typeof(LoggingView));
         nav.Register(typeof(AnotherViewModel), typeof(AnotherView));
  2. Follow a simple naming convention, and the service will automagically figure out what View to load when you tell it to load a certain View Model. The convention can be changed (it’s your code now after all). As long as you follow the scheme below, you’ll be good.

OK, then to use it, you’re basically just doing the following –

var vm = new [ClassName]ViewModel();
await Navigator.Instance.PushAsync(vm);


Bear in mind, there is no error checking on this code, so maybe tidy that up a little if you plan on using it. The only part you would need to be more careful of is probably the automatic View locating part.

Just to wrap this up. I played with Xamarin for a couple of days, but found the limitations to be too hard to work with. I’m sure there’s a lot of uses for it, but we don’t do a lot with web services at my work, so I haven’t got any excuses to work with it.

If you’ve got any comments on the code, or you’ve built anything cool in Xamarin that you’d like to show off, let me know in the comments.

Vertical Airspace

Sometimes when you’re working in WPF, you’ll come across an issue that just stops you dead in your tracks. The kind of issue that makes you wonder how in the hell this problem can still exist in version 4+ of a product. Well, today I’m going to talk about one of those, namely Airspace.

Odds are that you haven’t heard of airspace, it’s one of those weird issues that you don’t need to know about until you need to know about it. Here’s the best simple explanation I could find about it. Basically, the problem rears its head when you’re trying to host a WinForms control inside a WPF hosting window or vice-versa. What happens is that the WinForms control has to be on the topmost layer, no matter what. Even if you try to change the z-index of another control, the WinForms control will always be on top. In reality, this means you can’t show any modal dialogs on top of any WinForms controls.

OK, OK, I know, you’re probably thinking this won’t affect me, it doesn’t really matter, I never use any WinForms controls inside my super-shiny-awesome WPF apps. Welp, maybe you’re right, but maybe, just maybe you’re going to use the WebBrowser control!


The WebBrowser control is a WinForms control (they don’t like to publicize that fact) that’s basically been wrapped for WPF.

I found this out myself just a few weeks ago, when I tried to throw some Markdown preview into my application. I thought the easiest thing to do would be to just throw in a standard-issue WebBrowser and render to it. Little did I know the rabbit hole I was about to fall into. All worked great until I tried to put my dialog on top of the browser and it just didn’t. I can’t show the original example here, but I threw together a quick example of what happened.


The highlighted box above has a bunch of standard WPF controls, while the WebBrowser takes up most of the space. When I hit the settings button it’s supposed to show a dialog on top of the entire window and darken the background a little bit. As you can see, the dialog can’t be seen as it’s covered by the WebBrowser but you can see the top part has the darkened background.

I did a little searching on how to fix the problem, thinking it would be an easy fix, which led me here and to a false solution here which stated it would be fixed in WPF 4.5, but it never came to be. I guess the official solution is waaaaay more complicated than it looks. Some people recommended just hiding the WinForms control when showing a dialog, which, in theory would work, but is not very subtle, especially if it’s not a full-screen dialog.

I went further into the internet and came across both CEFSharp and Awesomium which both seem super powerful and useful, but also over-powered for what I needed. Add to the fact that both have a lot of JS native code calls to do the more complicated parts, and I was starting to get a little scared off.

Finally, after some thinking, I decided I should try and fix the problem that I’m actually having, that I can’t show a dialog on top, so I came up with a pretty hacky little workaround that I wanted to share with everyone.

AirspacePanel aims to fix this problem by allowing you to host any content you like inside of it, and swapping out the content for a screenshot of the content when you need to place a modal dialog on top of it. This swap is done by using the dependency property FixAirspace. Here’s an example –


<asf:AirspacePanel FixAirspace="{Binding FixAirspace}">
    <WebBrowser x:Name="Browser" />

The entire project (in all it’s tiny glory) is available here on github which contains a sample project if you just want to see it. Here’s the screenshot from above all fixed with AirspacePanel.


You can also find it in Nuget through Visual Studio.

If anyone does give it a shot, let me know what you think. For me, it fixes a small annoyance, but I’d like to hear some other opinions.

Command-Line into ClickOnce


The company I’m with right now has totally bought into ClickOnce. I’ll admit that before I started working here I’d never heard of it. In theory it seems like a nice idea, and generally it seems to work well, but like a lot of Microsoft products, it’s kind of stagnating. I get the impression Microsoft is calling it done on this one, and if you google it, you’ll get some great results like these.

I’m not here right now to discuss the merits or downsides of ClickOnce, I’m just going to describe my (painful) experiences attempting to get command-line access into my application.

I’m a big fan of shortcuts, especially AutoHotKey, so I thought I’d create a shortcut to some of the most-used functions. I’ve done it before and it worked great, how much different could this be…

First things first, ClickOnce doesn’t tell you where it’s installed, so part one of this, is find the application you’ve deployed. The best way I could find to do this was to go to the Start Menu and locate the shortcut. Normally, this would be just a shortcut, and point you in the right direction, right? Nope, it’s actually an application reference, which it turns out, you can use to reference the real app.

On my Win 8.1 system I found this shortcut here – C:\Users\**YOU**\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\MyAwesomeApp.

The application I was trying to use here was a WinForms app, so YMMV if you’re trying this in WPF or straight up in a console app. Hopefully it’s similar enough that you can hack along here with me.

OK, locate the application entry point, in WinForms this is Program.cs. You’ll need to add a check for the args being passed in and, in my case, attach a console app so the user can see what’s actually happening as you’re working.

static class Program
    private static extern bool AttachConsole(int dwProcessId);
    private const int ATTACH_PARENT_PROCESS = -1;

    static void Main()

    private static void Initialize()
        // Read the command line arguments
        string[] args = AppDomain.CurrentDomain.SetupInformation.ActivationArguments.ActivationData;

        if (args == null || args.Length == 0)
            Application.Run(new Form1());
            // Only attach the console when using the command line

            CmdLineHandler cmdLineObject = new CmdLineHandler(args[0].Split('!'));

So, as you can see, I’m PInvoking the console control (who doesn’t love a good pinvoke). In Initialize(), I’m checking to see if any args are being passed to the ClickOnce app using


This way I can start up the regular old forms app if no args are passed, or jump to the console app if there are. Then I’ve created a nice CmdLineHandler class to handle all the string parsing magic to make this work.

Here’s the important part, I’m serious, don’t skim over this part – you CANNOT use comma or space to separate your args when command-lining into a ClickOnce app. OK, I think I’ve bolded enough to make that stand out. For some reason, the comma in your args get’s swallowed. I’m not sure where it happens, but I know it happened every single time (I tried a bunch of different ways). Feel free to give it a try, hey, if you get it working, let me know!

OK, so I’m splitting those input args by !, which looks weird, but works fine. Once you’ve got the args, it’s up to you, I’m sure you’ll be fine.

Here’s the final command to my app

C:\Users\**YOU**\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\MyAwesomeApp -option1!yes!-option2!no

Again, I’m not endorsing ClickOnce, or this really, but if you’re stuck in this situation at work, this is my workaround.

Expression Trees

If you’re anything like me, you’re lazy.

I choose a lazy person to do a hard job. Because a lazy person will find an easy way to do it.

– Bill Gates

Apparently, in the world of programming, being lazy is a virtue, so maybe by passing this along I’m being virtuous?!?

I came across a library I had to implement this week that had function calls down to some hardware. Now since this library was a wrapper for a C (or C++) library, and whoever wrote it didn’t want to put a lot of work into it, every function called would return an ErrorCode enum.

Now if I was to be a copy and paste wizard, I’d just write myself about 20 of these

var errorCode = HARDWARE.GetSomeInformation(1, true, ref why???);

if (errorCode != ErrorCodes.None)
    Throw new CustomErrorException();

Obviously, that sucks. Nobody wants to do this. If you want to do this, you’re crazy. Even if you’ve got all the time in the world, or you’re a Regex master, or you’re using a tool called NimbleText you still shouldn’t do this. I think this is what people refer to as brittle code. If anything ever changes, or you realize you wanted to do something else with that error code, you’ve got to change it in 20 places (or more, how big is this library anyway?).

I decided to wrap this error call in a nice function delegate that would allow me to put all my ErrorCode handling in one place and just get on with stuff. Here’s what I ended up with

HARDWAREWrapper(() => HARDWARE.GetSomeInformation(1, true, ref why???));

private void HARDWAREWrapper(Func<ErrorCodes> functionToCall)
    var errorCode = functionToCall();

    if (errorCode != ErrorCodes.None)
        Throw new CustomErrorException();

OK, much better, now if anything in those function calls changes generically, or I decide to add more information to my exception we’re all good. It’s just one spot. Single responsibility principle? Not sure if that exactly counts, but my goal today is to sound smart by using lots of programming words (let me know how you think I’m doing).

So, here we go with the downside of this, since this is a generic call, it’s a little harder to figure out what exactly caused the exception in the first place. Since any code could be in that function (as long as it returns ErrorCodes) it’s a little harder to debug, especially if this is in the field and all.

Luckily I came across something pretty sweet called an Expression. I can wrap the Func<ErrorCodes> in this Expression to allow me access to the actual text of the function for debugging purposes. I’m sure that’s not what it was designed for, but it’s what I’m using it for today. Here’s the code –

HARDWAREWrapper(() => HARDWARE.GetSomeInformation(1, true, ref why???));

private void HARDWAREWrapper(Expression<Func<ErrorCodes>> expression)
    var functionToCall = expression.Compile();

    var errorCode = functionToCall();

    if (errorCode != ErrorCodes.None)
        Throw new CustomErrorException();

As you can see, the call to the function doesn’t actually change, even though it’s wrapped in an Expression, but it does allow you to log the function delegate contents to file. With Expression.ToString() I get access to the line HARDWARE.GetSomeInformation(1, true, ref why???) in text format.

Awesome. OK, that’s all I’ve got. If anyone out there is listening, let me know what you think!

Visual studio snippets

I only recently graduated from part-time coder to full-time coder. My previous job was part engineer, part programmer, so I never really spent enough time looking into some of the nicer features of Visual Studio.

Luckily, that’s all changed now. Snippets have by far become my favorite coding shortcut. At first, I didn’t think I had much use for it. If I had to create a long chunk of code I’d just revisit an older project and copy it over, no problems right? Well, now that I’m always coding I see what a waste of time that was.

I’m sure you can find out how to do these just about anywhere, in fact, I’m certain MSDN has a nice section on it, but I’m going to run you through it all the same.

Here’s my favorite snippet, it’s for an IConverter to be used in your xaml. These always have exactly the same starting point, so they’re perfect.

Step one is create a new file with the extension .snippet. In this case, I named it IConverter.snippet.

Now copy all the following code into it and I’ll go through it step by step. I’ve highlighted all the sections you’ll want to change.

<?xml version="1.0" encoding="utf-8"?>
<CodeSnippets xmlns="">
  <CodeSnippet Format="1.0.0">
        <Code Language="csharp"><![CDATA[**[ValueConversion(typeof($type1$), typeof($type2$))]
    public class $class$ : IValueConverter
        public object Convert(object value, Type targetType, object parameter, CultureInfo culture)
            if (!(value is $type1$))
                return 0;


        public object ConvertBack(object value, Type targetType, object parameter, CultureInfo culture)
            throw new NotImplementedException();

OK, it’s pretty straightforward from here. Give the snippet a Name and Author, although they don’t really matter too much. More importantly, replace SnippetShortcut with something useful, in my case I used conv. This is the shortcut you’ll use in Visual Studio to run the snippet. All you need to do it type conv[TAB][TAB] and you’re up and running.

In the declarations section,


You can declare all the variables in your code snippet, with an ID and Default Value. The ID can be used in the codeblock below. You have to surround the ID in $, like this $type1$ for the snipper to find and replace it correctly. The default value should be something sensible. I like to try and make that a useable value, so if I forget what that should be, the default gives me a hint.

Finally, replace the section that is highlighted bold (don’t delete the [CDATA[ or $end$ sections or it’ll break). Make sure to replace any section that might be a variable in your code block with the correct $variablename$.

OK, so you’re all done, now what? Well, open up Visual Studio, goto Tools —> Code Snippets Manager, click Add and locate your snippet folder. That’s it.

Now, after all that hard work you can relax and enjoy the beauty of code snippets.