Adventures in C#8 – The annotation for nullable reference types should only be used in code within a ‘#nullable’ context.

If you’re playing around with C# 8, and have been for a while, you may spot this message when using nullable reference types:

Warning CS8632 The annotation for nullable reference types should only be used in code within a ‘#nullable’ context.

There’s a couple of things you can check to fix this; the first is to check that you are, in fact, using C# 8. It’s likely that you are, because there is a separate error if you are not, but just for my own benefit:

<LangVersion>preview</LangVersion>

The next, and more likely step, is that you’re using the previous declaration for Nullable Reference Types; it used to look like this:

<NullableReferenceTypes>true</NullableReferenceTypes>

But no more! Now it looks like this:

<NullableContextOptions>enable</NullableContextOptions>

And, as if by magic – the error disappears!

But why? Well, it looks like they are making the setting more nuanced than simply on or off.

Working out when something broke using Git Bisect and Git Log

On 5th Feb, I attended DDD North. There were many excellent talks on the day, but in this post, I’m delving into something that I saw in one of the grok talks (I gave one myself). It concerns a feature of git that I’d never heard about until then, and it’s proved to be enormously useful. This post is, as all of my posts are, really for myself.

Something has broken

You’re giving the app a quick test, and suddenly you realise that something in the application has stopped working (or maybe it’s started, but it shouldn’t have). You remember that two days ago, this particular thing was working fine. This post is going to cover manually using git bisect to determine when your code was broken.

Get the good commit

To use git bisect, you need a good commit and a bad commit. Obviously, git doesn’t know what’s good or bad, so you could use it to determine when something was changed, added, or removed. Providing the behaviour is different, you can identify it.

If you know the date of the ‘good’ commit, you can ask git for the logs after that date. To list commits after a given date:

git log --after="2019-03-19T00:00:00-00:00"

This will give you all the commits (you can add a ‘–before’, too if you’re dealing with a long time ago. When you get the git Commit Id, make a note of it; for example:

Bc1afa359dc98b899cc4194ec8130a6229643172

Let’s do some bisecting

You start the bisect process by:

git bisect start

At all times through this process, git does precisely nothing to help you along, so if you even get an acknowledgement of your command you should consider yourself lucky!

The next step is to tell git that the current release is bad:

git bisect bad

You can give it a commit here that isn’t the current release, if you so choose.

The final step to start the process is to tell it where to start – i.e. what was the last known good release. This is your commit Id from earlier:

git bisect good Bc1afa359dc98b899cc4194ec8130a6229643172

What git does here is a binary chop. So the first code it will check out will be halfway between the two commits. It then tells you nothing (as usual). Now that it’s checked out the code for you, run your application with the code and see whether it is ‘good’ or ‘bad’. Once you’ve determined that, go back to git and tell it:

git bisect good

Or:

git bisect bad

Each time you tell it it’s good or bad, it will check out new code, and you need to test again:

(apologies for all the redacting)

Finally, it will determine which commit was the first bad one, and report back:

$ git bisect bad
Eea8e3d72fcb26cdebb2dbf2c13fdd88d7f3782a is the first bad commit
commit Eea8e3d72fcb26cdebb2dbf2c13fdd88d7f3782a
Author: Ian Kilmister <[email protected]>
Date:   Wed Mar 20 11:49:18 2019 +0100

    New lyrics

Once you’re done, you need to reset – this will book out the code you had out before you started:

git bisect reset

Read a Document in a .Net Core Application Baked into the Project File

This isn’t a difficult thing to achieve, but it is one that frequently has me reaching for Google to get the exact syntax. By creating this post (I create all my posts on OneNote first), it’s now offline for me.

There’s more than one way to do this; I’ve covered two. All the examples here are using a CSV file from an Assets folder:

Embedded Resource

These are embedded into the compiled assembly; which means in order to change them, you would need to recompile your code. This works well for images, but less well for data files.

First, you need to add your resource to your project, and set the properties to be an embedded resource:

Because it’s embedded, copying makes no sense. To read the file:

public string GetResourceTextFile(string filename)
{
    using Stream stream = this.GetType().Assembly.
                       GetManifestResourceStream($"SalesOrder.Generate.{filename}");
    using StreamReader sr = new StreamReader(stream);                           
            
    return sr.ReadToEnd();
}

Content

This allows you to include a file within your project; however, content files are not compiled. Combined with the “Copy To Output Directory” they place files in the binary directory of your project. The advantage here is that you can change this file after compilation:

Because this just copies the file to the executing directory, it’s easier to read the file:

public string GetContentTextFile(string filename)
{
    return File.ReadAllText($"Assets/{filename}");
}

References

https://stackoverflow.com/questions/145752/what-are-the-various-build-action-settings-in-visual-studio-project-properties

My XUnit Tests won’t run in a .Net Standard 2.0 Class Library

Firstly, this isn’t a bug, or something that you might have done wrong; it’s intentional. Essentially, you can’t run a .Net Standard Library, so your tests aren’t runnable.

Okay – so I want to convert to .Net Core 3.0!

Yep – that’s exactly what you want, and it’s this easy; open up the csproj file – it’ll look like this:

<Project Sdk="Microsoft.NET.Sdk">
  <PropertyGroup>
    <TargetFramework>netstandard2.0</TargetFramework>
  </PropertyGroup>

And replace it with this:

<Project Sdk="Microsoft.NET.Sdk">
  <PropertyGroup>
    <TargetFramework>netcoreapp3.0</TargetFramework>
  </PropertyGroup>

And that’s it – your tests should now run!

The Evolution of the Switch Statement (C#8)

Most languages have a version of the switch statement as far as I’m aware; I must admit, I don’t remember one from Spectrum Basic, but ever since then, I don’t think I’ve come across a language that doesn’t have one. The switch statement in C was interesting. For example, the following was totally valid:

switch (value)
{
    case 1:
        printf("hello ");
    case 2:
        printf("world");        
}

If you gave it a value of 1 would print “hello world”. When C# came out, they insisted on using breaks at the end of case statements, or having no code (admittedly there were a few bugs in C caused by accidentally leaving break statements out):

            int value = 1;
            switch (value)
            {
                case 1:
                    Console.Write("hello ");
                    break;
                case 2:
                    Console.Write("world");
                    break;
            }

Anyway, fast forward around 17 years to C# 7.x, and it basically has the same switch statement; in fact, as far as I’m aware, you could write this switch statement in C# 1.1 and it would compile fine. There’s nothing wrong with it, so I imagine MS were thinking why fix it if it’s not broken.

There are limitations, however; for example, what if I want to return the string, like this:

            int value = 1;
            string greeting = string.Empty;
            switch (value)
            {
                case 1:
                    greeting = "hello ";
                    break;
                case 2:
                    greeting = "world";
                    break;
            }

            Console.WriteLine(greeting);

Now it looks a bit cumbersome. What if we could write it like this:

            int value = 1;
            string greeting = value switch
            {
                1 => "hello ",
                2 => "world",
                _ => string.Empty
            };

            Console.WriteLine(greeting);

From C# 8, you can do just that. The switch statement will return its value. The case syntax is disposed of, and there’s no need for a break statement (which, to be fair, can encourage people to write large swathes of code inside the switch statement – if you don’t believe me then have a look in the Asp.Net Core source!).

And that’s not all. Pattern matching has also been brought in; for example, take the following simple class structure:

    interface IAnimal
    {
        void Eat();
        void Sleep();            
        string Name { get;}
    }

    class Dog : IAnimal
    {
        public string Name { get => "Fido"; }

        public void Eat()
        {
            Console.WriteLine("Dog Eats");
        }

        public void Sleep()
        {
            Console.WriteLine("Dog Sleeps");
        }
    }

    class Cat : IAnimal
    {
        public string Name { get => "Lemmy"; }

        public void Eat()
        {
            Console.WriteLine("Cat Eats");
        }

        public void Sleep()
        {
            Console.WriteLine("Cat Sleeps");
        }
    }

We can put that into a switch statement like this:

            IAnimal animal = new Cat();
            string greeting = animal switch
            {
                Dog d => $"hello dog {d.Name}",
                Cat c => $"hello cat {c.Name}",                
                _ => string.Empty
            };

            Console.WriteLine(greeting);

We can actually do better that this (obviously better is a relative term). Let’s say that we wanted to do something specific for our particular cat:

            IAnimal animal = new Cat();
            string greeting = animal switch
            {
                Dog d => $"hello dog {d.Name}",
                Cat c when c.Name == "Lemmy" => $"Hello motorcat!",
                Cat c => $"hello cat {c.Name}",                
                _ => string.Empty
            };

            Console.WriteLine(greeting);

It’s a bit of a silly and contrived example, but it does illustrate the point; further, if you switch the case statements around for the general and specific form of Cat, you’ll get a compile error!

A C# Developer’s Guide to: ReactJS – Part 2 – Moving Controls

Following on from my previous post, I’m going to extend our ReactJS application by adding some boxes, and allowing the user to re-arrange them on the screen.

Concepts

There are two key concepts to consider when working with React, and we’ll cover one of them in this post; that is: state.

React has a special property of each component known as state. If you use that to bind any of the UI to, then React will refresh that component when the state changes.

Moving a UI Element

Okay – that sounds great, but how would we do this in practice?

Imagine that you have a HTML box drawn on the screen; you might have something like this:

<div style="height:100px; width:200px; background:#0000FF" />

We can draw a box. If you use this code, then your box will be on the top left of the screen; so we can tell it not to be by specifying the left and top; let’s try defining a CSS style:

<div style="left:10px; top:20px;height:100px; width:200px; background:#0000FF" />

With React, we can set those values to a value derived from the state; for example:

render() {
    const top = this.state.newY;
    const left = this.state.newX;

    const myStyle = {
        height: '100px',
        width: '200px', 
        top: `calc(${top}px)`,
        left: `calc(${left}px)`, 
        borderStyle: 'solid',
        borderColor: 'blue', 
        position: 'absolute',
        zIndex: 1,
        background: 'blue'
    };

    return ( 
        <div> 
            <div style={myStyle}>source</div>
        </div>
    );
}

What this means is that every time I change the state values, the values in the style will update.

To illustrate this, if we write a React application like this (only the timeout function is new):

render() {
    const top = this.state.newY;
    const left = this.state.newX;
    const myStyle = {
        height: '100px',
        width: '200px', 
        top: `calc(${top}px)`,
        left: `calc(${left}px)`, 
        borderStyle: 'solid',
        borderColor: 'blue', 
        position: 'absolute',
        zIndex: 1,
        background: 'blue'
    };

    setTimeout(() => {
        this.setState({newX: this.state.newX + 1, newY: this.state.newY + 1});
    }, 50);

    return ( 
        <div> 
            <div style={myStyle}></div>
        </div>
    );
}

We can make out box move diagonally across the screen:

Screenshot included because you couldn’t possibly imaging what a blue rectangle looks like.

Start the app using VS Code

A quick note on VS Code. If you select Terminal -> New from the menu, you can run the React application directly from the IDE; and the best part is that if there’s something already running on the port, VS Code will just give you a new one:

A C# Developer’s Guide to: ReactJS – Part 1 – Create & Run

I thought it might be time for another of my patented “C# Developer’s Guide to…”. Basically, a guide to React for someone that understands programming (C# ideally), but doesn’t know anything about ReactJs.

There are a couple of VS templates for React; however, after working with it for a while, you realise that most of the React people live in the command line; there’s a tool called create-react-app:

You can use Powershell or Bash as your command line of choice. Start by installing the tool:

npm i -g create-react-app

This assumes that you have installed npm.

Create

To create a new application, navigate to the directory that you wish to create the application in and type:

create-react-app [app-name]

For example:

create-react-app context-menu-test

Once created, it tells you how you can start the app:

Run

First navigate into the directory that it’s just created (otherwise you’ll get errors when you run).

npm start

Will launch the server, and navigate to the Url with the default browser:

cd context-menu-test
npm start

This runs on port 3000 by default, so if you run a couple of instances then you might get the error:

Something is already running on Port 3000

To fix this, in VS Code, open the folder:

node_modules/react-scripts/scripts/start.js

Find the following line (at the time of writing):

const DEFAULT_PORT = parseInt(process.env.PORT, 10) || 3000;

You can change the port 3000 to something that’s free.

Edit App.js:

class App extends Component {
  render() {
    return (
      <div className="App">
        <div>
          <p>Test</p>
        </div>
      </div>
    );
  }
}

Save and see the updated screen appear (which I think you’ll agree is much better than the default screen).

Creating a Windows Service using .Net Core 2.2

Up until very recently, creating a Windows Service was the domain of the .Net Framework. However, since the release of the Windows Compatibility Pack that has all changed. In this article, we’ll create a .Net Core Windows Service from scratch.

I’m using the preview version of Visual Studio 2019 for this post. As far as I’m aware, there is absolutely no functional difference between this and VS2017; however, the initial New Project screen does look a little different.

Create the Project

There are no “New Windows Service (.Net Core)” options in Visual Studio, so we’re just going to create a console application (everything is a console application in .Net Core):

The .Net Core application can target .Net Core 2.2:

Windows Compatibility

The next step is to install the Windows Compatibility NuGet package:

Install-Package Microsoft.Windows.Compatibility

Write the Service

Let’s start with the main method:

static void Main(string[] args)
{
    using (var service = new TestSevice())
    {
        ServiceBase.Run(service);
    }
}

You’ll need to Ctrl-. ServiceBase. TestService doesn’t exist yet, so let’s create that:

internal class TestSevice : ServiceBase
{
    public TestSevice()
    {
        ServiceName = "TestService";
    }

    protected override void OnStart(string[] args)
    {
        string filename = CheckFileExists();
        File.AppendAllText(filename, $"{DateTime.Now} started.{Environment.NewLine}");
    }

    protected override void OnStop()
    {
        string filename = CheckFileExists();
        File.AppendAllText(filename, $"{DateTime.Now} stopped.{Environment.NewLine}");
    }

    private static string CheckFileExists()
    {
        string filename = @"c:\tmp\MyService.txt";
        if (!File.Exists(filename))
        {
            File.Create(filename);
        }

        return filename;
    }

}

Not exactly a complicated service, I’ll grant you.

Installing

For Framework apps, you could use InstallUtil, but if you try that on a Core app, you get an annoyingly vague error! Instead, you need to find the place where the binary has been compiled; for example:

C:\Users\pcmic\source\repos\ConsoleApp3\ConsoleApp3\bin\Debug\netcoreapp2.2

Now, launch a command prompt as admin, and type the following:

>sc create [service name] binpath=[full path to binary]

For example:

>sc create pcmtestservice binpath=C:\Users\pcmic\source\repos\ConsoleApp3\ConsoleApp3\bin\Debug\netcoreapp2.2\ConsoleApp3.exe

You should get the response:

[SC] CreateService SUCCESS

You can then either start the service from here:

>sc start pcmtestservice

Or locate it in the services utility and start it from there. You should now be able to start and stop the service and see it logging the events as you do so.

If you need to remove the service, use:

>sc delete pcmtestservice

References

https://stackoverflow.com/questions/7764088/net-console-application-as-windows-service

Short Walks – MakeCert is dead – long live New-SelfSignedCertificate

If you wanted to produce a self-signed certificate, the way to do this used to be makecert. It was part of the Windows SDK. Since Microsoft removed the Visual Studio Command Prompt (not sure why), you would run it like so:

"C:\Program
Files (x86)\Microsoft Visual
Studio\2017\Professional\Common7\Tools\vsdevcmd"

And then you could call MakeCert; for example:

makecert -r -pe
"CN=testcert" -b 12/12/2018 -e 12/12/2021 -sky signature -a sha256
-len 2048 -ss my -sr LocalMachine

However, one thing that you’ll notice now if you visit the MakeCert page is that it’s deprecated. The new way is a Powershell command; which can be used like so:

New-SelfSignedCertificate -Subject "CN=testcert" -KeySpec "Signature" -CertStoreLocation "Cert:\CurrentUser\My"

Short Walks – 406 Error While Creating a .Net Core Api

I’m afraid this is another of those: “What has AddMvc()” ever done for us posts.  Today, I was creating a brand new .Net Core 2.1 Api and, instead of calling AddMvc in ConfigureServices, I instead used AddMvcCore:

public void ConfigureServices(IServiceCollection services)
{
    services.AddMvcCore();

When I tried to access the API, I got a 406 error.  The fix is very simple (and one that using AddMvc() does for you:

public void ConfigureServices(IServiceCollection services)
{
    services.AddMvcCore()
            .AddJsonFormatters();