Using C# Source Generators with Microsoft Feature Management Feature Flags

C# Source Generators allow you to generate and compile source code dynamically. You can read more about them in this previous article.

In my series on Microsoft Feature Management, part 2 showed how to reduce magic strings by using an enum.

Using this approach still requires that the enum and the configuration file be manually kept in sync.

One use case for C# Source Generators (that are currently in preview) is to generate code from a file in the project, for example generating a class from an XML settings file.

We can process any external file in a source generator, including json files such as appsettings.json.

This means that if the feature flags are defined in the appsettings.json file, we can create a source generator to read this file and output an enum containing all the features that have been defined in the appsettings.json. This means that the feature flags in the app will always be in sync with configuration file. It also means that if you were to remove a flag from the appsettings.json then compilation would break if you were still referencing the feature flag in the app.

Creating a C# Source Generator

To get this all to work in Visual Studio 2019 Preview first create a new .NET Standard 2.0 class library project called FeatureFlagGenerator.

Modify the FeatureFlagGenerator.csproj to the following:

<Project Sdk="Microsoft.NET.Sdk">
    <PropertyGroup>
        <TargetFramework>netstandard2.0</TargetFramework>
        <LangVersion>preview</LangVersion>
    </PropertyGroup>
    
    <PropertyGroup>
        <RestoreAdditionalProjectSources>https://pkgs.dev.azure.com/dnceng/public/_packaging/dotnet5/nuget/v3/index.json ;$(RestoreAdditionalProjectSources)</RestoreAdditionalProjectSources>
    </PropertyGroup>
    
    <ItemGroup>
        <PackageReference Include="Microsoft.CodeAnalysis.CSharp.Workspaces" Version="3.6.0-3.20207.2" PrivateAssets="all" />
        <PackageReference Include="Microsoft.CodeAnalysis.Analyzers" Version="3.0.0-beta2.final" PrivateAssets="all" />
    </ItemGroup>
</Project>

Next add a new class called FeatureFlagEnumGenerator with the following code:

using System.Collections.Generic;
using System.Linq;
using System.Text;
using Microsoft.CodeAnalysis;
using Microsoft.CodeAnalysis.Text;

namespace FeatureFlagGenerator
{
    [Generator]
    public class FeatureFlagEnumGenerator : ISourceGenerator
    {
        public void Execute(SourceGeneratorContext context)
        {
            IEnumerable<AdditionalText> files = context.AdditionalFiles.Where(at => at.Path.EndsWith("appsettings.json"));

            AdditionalText appsettings = files.First();

            StringBuilder enumString = new StringBuilder();

            enumString.AppendLine("namespace GeneratedCode");
            enumString.AppendLine("{");

            enumString.AppendLine("   public enum GeneratedFeatureFlags");
            enumString.AppendLine("   {");

            // Doing this manually because referencing JsonDocument etc causes unknown errors
            // Not very robust code, POC only
            bool isInFeatureManagamentSection = false;
            foreach (var line in appsettings.GetText(context.CancellationToken).Lines)
            {               
                if (isInFeatureManagamentSection && line.ToString().Contains("}"))
                {
                    isInFeatureManagamentSection = false;
                }

                if (isInFeatureManagamentSection)
                {
                    int featureNameStartIndex = line.ToString().IndexOf('"')+1;
                    int featureNameEndIndex = line.ToString().IndexOf(':')-1;
                    enumString.AppendLine(line.ToString().Substring(featureNameStartIndex, featureNameEndIndex - featureNameStartIndex) + ",");
                }

                if (line.ToString().Contains("FeatureManagement"))
                {
                    isInFeatureManagamentSection = true;
                }
            }

            enumString.AppendLine("   }");
            enumString.AppendLine("}");

            SourceText sourceText = SourceText.From(enumString.ToString(), Encoding.UTF8);
            const string desiredFileName = "GeneratedFeaturesEnum.cs";
            context.AddSource(desiredFileName, sourceText);
        }

        public void Initialize(InitializationContext context)
        {
            // Advanced usage
            
        }
    }
}

Essentially the preceding code will search for the appsettings.json file in the project where the source generator is being used and then find all the configured feature flag names - for some reason trying to use System.Text.Json was causing errors so I implemented a very clunky solution using string manipulation.

Using a C# Source Generator

Now the generator is defined, it can be used in another project.

Go and add a new ASP.NET Core MVC project to the solution and modify the .csproj file to the following:

<Project Sdk="Microsoft.NET.Sdk.Web">

  <PropertyGroup>
      <TargetFramework>net5.0</TargetFramework>
      <LangVersion>preview</LangVersion>
  </PropertyGroup>

  <ItemGroup>
    <AdditionalFiles Include="appsettings.json">
      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
    </AdditionalFiles>
  </ItemGroup>

  <ItemGroup>
    <PackageReference Include="Microsoft.FeatureManagement.AspNetCore" Version="2.0.0" />      
  </ItemGroup>

  <ItemGroup>
    <ProjectReference Include="..\FeatureFlagGenerator\FeatureFlagGenerator.csproj" OutputItemType="Analyzer" ReferenceOutputAssembly="false" />
  </ItemGroup>

</Project>

There is a secret ingredient in the preceding project file and that is the AdditionalFiles Include="appsettings.json" part. This makes the file available to C# Source Generators and took me a while to work out.

Now when the web application is built, the source generator will run and create an enum with every configured flag name.

For example given the following appsettings.json:

{
  "Logging": {
    "LogLevel": {
      "Default": "Information",
      "Microsoft": "Warning",
      "Microsoft.Hosting.Lifetime": "Information"
    }
  },
  "AllowedHosts": "*",
  "FeatureManagement": {
    "Printing": true,
    "LiveChat": true
  }
}

This will generate and compile the following into the web application dll:

namespace GeneratedCode
{
  public enum GeneratedFeatureFlags
  {
    Printing,
    LiveChat,
  }
}

Notice the Printing and LiveChat enum values come from the appsettings FeatureManagement section.

Now the feature flags can be referenced in code:

if (await _featureManager.IsEnabledAsync(nameof(GeneratedCode.GeneratedFeatureFlags.Printing)))
{
    ViewData["PrintMessage"] = "On";
}
else
{
    ViewData["PrintMessage"] = "Off";
}

Or in the UI:

<feature name="@nameof(GeneratedCode.GeneratedFeatureFlags.Printing)">
    <li class="nav-item">
        <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Print">Print Preview</a>
    </li>
</feature>

It is these kinds of use cases, where there would otherwise be manual boilerplate code work, that C# Source Generators will be really useful.

SHARE:

Reducing Magic Strings with Microsoft Feature Toggles (Microsoft.FeatureManagement)

This is the second part in a series.

One of the downsides of Microsoft’s feature flags/feature toggle implementation is the reliance on magic strings in code to query the status of features.

For example, the following code checks for the state of the Printing feature:

if (await _featureManager.IsEnabledAsync("Printing"))
{
    ViewData["PrintMessage"] = "On";
}
else
{
    ViewData["PrintMessage"] = "Off";
}

Notice in the preceding code that the feature is referenced with the string “Printing”. These kind of magic strings in code can be problematic because they have no sematic compiler meaning, for example if the feature is mentioned in multiple places and the name of the feature changes in configuration you would need to update all the strings in the codebase. It is easy to miss one and then end up with errors in the app.

In the documentation, the recommended approach is to define an enum containing the feature you want to control, for example:

public enum Features
{
    Printing,
    QuickQuotes,
    OnlineChat
}

Now the code can be refactored to make use of the enum in conjunction with a nameof expression:

if (await _featureManager.IsEnabledAsync(nameof(Features.Printing)))
{
    ViewData["PrintMessage"] = "On";
}
else
{
    ViewData["PrintMessage"] = "Off";
}

There are a number of things that can (and should) happen when using feature toggles/flags. One of which is that when a feature becomes permanent then the toggle and associated code should be removed. First this means removing it from configuration, which as I mentioned in part 1 will not actually break anything. The next step will be to remove it from the enum, doing this will actually break compilation (which is a good thing) if the enum value is referenced in a nameof expression. The next step is to fix compilation errors by removing the conditional code that relies on the flag/toggle.

In part 1 I mentioned the <feature> tag helper that can be used to enable/disable UI elements based on a flag:

<feature name="Printing">
    <li class="nav-item">
        <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Print">Print Preview</a>
    </li>
</feature>

Notice in the preceding HTML that the name of the feature is still represented as a magic string “Printing”. If the enum value and config are removed, compilation will not beak if this HTML still exists. We can fix this by one again using nameof in the view:

<feature name="@nameof(Features.Printing)">
    <li class="nav-item">
        <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Print">Print Preview</a>
    </li>
</feature>

Now if we removed Printing from the enum we get a compilation error and we won’t accidentally release an app to production with incorrect UI elements.

SHARE:

Using the Microsoft Feature Toggle Library in ASP.NET Core (Microsoft.FeatureManagement)

This is the first part in a series.

As the creator of the open source FeatureToggle library for .NET it was with some interest that I learned about Microsoft’s own offering to the feature toggle/feature flags landscape.

Microsoft’s offering is know as “.NET Core Feature Management” and they use the term “feature flag” in place of other terms such as “feature toggle”, “feature flappers”, etc. The root namespace is “Microsoft.FeatureManagement”.

Getting Started with Microsoft Feature Management in ASP.NET Core

As a simple example, first create a new ASP.NET Core 3.1 web app project with MVC support and then install the Microsoft.FeatureManagement.AspNetCore NuGet package.

Next, open the Startup.cs class, add a using directive for Microsoft.FeatureManagement and opt in to feature management:

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

    services.AddFeatureManagement();
}

Feature flags can be configured in a number of ways, such as in appsettings.json, environment variables, and Azure App Configuration.

Open up the appsettings.json file and add a feature to be managed in a new FeatureManagement section, for example the following defines a feature called “Printing”:

{
  "Logging": {
    "LogLevel": {
      "Default": "Information",
      "Microsoft": "Warning",
      "Microsoft.Hosting.Lifetime": "Information"
    }
  },
  "AllowedHosts": "*",
  "FeatureManagement": {
    "Printing": true    
  }
}

In the preceding json, the Printing feature is currently enabled.

Now we have a configured feature flag/toggle we can make use of it.

Programmatically Querying Feature Flags in Controllers

If you want to execute conditional logic in a controller based on a feature flag you can do this by first getting a reference to an IFeatureManager. To do this, it can be injected as a dependency into the controller via its constructor:

private readonly IFeatureManager _featureManager;

public HomeController(IFeatureManager featureManager)
{
    _featureManager = featureManager;
}

Now the IFeatureManager instance can be used to make decisions by calling the IsEnabledAsync method and providing the string of the feature that was configured in the appsettings.json – in this case “Printing”:

public async Task<IActionResult> Index()
{
    if (await _featureManager.IsEnabledAsync("Printing"))
    {
        ViewData["PrintMessage"] = "On";
    }
    else
    {
        ViewData["PrintMessage"] = "Off";
    }
    
    return View();
}

In the view you could have some HTML to output the contents of the PrintMessage viewdata:

@{
    ViewData["Title"] = "Home Page";
}

<div class="text-center">
    <h1 class="display-4">Welcome</h1>
    <p>Printing is currently @ViewData["PrintMessage"]</p>
</div>

In the controller you could of course do more complex logic such as calling additional services or swapping out different algorithms based on flags.

One thing I don’t like about the FeatureManagement library is that if a flag is not configured or has a typo in the configuration, then the app still runs with the feature disabled. By default I think that the app should error if  a flag is used but not defined, otherwise you may have important features/logic that does not get executed; I feel that the sooner you know about these kind of problems the better. For example some important legal/compliance text might not be shown when it should.

Managing Controllers and Actions Based On Feature Flags

Controller actions (and also entire controllers) can be enabled/disabled based on a feature flag. For example an action could be based on the printing flag by decorating the action method with the [FeatureGate] attribute as follows:

[FeatureGate("Printing")]
public IActionResult Print()
{
    return View();
}

If you try and navigate to the Print page when the flag is disabled then you’ll get a 404.

You can also apply the [FeatureGate] attribute at the controller class level and this will affect all actions contained therein.

Modifying The Generated HTML View Based On Flags

Sometimes you will have UI elements that should be shown or hidden based on a flag.

To control an entire block of HTML you can surround it with the <feature> tag. To enable this you should first modify _ViewImports.cshtml and add the tag helper as shown below:

@using WebApplication1
@using WebApplication1.Models
@addTagHelper *, Microsoft.AspNetCore.Mvc.TagHelpers
@addTagHelper *, Microsoft.FeatureManagement.AspNetCore

As an example, suppose we have disabled the Print page/action – we would also not want it to be displayed in the menu. We could modify the HTML to make the menu item display conditional on the Printing flag:

<nav class="navbar navbar-expand-sm navbar-toggleable-sm navbar-light bg-white border-bottom box-shadow mb-3">
    <div class="container">
        <a class="navbar-brand" asp-area="" asp-controller="Home" asp-action="Index">WebApplication1</a>
        <button class="navbar-toggler" type="button" data-toggle="collapse" data-target=".navbar-collapse" aria-controls="navbarSupportedContent"
                aria-expanded="false" aria-label="Toggle navigation">
            <span class="navbar-toggler-icon"></span>
        </button>
        <div class="navbar-collapse collapse d-sm-inline-flex flex-sm-row-reverse">
            <ul class="navbar-nav flex-grow-1">
                <li class="nav-item">
                    <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Index">Home</a>
                </li>
                <li class="nav-item">
                    <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Privacy">Privacy</a>
                </li>
                <feature name="Printing">
                    <li class="nav-item">
                        <a class="nav-link text-dark" asp-area="" asp-controller="Home" asp-action="Print">Print Preview</a>
                    </li>
                </feature>
            </ul>
        </div>
    </div>
</nav>

Notice in the preceding HTML that the <feature name=”Printing”> element is wrapped around the printing menu option <li>.

Now if the Printing flag is set to true the menu item will be shown, otherwise it won’t appear in the rendered HTML.

SHARE:

Pretty Method Display in xUnit.net

One little-known feature of the xUnit.net testing framework is the ability to write test method names in a specific way and then have them converted to a ‘pretty’ version for example in Visual Studio Test Explorer.

Take the following test method:

using ClassLibrary1;
using Xunit;

namespace XUnitTestProject2
{
    public class CalculatorShould
    {
        [Fact]
        public void Add2PositiveNumbers()
        {
            var sut = new Calculator();

            sut.Add(1);
            sut.Add(1);

            Assert.Equal(2, sut.Value);
        }
    }
}

By default, this will look like the following screenshot in Visual Studio Test Explorer:

Default xUnit.net Test Method Name Display

The first thing that can be modified to to simplify the test method name display to only display the test method name and not the preceding namespace and class name, for example “XUnitTestProject2.CalculatorShould.Add2PositiveNumbers” becomes more simply “Add2PositiveNumbers” by making a simple configuration change.

Displaying Only Test Method Names in xUnit.net Tests

To control the rendering of method names in xUnit.net, the first thing to do is add a new file called “xunit.runner.json” to the root of the test project and set the Copy To Output Directory property to Copy if newer. This will make this file copy to the output bin directory. Once this is done, if you open the project file you should see something like:

<ItemGroup>
  <None Update="xunit.runner.json">
    <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
  </None>
</ItemGroup>

Next, modify the json file to the following:

{
  "$schema": "https://xunit.net/schema/current/xunit.runner.schema.json",
   "methodDisplay": "method"
}

Notice in the preceding json configuration the methodDisplay has been set to “method”, this will prevent the namespace and class being prepended to the method name in Test Explorer.

Now if you head back to Test Explorer you should see the following:

Method name only display in xUnit.net tests

Enabling Pretty Method Names in xUnit.net

In addition to shortening test method name display we can also make use of xUnit.net’s “pretty method display”.

To enable this feature, modify the json configuration file and add the "methodDisplayOptions": "all" configuration as follows:

{
  "$schema": "https://xunit.net/schema/current/xunit.runner.schema.json",
  "methodDisplay": "method",
  "methodDisplayOptions": "all"
}

Now the previous test can be renamed to “Add_2_positive_numbers” as follows:

[Fact]
public void Add_2_positive_numbers()
{
    var sut = new Calculator();

    sut.Add(1);
    sut.Add(1);

    Assert.Equal(2, sut.Value);
}

In test explorer this test method will show up as “Add 2 positive numbers” as the following screenshot shows:

xUnit.net pretty method display names

You can use other items in the test method name, for example you can use the monikers eq, ne, lt, le, gt, ge that get replaced with =, !=, <, <=, >, >= respectively, for example a test name of “Have_a_value_eq_0_when_multiplied_by_zero” would be displayed as “Have a value = 0 when multiplied by zero”. Here the eq has been replaced with =.

You can also use ASCII or Unicode escape sequences, for example the test name “Divide_by_U00BD” gets displayed as “Divide by ½” and the test “Email_address_should_only_contain_a_single_U0040” gets displayed as “Email address should only contain a single @”, or “The_U2211_of_1U002C_2_and_3_should_be_6” becomes “The ∑ of 1, 2 and 3 should be 6”:

xUnit Pretty methods

You could also combine the "methodDisplay": "classAndMethod" to create something like  and the following:

namespace Given_a_cleared_calculator
{
    public class when_a_number_gt_0_is_added
    {
        [Fact]
        public void then_the_value_should_be_gt_0()
        {
            // etc.
        }

        [Fact]
        public void then_the_value_should_eq_the_one_added()
        {
            // etc.
        }
    }
}

This would produce the following tests in Test Explorer:

xUnit.net Pretty Display Names

If you want to learn more about writing tests with xUnit.net check out my Pluralsight course today.

SHARE:

Simplifying Parameter Null and Other Checks with the Pitcher Library

In a previous post I looked at the GaurdClauses library that can simplify the usual guard checks we sometimes need to write. In the comments someone mentioned the Pitcher library that accomplishes the same thing so I thought I’d check it out here.

First, the NuGet package needs to be installed and a using Pitcher; directive added, then we can make use of the library.

As an example, without the library you might end up with some code like the following:

public static void AddNewPerson(string name, int ageInYears)
{
    if (string.IsNullOrWhiteSpace(name))
    {
        throw new ArgumentException($"Cannot be null, empty, or contain only whitespace.", nameof(name));
    }

    if (ageInYears < 1)
    {
        throw new ArgumentOutOfRangeException(nameof(ageInYears), "Must be greater than zero.");
    }

    // Add to database etc.
}

This is just some boilerplate type code to check for null/empty strings and that the age that’s passed in is positive.

With the Pitcher library this could be refactored to the following (I’ve left the GaurdClauses code commented as a comparison):

public static void AddNewPerson(string name, int ageInYears)
{
    // GuardClauses version:
    // Guard.Against.NullOrWhiteSpace(name, nameof(name));
    // Guard.Against.NegativeOrZero(ageInYears, nameof(ageInYears));


    // Pitcher version:
    Throw.ArgumentNull.WhenNullOrWhiteSpace(name, nameof(name));

    Throw.ArgumentOutOfRange.WhenLessThan(ageInYears, mustBeMoreThan:0, nameof(ageInYears));
    // or more simply:
    Throw.ArgumentOutOfRange.WhenNegativeNumber(ageInYears, nameof(ageInYears));            
}

Pitcher allows you to throw either an ArgumentOutOfRangeException or an ArgumentNullException when using this syntax and also provides ways to throw other exception types when a given condition is true, for example:

Throw.When(ageInYears == 42, new InvalidOperationException("This age has no meaning"));

You can find the source code and more examples on GitHub and don’t forget to say hi/thanks to the project maintainer Alex Kamsteeg and tell him you heard about Pitcher here :)

SHARE:

New Pluralsight Course: Creating Automated Browser Tests with Selenium in C#

My newest Pluralsight course was just published and you can start watching today. Selenium is a tool that allows you to automate a web browser and simulate an end-user interacting with your web app. You can combine Selenium with a test framework such as xUnit.net to create tests that check your web app is working as expected.

Automated browser tests can compliment your other types of tests such as unit and integration tests.

From the course description: “Unit and integration tests can help you catch a range of bugs, but not all of them. Even if your unit and integration tests pass, you could still deploy your web app to production and find it doesn’t work as expected. In this course, Creating Automated Browser Tests with Selenium in C#, you will gain the ability to create tests that automate the browser and simulate a real person using your web app. First, you will learn how to set up your test project and write your first test. Next, you will discover how to interact with web page elements from your tests, such as clicking a button or typing text. Finally, you will explore how to create a suite of automated web tests that are easier to maintain over time. When you are finished with this course, you will have the skills and knowledge of Selenium automated browser testing needed to help ensure your web app is working as expected before you release it to production.”

Check out the course today and if you’re not a Pluralsight member you can currently start watching for free with a Pluralsight Free Trial with Unlimited Access .

SHARE:

Adding Tuple Support to .NET Classes in C#

Edit: Updated to improve clarity (thanks to Paulo in the comments for helping to improve his article).

Tuples in C# are objects that can be created with a specific syntax. You don’t have to declare tuple types first like you do with classes for example, they can instead be created using a lightweight C# syntax.

A tuple is a object that holds a number of arbitrary data items and which has no custom behaviour. In contrast, a class or struct can have both data and custom behaviour.

For example the following creates a tuple with 2 string values:

(string, string) names = ("Sarah", "Smith");
Console.WriteLine($"First name: '{names.Item1}' Last name: '{names.Item2}'");

This code produces the output: First name: 'Sarah' Last name: 'Smith'

In the preceding code, the items inside the tuple don’t have names so they are referred to as Item1 and Item2 but you could also name the items, for example:

(string firstName, string lastName) names = ("Sarah", "Smith");
Console.WriteLine($"First name: '{names.firstName}' Last name: '{names.lastName}'");

If you had a rich Person class that had both data and behaviour, you could also add support for tuple-like deconstruction and unpackaging of a Person instance into variables just like you would do with a tuple instance.

Consider the following class:

class Person
{
    public string FirstName { get; set; }
    public string LastName { get; set; }
    public int AgeInYears { get; set; }
    public string FavoriteColor { get; set; }
    
    // methods etc.
}

We could create a tuple as before containing the first and last name as follows:

var sarah = new Person
{
    FirstName = "Sarah",
    LastName = "Smith",
    AgeInYears = 42,
    FavoriteColor = "red"
};

(string firstName, string lastName) names = (sarah.FirstName, sarah.LastName);
Console.WriteLine($"First name: '{names.firstName}' Last name: '{names.lastName}'");

This is however a little clunky, we can modify the Person class to provide support for a Person to have tuple-like deconstruction and unpacking semantics. To do this a public void method called Deconstruct can be added, for example:

class Person
{
    public string FirstName { get; set; }
    public string LastName { get; set; }
    public int AgeInYears { get; set; }
    public string FavoriteColor { get; set; }

    // methods etc.

    public void Deconstruct(out string firstName, out string lastName)
    {
        firstName = FirstName;
        lastName = LastName;
    }
}

Now the code could be changed to:

var (firstName, lastName) = sarah;
Console.WriteLine($"First name: '{firstName}' Last name: '{lastName}'");

You could also add this deconstruction/unpackaging support to a class you can’t change by declaring an extension method such as:

static class PersonExtensions
{
    public static void Deconstruct(this Person person, out string firstName, out string lastName)
    {
        firstName = person.FirstName;
        lastName = person.LastName;
    }
}

Or as another example, you could add tuple-like deconstruction & unpackaging support for the .NET String type:

static class StringExtensions
{
    public static void Deconstruct(this string s, out string original, out string upper, out string lower, out int length)
    {
        original = s;
        upper = s.ToUpperInvariant();
        lower = s.ToLowerInvariant();
        length = s.Length;
    }
}

And then write:

var (original, upper, lower, length) = "The quick brown fox";
Console.WriteLine($"Original: {original}");
Console.WriteLine($"Uppercase: {upper}");
Console.WriteLine($"Lowercase: {lower}");
Console.WriteLine($"Length: {length}");

As Paulo points out in the comments there is no actual tuple instance per-se involved here, if look at the the decompiled source that Paulo links to you can see the Person has been unpackaged into multiple variables.

If you want to learn a load more C# tips check out my C# Tips and Traps course today. You can even currently start watching with a Pluralsight Free Trial with Unlimited Access .

SHARE:

Variables? We Don’t Need No Stinking Variables - C# Discards

C# 7.0 introduced the concept of discards. Discards are intentionally unused, temporarily dummy variables that we don’t care about and don’t want to use.

For example, the following shows the result of an addition being discarded:

_ = 1 + 1;

Note the underscore _ this is the discard character.

Given the preceding example, you cannot access the result of this addition, for example:

WriteLine(_); // Error CS0103  The name '_' does not exist in the current context 

Using C# Discards with Out Parameters

A more useful example is when you are working with a method that has one or more out parameters and you don’t care about using the outputted value.

As an example, consider one of the many TryParse methods in .NET such as int.TryParse. The following code show a method that writes to the console whether or not a string can be parsed as an int:

static void ParseInt()
{
    WriteLine("Please enter an int to validate");
    string @int = ReadLine();
    bool isValidInt = int.TryParse(@int, out int parsedInt);
    
    if (isValidInt)
    {
        WriteLine($"{@int} is a valid int");
    }
    else
    {
        WriteLine($"{@int} is NOT a valid int");
    }
}

The preceding method can be written using a discard because the out int parsedInt value is never used:

static void ParseIntUsingDiscard()
{
    WriteLine("Please enter an int to validate");
    string @int = ReadLine();

    if (int.TryParse(@int, out _))
    {
        WriteLine($"{@int} is a valid int");
    }
    else
    {
        WriteLine($"{@int} is NOT a valid int");
    }
}

For example we could create an expression bodied method using a similar approach:

static bool IsInt(string @int) => int.TryParse(@int, out _);

If you have a method that returns a lot of out values such as:

private static void GenerateDefaultCity(out string name, out string nickName, out long population, out DateTime founded)
{
    name = "London";
    nickName = "The Big Smoke";
    population = 8_000_000;
    founded = new DateTime(50, 1, 1);
}

In this case you might only care about the returned population value so you could discard all the other out values:

GenerateDefaultCity(out _,out _, out var population, out _);
WriteLine($"Population is: {population}");

Using C# Discards with Tuples

Another use for discards is where you don’t care about all the fields of a tuple. For example the following method returns a tuple containing a name and age:

static (string name, int age) GenerateDefaultPerson()
{
    return ("Amrit", 42);
}

If you only cared about the age you could write:

var (_, age) = GenerateDefaultPerson();
WriteLine($"Default person age is {age}");

Simplifying Null Checking Code with Discards

Take the following null checking code:

private static void Display(string message)
{
    if (message is null)
    {
        throw new ArgumentNullException(nameof(message));
    }
    WriteLine(message);
}

You could refactor this to make use of throw expressions:

private static void DisplayV2(string message)
{
    string checkedMessage = message ?? throw new ArgumentNullException(nameof(message));

    WriteLine(checkedMessage);
}

In the preceding version however, the checkedMessage variable is somewhat redundant, this could be refactored to use a discard:

private static void DisplayWithDiscardNullCheck(string message)
{
    _ = message ?? throw new ArgumentNullException(nameof(message));
    
    WriteLine(message);
}

Using C# Discards with Tasks

Take the following code:

// Warning CS1998  This async method lacks 'await' operators and will run synchronously.
Task.Run(() => SayHello());

Where the SayHello method is defined as:

private static string SayHello()
{
    string greeting = "Hello there!";
    return greeting;
}

If we don’t care about the return value and want to discard the result and get rid of the compiler warning::

// With discard - no compiler warning
_ = Task.Run(() => SayHello());

If there are any exceptions however, they will be supressed:

await Task.Run(() => throw new Exception()); // Exception thrown
_ = Task.Run(() => throw new Exception()); // Exception suppressed

Pattern Matching with Switch Statements and Discards

You can also use discards in switch statements:

private static void SwitchExample(object o)
{
    switch (o)
    {
        case null:
            WriteLine("o is null");
            break;
        case string s:
            WriteLine($"{s} in uppercase is {s.ToUpperInvariant()}");
            break;
        case var _:
            WriteLine($"{o.GetType()} type not supported.");
            break;
    }
}

If you want to learn a load more C# tips check out my C# Tips and Traps course today. You can even currently start watching with a Pluralsight Free Trial with Unlimited Access .

SHARE:

Simplifying Parameter Null and Other Checks with the GuardClauses Library

Often you want to add null checking and other check code at the start of a method to ensure all the values passed into the method are valid before continuing.

For example the following method checks the name and age:

public static void AddNewPerson(string name, int ageInYears)
{
    if (string.IsNullOrWhiteSpace(name))
    {
        throw new ArgumentException($"Cannot be null, empty, or contain only whitespace.", nameof(name));
    }

    if (ageInYears < 1)
    {
        throw new ArgumentOutOfRangeException(nameof(ageInYears), "Must be greater than zero.");
    }

    // Add to database etc.
}

This “guard” kind of code can “clutter” the method and reduce readability.

One library I recently came across is the Guard Clauses library from Steve Smith.

Once this library is installed we could refactor the preceding code to look like the following:

public static void AddNewPerson(string name, int ageInYears)
{
    Guard.Against.NullOrWhiteSpace(name, nameof(name));
    Guard.Against.NegativeOrZero(ageInYears, nameof(ageInYears));

    // Add to database etc.
}

Passing a null name results in the exception: System.ArgumentNullException: Value cannot be null. (Parameter 'name')

Passing an empty string results in: System.ArgumentException: Required input name was empty. (Parameter 'name')

Passing in an age of zero results in: System.ArgumentException: Required input ageInYears cannot be zero or negative. (Parameter 'ageInYears')

The code is also more readable and succinct.

Out of the box the library comes with the following guards (taken from the documentation):

  • Guard.Against.Null (throws if input is null)
  • Guard.Against.NullOrEmpty (throws if string or array input is null or empty)
  • Guard.Against.NullOrWhiteSpace (throws if string input is null, empty or whitespace)
  • Guard.Against.OutOfRange (throws if integer/DateTime/enum input is outside a provided range)
  • Guard.Against.OutOfSQLDateRange (throws if DateTime input is outside the valid range of SQL Server DateTime values)
  • Guard.Against.Zero (throws if number input is zero)

You can also define your own reusable clauses:

// Define in this namespace so can use alongside built-in guards with no additional namespaces required
namespace Ardalis.GuardClauses
{
    public static class PositiveGuard
    {
        public static void Positive(this IGuardClause guardClause, int input, string parameterName)
        {
            if (input >= 0)
            {
                throw new ArgumentOutOfRangeException(parameterName, $"Required input {parameterName} cannot be positive.");
            }                           
        }
    }
}

And then in a method we can write:

public static void ReportNegativeTemperature(int temp)
{
    Guard.Against.Positive(temp, nameof(temp));
    // Do something
}

And if we pass a positive (or zero) temp we get: System.ArgumentOutOfRangeException: Required input temp cannot be positive. (Parameter 'temp')

This is one of those simple libraries that can make basic tasks easier/more readable.

If you check this out and use it make sure you say thanks to Steve on Twitter and let him know @robertsjason sent you ;)

SHARE:

Writing Azure Functions with Function Monkey: Using Commands Without Handlers

If you’ve read the previous articles on Function Monkey you may be wondering if you always need a command handler. Sometimes you may want to accept a request into the system (for example via HTTP) and that pass that request off for further processing. For example the HTTP data can be accepted and then the data (“command”) can be put on a queue for processing. This allows the potential scale-out of the function that processes queue messages to improve the overall throughput of the system.

Take the following example that allows an invoice to be submitted via HTTP. The submitted invoice is validated before simply being returned from the SubmitInvoiceCommandHandler. The output of the handler gets sent to a storage queue called “invoices”. Then we have a queue storage trigger creating and handling the ProcessInvoiceCommand.

using System.Net.Http;
using System.Threading.Tasks;
using AzureFromTheTrenches.Commanding.Abstractions;
using FluentValidation;
using FunctionMonkey.Abstractions;
using FunctionMonkey.Abstractions.Builders;
using FunctionMonkey.FluentValidation;
using Microsoft.Extensions.DependencyInjection;
using Microsoft.Extensions.Logging;

namespace FunctionApp2
{
    public class SubmitInvoiceCommand : ICommand<SubmitInvoiceCommand>
    {
        public string Description { get; set; }
        public decimal Amount { get; set; }
    }

    public class SubmitInvoiceCommandValidator : AbstractValidator<SubmitInvoiceCommand>
    {
        public SubmitInvoiceCommandValidator()
        {
            RuleFor(x => x.Description).NotEmpty();
            RuleFor(x => x.Amount).GreaterThan(0);
        }
    }

    public class SubmitInvoiceCommandHandler : ICommandHandler<SubmitInvoiceCommand, SubmitInvoiceCommand>
    {
        public Task<SubmitInvoiceCommand> ExecuteAsync(SubmitInvoiceCommand command, SubmitInvoiceCommand previousResult)
        {
            // We are not actually "handling" anything here, the handler is just returning the sam command
            return Task.FromResult(command);
        }
    }

    public class ProcessInvoiceCommand : ICommand
    {
        public string Description { get; set; }
        public decimal Amount { get; set; }
    }

    public class ProcessInvoiceCommandHandler : ICommandHandler<ProcessInvoiceCommand>
    {
        private readonly ILogger Log;

        public ProcessInvoiceCommandHandler(ILogger log)
        {
            Log = log;
        }

        public Task ExecuteAsync(ProcessInvoiceCommand command)
        {
            Log.LogInformation($"Processing invoice {command.Description} {command.Amount}");
            return Task.CompletedTask;
        }
    }

    public class FunctionAppConfiguration : IFunctionAppConfiguration
    {
        public void Build(IFunctionHostBuilder builder)
        {
            builder
                .Setup((serviceCollection, commandRegistry) =>
                {
                    serviceCollection.AddTransient<IValidator<SubmitInvoiceCommand>, SubmitInvoiceCommandValidator>();
                    commandRegistry.Register<SubmitInvoiceCommandHandler>();
                    commandRegistry.Register<ProcessInvoiceCommandHandler>();
                })
                .AddFluentValidation()
                .Functions(functions => functions

                    .HttpRoute("v1/SubmitInvoice", route => route
                        .HttpFunction<SubmitInvoiceCommand>(HttpMethod.Post)
                        .OutputTo.StorageQueue("invoices"))

                    .Storage(storage => storage
                        .QueueFunction<ProcessInvoiceCommand>("invoices"))                    
                );
        }
    }
}

If we POST the JSON { "Description" : "NAS",    "Amount" : 1000 } we get the following (abridged) output:

Executing HTTP request: {"method": "POST",  "uri": "/api/v1/SubmitInvoice"}
Executing 'SubmitInvoice' 
Executed 'SubmitInvoice'
Executing 'StqFnProcessInvoice' (Reason='New queue message detected on 'invoices')
Storage queue trigger function StqFnProcessInvoice processed a request.
Processing invoice NAS 1000.0
Executed 'StqFnProcessInvoice' 

At the moment the SubmitInvoiceCommandHandler is not doing anything useful, it’s just passing the command back out so it can be output to queue storage.

With Function Monkey you can do away with the command handler in these cases.

One way to do this is when configuring the function app by adding the NoCommandHandler() option in the build method. This means that the SubmitInvoiceCommandHandler class can be deleted:

using System.Net.Http;
using System.Threading.Tasks;
using AzureFromTheTrenches.Commanding.Abstractions;
using FluentValidation;
using FunctionMonkey.Abstractions;
using FunctionMonkey.Abstractions.Builders;
using FunctionMonkey.FluentValidation;
using Microsoft.Extensions.DependencyInjection;
using Microsoft.Extensions.Logging;

namespace FunctionApp2
{
    public class SubmitInvoiceCommand : ICommand<SubmitInvoiceCommand>
    {
        public string Description { get; set; }
        public decimal Amount { get; set; }
    }

    public class SubmitInvoiceCommandValidator : AbstractValidator<SubmitInvoiceCommand>
    {
        public SubmitInvoiceCommandValidator()
        {
            RuleFor(x => x.Description).NotEmpty();
            RuleFor(x => x.Amount).GreaterThan(0);
        }
    }

    public class ProcessInvoiceCommand : ICommand
    {
        public string Description { get; set; }
        public decimal Amount { get; set; }
    }

    public class ProcessInvoiceCommandHandler : ICommandHandler<ProcessInvoiceCommand>
    {
        private readonly ILogger Log;

        public ProcessInvoiceCommandHandler(ILogger log)
        {
            Log = log;
        }

        public Task ExecuteAsync(ProcessInvoiceCommand command)
        {
            Log.LogInformation($"Processing invoice {command.Description} {command.Amount}");
            return Task.CompletedTask;
        }
    }

    public class FunctionAppConfiguration : IFunctionAppConfiguration
    {
        public void Build(IFunctionHostBuilder builder)
        {
            builder
                .Setup((serviceCollection, commandRegistry) =>
                {
                    serviceCollection.AddTransient<IValidator<SubmitInvoiceCommand>, SubmitInvoiceCommandValidator>();
                    commandRegistry.Register<ProcessInvoiceCommandHandler>();
                })
                .AddFluentValidation()
                .Functions(functions => functions

                    .HttpRoute("v1/SubmitInvoice", route => route
                        .HttpFunction<SubmitInvoiceCommand>(HttpMethod.Post)
                        .Options(options => options.NoCommandHandler())
                        .OutputTo.StorageQueue("invoices"))

                    .Storage(storage => storage
                        .QueueFunction<ProcessInvoiceCommand>("invoices"))                    
                );
        }
    }
}

If we submit the same JSON request we get:

Executing HTTP request: {  "method": "POST",  "uri": "/api/v1/SubmitInvoice"}
Executing 'SubmitInvoice' 
Executed 'SubmitInvoice' 
Executing 'StqFnProcessInvoice' 
Storage queue trigger function StqFnProcessInvoice processed a request.
Processing invoice NAS 1000.0
Executed 'StqFnProcessInvoice'

Even though we don’t have an explicit handler now for the SubmitInvoiceCommand, the validation still takes place.

Another option is to implement the marker interface ICommandWithNoHandler in the command and then you don’t need the .NoCommandHandler() option.

Other Function Monkey articles:

SHARE: