ReturnTaskDirectlyAnalyzer

This Roslyn Analyzer warns developers about using the async/await pattern in a way where, instead of awaiting a Task, it could be returned directly.
This is better for many reasons, first and foremost because the compiler doesn't need to generate a state machine for the async method during compile time and awaiting the resulting Task is deferred to the caller.
Here's a few methods where this analyzer kicks in:

public async Task RunAsync() {
    await Task.Delay(1000);
}

public async Task RunAsync() {
    var random = new Random().Next();
    if(random % 2 == 0) {
        await Task.Delay(1000).ConfigureAwait(false);
        return;
    }
    
    await DoSomethingAsync(); 
}

public async Task<int> RunAsync() {
    var random = new Random().Next();
    if(random % 2 == 0) {
        return await GetSomeValueAsync();
    }
    
    return await GetSomeOtherValueAsync();
}

After applying the code fix which the analyzer supplies, these methods would be transformed to the following:

public Task RunAsync() {
    return Task.Delay(1000);
}

public Task RunAsync() {
    var random = new Random().Next();
    if(random % 2 == 0) {
        return Task.Delay(1000);
    }
    
    return DoSomethingAsync(); 
}

public Task<int> RunAsync() {
    var random = new Random().Next();
    if(random % 2 == 0) {
        return GetSomeValueAsync();
    }
    
    return GetSomeOtherValueAsync();
}

The analyzer will not report when a Task is awaited within a try or a using block.
To see a list of all supported cases, please have a look at the ReturnTaskDirectlyTests.cs file.

Installation

This analyzer is published in NuGet and can be installed using the following methods.
Using the .NET Core command-line interface:

dotnet add package ReturnTaskDirectlyAnalyzer

The NuGet package manager console:

Install-Package ReturnTaskDirectlyAnalyzer

Or by adding the PackageReference to your project manually and supplying the latest version:

<PackageReference Inlude="ReturnTaskDirectlyAnalyzer" Version="xxx" />

Contributing

If you identify a false positive (the analyzer reports an await when it shouldn't), please do the following:

1. Fork the repository (if you haven't already)
2. Create a new branch from master
3. Add a test demonstrating the false positive in the ShouldNotRaise section of the analyzer's tests.
4. Open a PR with the failing test which targets the master branch. I will then push a fix in your PR and complete it. If you know what the fix is and can supply it within the PR, even better :)

If you have identified a case where the analyzer does not report an await even though it should, follow the same steps as above, however add the test in the ShouldRaise section.

If you are ever not sure about the analyzer's behavior, are having trouble creating a test or have general questions, please feel free to open an issue.