Avoid allocations when chaining contexts

[jnyrup]

ba6028b:
#2607 added chaining of nested contexts.

It always creates a new Lazy<string> but we do allow a null Context

fluentassertions/Src/FluentAssertions/Execution/AssertionScope.cs

Line 316 in e760ba4

var identifier = Context?.Value;

So if both context are null, we can then then avoid creating a new context.

Also if one of the contexts are null, we can just return the other one.

Now we only need to create a new context in the last case where we have two non-null contexts.
As the lambda passed to the new Lazy<string> takes a closure over the other contexts, one needs to take some care to avoid always allocating the closure object.

• Dev: What's currently in develop
• Foo: The prettiest code using both switch expression and expression-bodied members.
• Bar: Similar to Foo, but avoids allocating lambda closure object unless needed using a local function.
• Baz: Just to show that the unwanted allocations in Foo are not due to switch expression not supporting block-bodies but due to Pay for an allocation/assignments of a capture class regardless of whether it's used dotnet/roslyn#20777

See how Bar is the only variant that is non-allocating when either contexts are null.

Method	a	b	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Dev	?	?	20.6564 ns	0.2046 ns	0.1814 ns	1.00	0.00	0.0268	168 B	1.00
Foo	?	?	3.0756 ns	0.0601 ns	0.0562 ns	0.15	0.00	0.0051	32 B	0.19
Bar	?	?	0.5815 ns	0.0201 ns	0.0178 ns	0.03	0.00	-	-	0.00
Baz	?	?	5.5472 ns	0.0305 ns	0.0255 ns	0.27	0.00	0.0051	32 B	0.19
Dev	?	inner	20.6674 ns	0.0672 ns	0.0561 ns	1.00	0.00	0.0268	168 B	1.00
Foo	?	inner	3.7963 ns	0.0971 ns	0.0909 ns	0.18	0.00	0.0051	32 B	0.19
Bar	?	inner	1.5008 ns	0.0229 ns	0.0191 ns	0.07	0.00	-	-	0.00
Baz	?	inner	6.3278 ns	0.0979 ns	0.0916 ns	0.31	0.00	0.0051	32 B	0.19
Dev	outer	?	21.0130 ns	0.1855 ns	0.1645 ns	1.00	0.00	0.0268	168 B	1.00
Foo	outer	?	3.8222 ns	0.0425 ns	0.0397 ns	0.18	0.00	0.0051	32 B	0.19
Bar	outer	?	1.4970 ns	0.0402 ns	0.0356 ns	0.07	0.00	-	-	0.00
Baz	outer	?	6.1180 ns	0.0834 ns	0.0740 ns	0.29	0.00	0.0051	32 B	0.19
Dev	outer	inner	22.6892 ns	0.2077 ns	0.1841 ns	1.00	0.00	0.0268	168 B	1.00
Foo	outer	inner	19.2641 ns	0.1288 ns	0.1142 ns	0.85	0.00	0.0268	168 B	1.00
Bar	outer	inner	19.5794 ns	0.4315 ns	0.4036 ns	0.86	0.02	0.0268	168 B	1.00
Baz	outer	inner	19.7202 ns	0.2628 ns	0.2194 ns	0.87	0.01	0.0268	168 B	1.00

code

BenchmarkRunner
    .Run<SwitchBenchmark>(
        ManualConfig
        .CreateMinimumViable()
        .AddValidator(DeferredExecutionValidator.DontFailOnError)
        .AddDiagnoser(MemoryDiagnoser.Default)
    );

public class SwitchBenchmark
{
    public IEnumerable<object?[]> Values =>
    [
        [null, null],
        [new Lazy<string>("outer"), null],
        [null, new Lazy<string>("inner")],
        [new Lazy<string>("outer"), new Lazy<string>("inner")]
    ];

    [Benchmark(Baseline = true)]
    [ArgumentsSource(nameof(Values))]
    public Lazy<string>? Dev(Lazy<string>? a, Lazy<string>? b) => Impl.Dev(a, b);

    [Benchmark]
    [ArgumentsSource(nameof(Values))]
    public Lazy<string>? Foo(Lazy<string>? a, Lazy<string>? b) => Impl.Foo(a, b);

    [Benchmark]
    [ArgumentsSource(nameof(Values))]
    public Lazy<string>? Bar(Lazy<string>? a, Lazy<string>? b) => Impl.Bar(a, b);

    [Benchmark]
    [ArgumentsSource(nameof(Values))]
    public Lazy<string>? Baz(Lazy<string>? a, Lazy<string>? b) => Impl.Baz(a, b);
}

static class Impl
{
    public static Lazy<string> Dev(Lazy<string>? outer, Lazy<string>? inner)
    {
        return new Lazy<string>(() => JoinContext(outer, inner));

        static string JoinContext(params Lazy<string>?[] contexts) =>
                string.Join("/", contexts.Where(ctx => ctx is not null).Select(x => x.Value));
    }

    public static Lazy<string>? Foo(Lazy<string>? outer, Lazy<string>? inner) => (outer, inner) switch
    {
        (null, null) => null,
        ({ } a, null) => a,
        (null, { } b) => b,
        ({ } a, { } b) => new(() => a.Value + "/" + b.Value)
    };

    public static Lazy<string>? Bar(Lazy<string>? outer, Lazy<string>? inner)
    {
        return (outer, inner) switch
        {
            (null, null) => null,
            ({ } a, null) => a,
            (null, { } b) => b,
            ({ } a, { } b) => Combine(a, b)
        };

        static Lazy<string> Combine(Lazy<string> outer, Lazy<string> inner) =>
            new(() => outer.Value + "/" + inner.Value);
    }

    public static Lazy<string>? Baz(Lazy<string>? outer, Lazy<string>? inner)
    {
        if (outer is null)
        {
            if (inner is not null)
            {
                return inner;
            }

            return null;
        }

        if (inner is null)
        {
            return outer;
        }

        return new(() => outer.Value + "/" + inner.Value);
    }
}

694ca94:
In #1939 (comment) I avoided allocating a new lambda closure object per iteration using, by creating a Func outside the loop, which the compiler sees at non-capturing and then allocates at most in a static field.
We can improve this by utilizing local functions, which is also marked as static to enforce it doesn't take any closure.

Inspect the lowered code in SharpLab to see the differences between using a lambda directly inside the loop, creating a Func outside the loop and now using a static local function.

Benchmarking:

Method	Step	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Func	?	0.6533 ns	0.0122 ns	0.0108 ns	1.00	0.00	-	-	NA
LocalFunction	?	0.0396 ns	0.0101 ns	0.0095 ns	0.06	0.01	-	-	NA
Func	System.Object	9.8370 ns	0.1420 ns	0.1259 ns	1.00	0.00	0.0140	88 B	1.00
LocalFunction	System.Object	9.3965 ns	0.0707 ns	0.0662 ns	0.96	0.02	0.0140	88 B	1.00

code

BenchmarkRunner.Run<Allocations>();

[MemoryDiagnoser]
public class Allocations
{
    [ParamsSource(nameof(ValuesForStep))]
    public object? Step { get; set; }

    public IEnumerable<object?> ValuesForStep => [new(), null];

    [Benchmark(Baseline = true)]
    public object? Func() => TryToProveNodesAreEquivalent(Step);

    [Benchmark]
    public object? LocalFunction() => TryToProveNodesAreEquivalent_New(Step);

    public object? TryToProveNodesAreEquivalent(object? step)
    {
        Func<object, GetTraceMessage> getMessage = static step => _ => $"Equivalency was proven by {step.GetType().Name}";

        if (step is not null)
        {
            return getMessage(step);
        }

        return null;
    }

    public object? TryToProveNodesAreEquivalent_New(object? step)
    {
        if (step is not null)
        {
            return GetMessage(step);

            static GetTraceMessage GetMessage(object step) =>
                _ => $"Equivalency was proven by {step.GetType().Name}";
        }

        return null;
    }

    public delegate string GetTraceMessage(object node);
}

IMPORTANT

• If the PR touches the public API, the changes have been approved in a separate issue with the "api-approved" label.
• The code complies with the Coding Guidelines for C#.
• The changes are covered by unit tests which follow the Arrange-Act-Assert syntax and the naming conventions such as is used in these tests.
• If the PR adds a feature or fixes a bug, please update the release notes with a functional description that explains what the change means to consumers of this library, which are published on the website.
• If the PR changes the public API the changes needs to be included by running AcceptApiChanges.ps1 or AcceptApiChanges.sh.
• If the PR affects the documentation, please include your changes in this pull request so the documentation will appear on the website.
  • Please also run ./build.sh --target spellcheck or .\build.ps1 --target spellcheck before pushing and check the good outcome

[github-actions]

Qodana for .NET

It seems all right 👌

No new problems were found according to the checks applied

💡 Qodana analysis was run in the pull request mode: only the changed files were checked
☁️ View the detailed Qodana report

Contact Qodana team

Contact us at qodana-support@jetbrains.com

• Or via our issue tracker: https://jb.gg/qodana-issue
• Or share your feedback: https://jb.gg/qodana-discussions

[coveralls]

Pull Request Test Coverage Report for Build 8409588228

Details

• 11 of 11 (100.0%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.002%) to 97.557%

---

Totals
Change from base Build 8361920663:	0.002%
Covered Lines:	12003
Relevant Lines:	12187

---

💛 - Coveralls