What is the rationale for preventing null propagation with method groups?

[jnm2]

SharpLab

using System;

class C
{
    Action M(IDisposable disposable)
    {
        // ❌ CS0023: Operator '?' cannot be applied to operand of type 'method group'
        //               ↓
        return disposable?.Dispose;
    }
}

This is the code that must currently be used instead:

using System;

class C
{
    Action M(IDisposable disposable)
    {
        return disposable is not null ? disposable.Dispose : null;
    }
}

[alrz]

What is the rationale for preventing null propagation with method groups?

Not implemented.

[jnm2]

Meaning the compiler doesn't yet comply with the spec, or the spec work wasn't done?

[alrz]

The latter of course. I'd be surprised if something as big as this was just slipped after seven years.

[Rekkonnect]

I had no idea this doesn't work. With a bit of tweaking, I also discovered that ?? on method groups doesn't work either (in such context).

[jnm2]

Oh, I remember this. This one didn't super surprise me though, since the new Action(...) part would have to be communicated through the ?? operator by target-typing. I guess maybe the same thing is true of the ?. operator in some sense that I'm not seeing.

[Rekkonnect]

Whatever holds true, target type is visible and explicit. Method group should be bound as early as possible, enabling use of ?. and ??. Almost everyone would expect this behavior to be valid.

[jnm2]

I don't think so actually in the case of ??. Target typing doesn't normally get communicated through the ?? operator to either of its operands:

class C
{
    C M()
    {
        // ❌ CS8754: There is no target type for 'new()'
        return new() ?? null;
    }
}

class C
{
    C M()
    {
        // ❌ CS0019: Operator '??' cannot be applied to operands of type '<null>' and 'new()'
        return null ?? new();
    }
}

So I wouldn't expect target typing to SomeDelegate to happen for a method group operand, either.

[Rekkonnect]

You're right, target typing for ?? is another proposal on its own, and improving ?. for method groups should also enable ?? for method groups once ?? target typing is on, or at least closer to that step.

[333fred]

I'd be willing to champion changing this.

[jnm2]

dotnet/csharpstandard#654 explores whether the spec actually currently allows this, via target-typed conditional expressions, and ?. being defined as lowering to a conditional expression. I'm not sure there's been a conclusion yet.