Add `IMethodSymbol Deconstruct` to `IDeconstructionAssignmentOperation`

[B1Z0N]

Background and Motivation

We have it like that right now:

    public interface IDeconstructionAssignmentOperation : IAssignmentOperation
    {
    }

Proposed API

Would be nice for the compiler to do the heavy lifting and get the symbol to the proper method being called:

    public interface IDeconstructionAssignmentOperation : IAssignmentOperation
    {
+       IMethodSymbol Deconstruct;
    }

PS: I simplified the issue scheme because it's far too simple of a change. What do you think?

[333fred]

PS: I simplified the issue scheme because it's far too simple of a change. What do you think?

@B1Z0N could you please fill out the entire template? Every change needs a justification; what is the scenario you're trying to use this in? What do you have to write today? How would this API change that?

[CyrusNajmabadi]

Championing this. IDE needs this in analyzers like the "remove unused members" analyzer to know that the .Deconstruct method is actually used.

[CyrusNajmabadi]

That said, it seems like we need to expose something more like DeconstructInfo which ahs the target, as well as deep nested info about conversions

[CyrusNajmabadi]

Ok. I ahve updated the proposal. It adds a CommonDeconstructionInfo which maps to the C# specific DeconstructionInfo: https://github.com/dotnet/roslyn/blob/main/src/Compilers/CSharp/Portable/Compilation/DeconstructionInfo.cs. Just like the C# version, it exposes the 'Method' and 'Conversion' (though as a CommonConversion as we're in the common layer), as well as nested infos.

[333fred]

@CyrusNajmabadi we have extension methods for getting C# and VB specific conversions off an IConversionOperation; do you want to propose similar extensions for getting language-specific deconstruction info?

[CyrusNajmabadi]

That makes a lot of sense @333fred . will update.

[333fred]

API Review

• Could we just have the extension method?
  • Wouldn't work well. The pain point is that we have to do language-specifics
• Names look good and match up with existing public API for DeconstructionInfo.
• If we need to, we'll look at calculating for CommonDeconstructionInfo lazily to avoid expensive work.

Conclusion: Approved