csharplang/meetings/2022/LDM-2022-07-27.md

C# Language Design Meeting for July 27th, 2022

Agenda

1. scoped for local declarations

Quote of the Day

• nervous chuckling "That's actually an interesting question!" "I know, it's just when you say scary things."

Discussion

scoped for local declarations

https://github.com/dotnet/roslyn/issues/62039
https://github.com/dotnet/csharplang/issues/1147

Today, we looked at allowing scoped as a local modifier in all local declaration context, such as usings, foreach iterator variables, and such. We like the idea in principle: users don't think of the various different grammatical ways of creating a local variable differently, they think of them all as locals. There were a couple more in-depth points of discussion about the implementation:

• Is scoped or not something that var should be able to infer? This matters for deconstruction, as in var (x, y) = MethodCall();. After some consideration, we don't think scopedness is something we can or should infer. It might also vary between each of the deconstruction variables. Given this, we think that scoped needs to explicitly declared, not inferred in any local case.
• The specification for scoped doesn't particularly line up with how we consider modifiers. We'd like to think of this as a new kind of modifier, a modifier that must go directly on a type, rather than a modifier that can go anywhere in the modifiers list of a method. Unlike method modifiers, scoped is specifically a modifier on either the return value or a parameter value, and should be applied in those locations. scoped unsafe Span<int> M() => ...;, for example, isn't legal. The scoped affects the return, while the unsafe generally affects the method body. We therefore think that the specification (and Roslyn implementation) should match this intuition.

Conclusion

We approve the general concept in this issue, but the grammar specification needs to be updated.