Proposal: Type pattern for generic types

[gulshan]

Now, type patterns do not work with generic types. A type argument is required to determine actual type. Given the following generic classes (using closed and record syntax)-

public closed abstract class Option<T> { }
public sealed class Some<T>(T Value) : Option<T> { }
public sealed class None<T>() : Option<T> { }

The code I want is write is-

var x = SomeGenericMethod();
if(x is Some<int> i) // possible now, the type is needed to be mentioned explicitly
    Write(i.Value);
if(x is Some i) // proposing here. full type will be inferred
    Write(i.Value);
if(x is Some<var> i) // or this one if previous one is impossible
    Write(i.Value);
if(x is Some<_> i) // or this one
    Write(i.Value);

Grammar will be-

pattern
    : type_pattern
    ;

type_pattern
    : type identifier
    | generic_type identifier
    ;

[YotaXP]

if(x is Some i) would not be possible, as it would be ambiguous between Some<T>, Some<T,U>, and a non-generic Some class. The other options seem fine to me so far, though.

[orthoxerox]

@YotaXP why not allow it if the compiler can infer that there's only one applicable deconstructor?

upd: shouldn't've answered from my phone. I see it's not a deconstructor, it's a typeswitch.

[gafter]

@gulshan Can you please provide the proposed translation of this construct, as well as some realistic examples that illustrate the need this fulfils?

[alrz]

I think it's only possible to omit the type argument when it's known to the compiler.

[weitzhandler]

Named parameters or omission by index looks fine to me.
But I vote for the general idea of including non-inferred types only.