LDM Notes for August 27th, 2025

[333fred]

https://github.com/dotnet/csharplang/blob/main/meetings/2025/LDM-2025-08-27.md

Agenda

• Type inference in patterns
• Type value conversion
• Union syntax

[KennethHoff]

In relation to Type inference in patterns, I think a syntax like this would make more sense, and I can see a path [ syntactically ] from that syntax to #6308

if (myList is List<?>)

[jnm2]

What is the concept that the language has allowed without ceremony for 20 years?

[CyrusNajmabadi]

The language has allowed generic inference with method calls without any <...> ceremony :)

[jnm2]

Ah, gotcha. It's still hard to get my head around abstracting that from methods to types.

[TahirAhmadov]

I would vote for the typeof syntax. Other than it being a different symbol type, the usage is probably going to be less frequent, and consistency with type of has a value of its own. Most importantly, how do we disambiguate between different number of generic arguments?

[jnm2]

I don't expect it to be common for multiple non-zero arities to be possible in the strong type being pattern matched against. I do see this being pretty frequently used with DUs though, not that much less frequent in some domains.

[TonyValenti]

I'm a big fan of Enum syntax for unions.

[Unknown6656]

I agree. However, should anonymous unions be introduced, I'd be in favour of int | string.

[TonyValenti]

Yes to this as well.

[KennethHoff]

I think it's important for the LDM that the pattern syntax matches the construction syntax whenever possible, and we already have the pattern syntax:

if (value is string or int)

[Richiban]

I'm a big fan of Enum syntax for unions

I think enum styntax is problematic here because the members in an enum are declarations of new things, but in a union the cases are references to existing things. For this reason I don't think the two constructs should share syntax