C# Language Design Meeting for August 13th, 2025

Agenda

• Unions

Quote of the Day

• "The operation succeeded, but the patient died"

Discussion

Unions

Champion issue: https://github.com/dotnet/csharplang/issues/8928

Union overview:

• https://github.com/dotnet/csharplang/blob/c3325533e57dec6aec3266e066e39abf7260e87a/meetings/working-groups/discriminated-unions/union-proposals-overview.md

Specs:

• https://github.com/dotnet/csharplang/blob/eb5381bf21b884a0dca53c9e127c0f527503fdfc/proposals/closed-hierarchies.md
• https://github.com/dotnet/csharplang/blob/eb5381bf21b884a0dca53c9e127c0f527503fdfc/proposals/nominal-type-unions.md
• https://github.com/dotnet/csharplang/blob/eb5381bf21b884a0dca53c9e127c0f527503fdfc/proposals/case-declarations.md

Reviewed document:

• https://github.com/dotnet/csharplang/blob/eb5381bf21b884a0dca53c9e127c0f527503fdfc/meetings/working-groups/discriminated-unions/to-nest-or-not-to-nest.md

We spent the meeting discussing union type nesting. There are two axes to consider: closed hierarchies vs. unions, and nested vs. non-nested types. We examined all four combinations and, while we did not reach a decision, several ideas emerged:

1. For closed hierarchies, we're unsure whether we actually need a "brief" syntax. We think these are scenarios that are naturally more complex and don't have built-in assumptions, so we might not benefit much from an opinionated, more succinct syntax.
2. We're also unsure whether unions need a brief syntax for top-level cases. We think these might be the more opinionated types, much like records are more opinionated than classes, and that nesting might be the opinion that unions have.
3. For type inference, we aren't too concerned by this given the above defaults, but we are open to the new type inference improvements being proposed as part of this feature. We don't think they are blockers, though.
4. There's some similarity between the designs we've arrived at and Scala 3's design decisions, so we do have some prior art that we can examine further.