drdexebtjl • today at 4:51 PM

>Those interceptors and inline arrays via attributes instead of proper language grammar aren't that great either.

Yeah. Even when they add new grammar nowadays, it's always just something that trivially sugars away into previous grammar (see: records, `with` clones, extension properties, required, etc).

The moment they need something that it's slightly more complex... Out of scope. Even when it's completely necessary for the thing to be useful in practice.

For example, they added `required`, `record`s and property initializers, giving us good reasons to write `new Foo { A = a, B = b }` instead of `new Foo(a, b)`. A and B must be positive, so you'd write:

  public required int A { get; init => field = value > 0 ? value : throw ... ; }
  public required int B { get; init => field = value > 0 ? value : throw ... ; }

This is pretty standard C# code that you might see in an example for records.

But then the requirements change: A and B must be positive, or they must both be zero at the same time.

This cannot be expressed at all with initializers. You simply cannot add code that runs after all initializers are called. You're stuck chasing every single initialization of Foo and using a constructor or factory method instead. Shipped it as a public API? Too bad. Should have seen it coming!

The new features are filled with this sort of thing. As if Microsoft never used them beyond the most basic examples. Or maybe they did, and explicitly chose not to fix it and solve later.