Discussion: partial methods and interfaces

[alrz]

Currently, all partial methods require a body but in an interface, you might still want to keep it abstract:

partial interface I { [A] public partial void M(); }
partial interface I { [B] public partial void M(); }

Equilvanet to:

partial interface I { [A, B] public void M(); }

Note that partial void M(); follows the old semantic and optionally emits private void M() { } if it has an implementation part.

[YairHalberstadt]

This should also be the case for abstract methods in classes.

[RikkiGibson]

It feels like the suggestion is to decide whether a partial method with access modifiers is a DIM based on presence of an implementation part.

It would help if you could share some concrete use cases for this.

[alrz]

@RikkiGibson

The one in the opening post is my use case. I want to add some attributes to interface methods in separate partial decls.

Note the result would not use DIM in any ways, I only added public to use new semantics for partial.

[alrz]

new semantics for partial.

However that itself requires a body which I don't want to add.

On the other hand, partial void M() would actually emit a "private method" (if it has a implementation part) which was a little surprising.

[RikkiGibson]

Sorry, I was trying to ask what kinds of applications or features are you having difficulty implementing because of this limitation of the partial methods design.

[alrz]

Nothing fancy, just wanted to add some framework-specific attributes to an interface I linked from another project.

I did manage to wokaround this though, using placeholder attributes where they were not available.