Support overriding existing Methods Callbacks

[mattsse]

Currently, it's not possible to override MethodCallback in Methods because Methods::merge fails early if a callback with the name already exists

jsonrpsee/core/src/server/rpc_module.rs

Lines 355 to 357 in a0ce8d4

	for name in other.callbacks.keys() {
	self.verify_method_name(name)?;
	}

This makes it impossible to disable various callbacks dynamically when RpcModules (via server proc macro's into_rpc) are merged.

now that verify_method_name is public this could be solved by:

• support Methods::iter and Methods::into_iter so callbacks can be inserted individually.
• support Methods::merge_replace which replaces existing handlers

[niklasad1]

Hey @mattsse

The methods are really "static" i.e, after the server has been started then it's not possible change or override the methods.
Can you explain with an example what you want to do?

I think Methods::iter and Method::merge_replace make sense.

[mattsse]

The methods are really "static" i.e, after the server has been started then it's not possible change or override the methods.

right, I'm talking about the setup phase.

for example, you have a bunch of server traits and default implementations that are then merged via:

#[rpc(server))]
pub trait MyApi {
  async fn do_stuff() {}
}

#[rpc(server))]
pub trait MyApi2 {
  async fn do_other_stuff() {}
  
}

fn default_setup () -> RpcModule {
   let mut module = RpcModule::new(());
   module.merge(MyApi::new().into_rpc()).unwrap();
   module.merge(MyApi2::new().into_rpc()).unwrap();
   module
}

now RpcModule is a list of methods.

it would be useful to now replace certain methods with something like

#[rpc(server))]
pub trait Unsupported {
  async fn do_stuff() {
    return Err("Unsupported method: do_stuff")
  }
}

fn customize(m: Unsupported) -> RpcModule {
  default_setup().merge_replace(m.into_rpc())

}

So the motivation behind this is to make it easier to selectively replace certain methods in the default config.

I think Methods::iter and Method::merge_replace make sense.

cool, I can try adding these

[niklasad1]

sounds good 👍