Fix for null resolveProvider in codeAction

[fsommar]

The resolveProvider field doesn't need to be present even if the codeAction param is.

From the specification:

export interface CodeActionOptions extends WorkDoneProgressOptions {
    /**
     * CodeActionKinds that this server may return.
     *
     * The list of kinds may be generic, such as `CodeActionKind.Refactor`,
     * or the server may list out every specific kind they provide.
     */
    codeActionKinds?: CodeActionKind[];

    /**
     * The server provides support to resolve additional
     * information for a code action.
     *
     * @since 3.16.0
     */
    resolveProvider?: boolean;
}

[CLAassistant]

All committers have signed the CLA.

[nixel2007]

BTW why the left part of Either is not checked?

[fsommar]

BTW why the left part of Either is not checked?

I think it's fair to assume that the resolve provider is false if no options are being passed; either codeAction is enabled with a boolean for the most basic support, or options are provided where additional functionality like codeAction.resolve can be enabled.

Specification https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#codeActionOptions

[nixel2007]

But it can be disabled with the boolean as well

[fsommar]

But it can be disabled with the boolean as well

I'll use examples to illustrate how I'm thinking:

1. {
   "codeAction": true
   }

2. {
   "codeAction": {
   "resolveProvider": true
   }
   }

3. {
   "codeAction": {
   "resolveProvider": false
   }
   }

4. {
   "codeAction": {}
   }

5. {
   "codeAction": false
   }

6. {}

---

#	codeAction	codeAction.resolve
1	true	false
2	true	true
3	true	false
4	true	false
5	false	false
6	false	false

---

If you're referring to 5., it will be false for codeAction.resolve since provider != null && provider.isRight() will be false.

[nixel2007]

Oh, sorry, I missed that this method is about codeActionResolve and not codeAction itself. Yes, you are right and everything is correct