Report error on method name for chained method calls

[dhruvrajvanshi]

Fixes #31365

This is a first stab at fixing the issue. It reports the error at the name of the final method call.

Ideally, the error would be reported at the argument list span for the call but I couldn't find a straightforward way to get a span for the argument list. I'm willing to the PR if we really need reporting at the argument list.

[DanielRosenwasser]

I think the idea was that the argument list was supposed to be included in the range.

[DanielRosenwasser]

Ah, just saw your comment in the PR.

[DanielRosenwasser]

Spoke with @RyanCavanaugh - we think just erroring on the name itself is fine, but we'd like to error on the invoked expression itself even when it's not a property access. In other words, if you don't have a property access as the target, use node.expression for the error span.

[dhruvrajvanshi]

@DanielRosenwasser I'm not quite sure I understand.
Do you mean something like this?

function getCallErrorNode(node: CallLikeExpression): Node {
        if (isCallExpression(node)) {
            if (isPropertyAccessExpression(node.expression)) {
                return node.expression.name;
            } else {
                return node.expression
            }
        }
                
        return node;
}

Because this is breaking a lot of error messages (and they look worse IMO).

Ideally, the error should include the final method name and its argument list which could be done by including a span in NodeArray.

[DanielRosenwasser]

Yeah, something like that code.

Because this is breaking a lot of error messages (and they look worse IMO).

Maybe we'd do this for only Identifier nodes. What looks grosser?

[dhruvrajvanshi]

I've updated the PR so you can see for yourself. Most call argument arity errors exclude the argument list. It is more uniform with the new change though.

Is there a way to report diagnostics on spans instead of nodes?

[DanielRosenwasser]

I think this is a great first step. If we want to iterate further, we'll take another PR.