Change interior domain

[AndreasArvidsson]

• Interior domain is now same as content range (before shrinking to text content). "take inside" with the cursor on a function name will not select function body anymore.
• Interior of typed target now utilities every scope. "take inside funk" will select function body. "take inside state" will select all branch bodies in an if statement.

[pokey]

I'm a bit surprised that this actually works. For example, in:

function aaa(bbb: ccc) {
   ddd;
}

If I say "clear inside funk air", how come you don't get:

function aaa(|) {
   |
}

[pokey]

ohh I guess it's because (bbb: ccc) is surroundingPairInterior whereas this is explicitly requestiont interior?

[pokey]

So that means that even for {}-style languages, we still need to define interior in tree-sitter if we want "inside funk" to work?

[pokey]

Ok here's a pathological case:

function aaa(bbb = () => { return "ccc"; }) {
   ddd;
}

In this case, you'd expect "clear inside air" to give:

function aaa(bbb = () => { return "ccc"; }) {
   |
}

but instead you'd get:

function aaa(bbb = () => { | }) {
   |
}

I think this is an ok compromise, but just figured I'd flag to demonstrate why this approach is really just a hack that approximates what we actually want

[AndreasArvidsson]

Exactly. inside scope completely rely on Tree sitter.

Maybe I can work around this. It's not perfect, but I think this is a decent solution for now.

[AndreasArvidsson]

I had look at this and there's nothing in Tree sitter that can help us with this particular case with my suggested implementation. With that said I'm happy with this behavior for now. I think it's much better than today's implementation.

The only way to properly fix this is to use interior properties. I guess we could do both? Where it matters like function bodies we could use a property and for other things like if statements we use a separate scope? That would probably give us the best of both? Right now I can't think of any other case except functions where you could create the above problem. The question is if it worth the increased complexity for this edge case?