feat(command): select all siblings and children

[dead10ck]

This adds two new related commands:

select_all_siblings

Selects all siblings of the node under the cursor.

select_all_children

Selects all children of the node under the cursor.

select_all_children_in_selection

Selects all children of the node under the cursor, but only if the children are contained inside the starting selection.

Edit: saving this one for a future PR

[pascalkuthe]

after trying this out for a bit I think select_all_children would benefit a lot from only selecting children that are actually contained in the selection. The current implementation is also useful but i found that quite a lof of nodes have children I don't care about. If someone really wants all children inside the node they can do a-o first. That change would also make the command more predictable I think

[dead10ck]

after trying this out for a bit I think select_all_children would benefit a lot from only selecting children that are actually contained in the selection.

Hmm, but in this case, the command would basically never do anything when the selection is just a cursor, or really any time you don't happen to have an entire node contained in the selection.

But on the other hand, I could see this being useful for if you only want a subset of the children. I'll add a third command that does this.

[dead10ck]

Just added select_all_children_in_selection that only selects if all resulting ranges are contained within the starting selection.

[pascalkuthe]

This should be Adapted to cross tree sitter injection layer like the othrr motions now do, the rest lgtm

[pascalkuthe]

I think thus also needs to be able to cross injection layer so I would use the new tree cursor here

[dead10ck]

Sounds good, will do

[dead10ck]

Ok I've rebased and changed all the motions to use the new in-crate TreeCursor instead of the tree-sitter API directly. This required adding a child iterator in the TreeCursor module.

[pascalkuthe]

Suggested change

	let found = match self.named {
	true => self.cursor.goto_first_named_child(),
	false => self.cursor.goto_first_child(),
	};
	let found = self.cursor.goto_first_child_impl(self.named);

[pascalkuthe]

``

Suggested change

	if !found {
	return None;
	}
	return Some(self.cursor.node());
	return found.then_some(self.cursor.node());

[pascalkuthe]

Suggested change

	let found = match self.named {
	true => self.cursor.goto_next_named_sibling(),
	false => self.cursor.goto_next_sibling(),
	};
	let found = if self.named {
	self.cursor.goto_next_named_sibling()
	} else
	self.cursor.goto_next_sibling(),
	};

I don't love matching on bools. Nothing really wrong with it but this is one of the "multiple ways to do the same things so always doing it the same way make code more consistent".

may be worth making named a parameter (with an _impl` function) so you don't need the if else at all.

[pascalkuthe]

Suggested change

	// subsequent iterations, keep visiting siblings until we run out
	if found {
	return Some(self.cursor.node());
	}
	None
	// subsequent iterations, keep visiting siblings until we run out
	found.then_some(self.cursor.node())

[pascalkuthe]

Hmm I am not sure this is quite what I had in mind here. If I can tell this right this would just not do anything as soon as any child is not in the selection. Instead I would like to see this be sort of like anl intersection. We should select the children covered by a selection.

In pseudo code all_chidren.retain(|range| selection.contains(range) (just better implemented so its not O(n^2).

If this is too complex just omit this and I think I can implement this when solving #3007

[dead10ck]

Hmm, maybe I'll just leave it out for now

[pascalkuthe]

Every callsite has a fallback so lets just inline that code here to avoid duplication

Suggested change

	if !children.is_empty() {
	Some(children.into_iter())
	} else {
	None
	}
	if children.is_empty() {
	vec![fallback]
	} else {
	chidren
	}

[pascalkuthe]

I would just return a Vec and let the callsite call into_iter or if you really want to return an iterator use an anonymous impl Iterator instead this is a bit hard to read

[dead10ck]

Good suggestions, I took care of them.

[the-mikedavis]

Looks great! Sorry for throwing a wrench into this with the TreeCursor changes - thanks for rebasing and improving the TreeCursor while you were at it

[sudormrfbin]

I can finally stop splitting by commas inside a function call to select the arguments :) But it seems the docs weren't updated with the new alt-a and alt-I keybinds.