A-n, A-p, A-o, A-i all don't have have consistent selection "orientations"

[emilyyyylime]

What I mean by the selections' orientation is the relative order of the cursor and the anchor.

I think it's important for movement operations to be consistent in their orientation because it lets me better string together multiple operations without having to actively look out for where my cursor is.

Furthermore, if I'm using A-n etc. to navigate the document I usually want/expect the object I'm moving into to come into focus, while the current behaviour seems to "prefer" not moving document around at all.

[the-mikedavis]

The current behavior is to inherit the direction from the original selection:

helix/helix-core/src/object.rs

Lines 70 to 74 in 4ec2a21

	if range.head < range.anchor {
	Range::new(to, from)
	} else {
	Range::new(from, to)
	}

What's the behavior you're looking for?

[emilyyyylime]

@the-mikedavis I think next sibling and prev sibling should always be forwards and backwards, respectively.

I guess I don't actually care about direction when both nodes are on screen at once but if I'm going to a node that's mostly off-screen (say from one top-level function to the next), I'd want it to become mostly on screen, not just to see a little sliver of it selected

[krish-r]

helix/helix-core/src/object.rs

Line 116 in 26d9610

Range::new(from, to).with_direction(range.direction())

Does it make sense to use Direction::Backward instead of Direction::Forward / range.direction() for the following commands? This way, we can always see the start of a (long) function.

• select_prev_sibling
• select_next_sibling

The following Tree-sitter based commands could also benefit from using Direction::Backward instead.

• goto_next_function
• goto_next_class

[the-mikedavis]

Always using Direction::Backward would make select mode behave oddly when that is implemented (#1999), also see #10022 (comment). select_prev_sibling should always use Direction::Backward and select_next_sibling should always use Direction::Forward instead

[the-mikedavis]

Fixed in #10542