I guess this can be considered a (non human friendly) feature. Close it if it is (;

Well, R2 just creates a vertical stack with your code, which would be even harder to climb. :^) At a glance it made sense to me, because the cursor (next position) is set when a face is placed, so when you get to a direction keyword it doesn't go back and recalc from the previous face. Design wise, this could be justified either way. But when I looked at the code for below/across, I saw that you could trick it. e.g. view [base base below base across base below across below across base across base] by alternating direction keywords is keeps applying spacing even though no new face is added. That's a bug for sure, as they should be idempotent.

I don't have time to think through effects on grid layout mode, but I don't think changing this will break much code. It's possible someone leverages it, especially in combination with R2 code using the here: at idiom. The other word I used in R2 quite a bit was guide, when making layouts with complex rows or columns.

The stair layout is a great example for this, if a bit pathological. ;^)

Yeah well. I see why it is like this too. But it's not that hard to keep both the offset & size of the last laid out face, and then walk forth from it along the across/below chosen axis (in size+spacing increments). Almost the same complexity but more predictable.

If you create a branch to do that, I'll press Nenad to evaluate it.

I'll rescind my pathological comment as well, because this could apply directly to tree view layouts.