Whenever split horizontal is triggered, window is "pushed" down

[Nikki1993]

I am having hard time describing it well so I've recorded a video.

I got 2 keybinds.

cmd-alt-ctrl-h = 'split horizontal'
cmd-alt-ctrl-period = 'split vertical'

Whenever I want to split horizontally window gets pushed. Then I switch back to split vertical and it goes back up. If I trigger split horizontally and then close the window to the right and re-open it again, window gets correctly positioned from top to bottom. If I open another, it also correctly splits horizontally. But if I were to open a new window when the previous one is "pushed" down, the second window will also appear "pushed" down.

Screen.Recording.2024-01-03.at.18.48.51.mov

[fmasa]

@Nikki1993 This is because split creates new implicit horizontal container and the window is moved inside.

In default config, there is a indent-for-nested-containers-with-the-same-orientation = 30 which automatically idnets nested containers with the same orientation.

[nikitabobko]

I just got an idea that it would make sense to create paddings from all 4 sides for windows that are the only child of their parent container, regardless of the containers orientations

Showing 4 paddings would be a better visualization

Example 1:

h_tiles
  window1
  v_tiles
    window2  <- create all 4 paddings

Example 2

h_tiles
  window1
  h_tiles
    window2  <- create all 4 paddings

And apply indent-for-nested-containers-with-the-same-orientation only in cases when there are several children

Example 3

h_tiles
  window1
  h_tiles      <-- "indent" the container. indent-for-nested-containers-with-the-same-orientation
    window2
    window3

Example 4

h_tiles
  window1
  v_tiles      <-- no indentation
    window2
    window3

[nikitabobko]

Idea No. 2

When a window is the only child of its parent then pretend that there is a second child in the container. It would create a "gap" in the tiles. But the gap will make it clear what orientation the parent has.

Given this layout:

h_tiles
  window1
  h_tiles
    window2

It would be rendered as if window2 had a sibling:

h_tiles
  window1
  h_tiles
    window2
    fake_sibling_window

Given this layout:

h_tiles
  window1
  v_tiles
    window2

It still would be rendered as if window2 had a sibling:

h_tiles
  window1
  h_tiles       <-- It's unclear whether the layout should be applied
    window2
    fake_sibling_window

Correct me if I'm wrong, but the only reason why people create containers that have only a single child is to later add another window as a sibling. The empty gap in tiles would make this intention clear

And this idea makes it possible to use split command with enable-normalization-flatten-containers. enable-normalization-flatten-containers could be dropped. Effectively it would always be enabled

[nikitabobko]

Breaking change. indent-for-nested-containers-with-the-same-orientation is confusing, so I dropped it in 0.10.0