Add depth limit as argument for how often to autotile #41

Syphdias · 2022-10-27T17:59:17Z

Currently the script will autotile (vsplit/hsplit) every window focused on with no limit.

At a certain depth this behaviour is not desirably to me and I want to add more windows to the same container. Especially for terminals I prefer to have more space horizontally than vertically to see the entire line if possible.

To achieve this, I add a optional parameter to prevent further autotiling after the specified limit. For example with a 16:9 display and a depth limit of 1, it will split horizontally first and then only split vertically but in the same container.

 ________________
|        |       |
|--------|       |
|        |-------|
|--------|       |
|________|_______|

As a side effect, this enables stacking and tabbed layouts after reaching the limit.
The default behavior is still the same with an infinite limit for how deep to autotile.

Edit: I mainly made this for me. Not sure if you are interested in this.

Currently the script will autotile (vsplit/hsplit) every window focused on with no limit. At a certain depth this behaviour is not desirably to me and I want to add more windows to the same container. Especially for terminals I prefer to have more space horizontally than vertically to see the entire line if possible. To achieve this, I add a optional parameter to prevent further autotiling after the specified limit. For example with a 16:9 display and a depth limit of 1, it will split horizontally first and then only split vertically but in the same container. ``` ________________ | | | |--------| | | |-------| |--------| | |________|_______| ``` As a side effect, this enables stacking and tabbed layouts after reaching the limit. The default behavior is still the same with an infinite limit for how deep to autotile.

nwg-piotr · 2022-10-27T23:05:52Z

Well, I'm not sure. Need to give it a try and use for some time.

Syphdias · 2022-10-29T10:47:24Z

I unfortunately found a bug. It is caused by i3 not "merging" containers when not used.

What I mean by that is: If you open a window, vsplit, open a window, vsplit, etc. and then close everything but the last window, you end up with nested single vsplit containers.

I closed the upper ones and left the last one open. This results in a structure that looks like this:

      workspace: 3 (splitv)
        con:  (splitv)
          con:  (splitv)
            con:  (splitv)
              con:  (splitv)
                con: Tilix: ~ (splith)

output is "type: name (layout)"

This looks like depth 0 with only one window in the workspace but it is recognized as depth 4 or 5.

nwg-piotr · 2022-10-29T14:38:38Z

I don't care about i3 in the nwg-autotiling clone package. And actually I don't test on i3, as X11 sucks on my triple-headed setup with hybrid graphics. I'd gladly use your solution in nwg-shell, but if it comes to the autotiling repo & package, I need to rely on you.

Syphdias · 2022-10-29T17:57:19Z

Gaming keeps me on X11 for now. ;)

Apparently the containers not being flattened is an i3 bug: i3/i3#3503
I might be able to detect if there is only one node in the parent and not count it. Need to play around with it a bit.

How does sway handle the nested containers? Do they get merged/flattened?

nwg-piotr · 2022-10-29T22:28:43Z

Gaming keeps me on X11 for now. ;)

I only have one game installed (Witcher 2), but it behaves perfectly well on sway.

How does sway handle the nested containers? Do they get merged/flattened?

Wish I knew. ;)

Syphdias · 2022-10-29T22:37:34Z

How does sway handle the nested containers? Do they get merged/flattened?

Wish I knew. ;)

You could try this #41 (comment) and then run
https://github.com/Syphdias/i3nodes/blob/main/i3nodes.py (need python3.5 i think and i3ipc(docs says it works with sway as well))

nwg-piotr · 2022-10-29T22:49:56Z

$ print-or-nodes
root: root (splith)
  output: __i3 (output)
    workspace: __i3_scratch (splith)
  output: eDP-1 (output)
    workspace: 7 (splith)
      con: Add depth limit as argument for how often to autotile by Syphdias · Pull Request #41 · nwg-piotr/autotiling - Google Chrome (none)
  output: DP-1 (output)
    workspace: 1 (splith)
  output: HDMI-A-1 (output)
    workspace: 4 (splith)
      con:  (splitv)
        con:  (splith)
          con: ~/bin/print-or-nodes - Mousepad (none)
        con:  (splith)
          con: foot (none)
      con:  (splitv)
        con: bin (none)
        con:  (splith)
          con:  (splitv)
            con: foot (none)
          con:  (splitv)
            con: foot (none)
            con:  (splith)
              con: foot (none)

Syphdias · 2022-10-29T22:54:44Z

I closed the upper ones and left the last one open. This results in a structure that looks like this:

You did not close all the upper terminals but yes, looks like sway does not flatten the hierarchy as well. Otherwise there would be no containers with only one other container in it:

        con:  (splith)
          con:  (splitv)
            con: foot (none)

nwg-piotr · 2022-10-29T22:57:35Z

Tried on HDMI-A-1.

$ print-or-nodes
root: root (splith)
  output: __i3 (output)
    workspace: __i3_scratch (splith)
  output: eDP-1 (output)
    workspace: 7 (splith)
      con: Add depth limit as argument for how often to autotile by Syphdias · Pull Request #41 · nwg-piotr/autotiling - Google Chrome (none)
  output: DP-1 (output)
    workspace: 1 (splitv)
      con:  (splitv)
        con: nwg-shell-config – ui_components.py (none)
  output: HDMI-A-1 (output)
    workspace: 4 (splith)
      con: foot (none)

nwg-piotr · 2022-10-29T23:08:22Z

You mean this situation, right (HDMI-A-1)?

$ print-or-nodes
root: root (splith)
  output: __i3 (output)
    workspace: __i3_scratch (splith)
  output: eDP-1 (output)
    workspace: 7 (splith)
      con: Add depth limit as argument for how often to autotile by Syphdias · Pull Request #41 · nwg-piotr/autotiling - Google Chrome (none)
  output: DP-1 (output)
    workspace: 1 (splitv)
      con:  (splitv)
        con: nwg-shell-config – ui_components.py (none)
  output: HDMI-A-1 (output)
    workspace: 4 (splith)
      con:  (splitv)
        con: foot (none)

Too bad! I like the behavior at the depth limit 1.

With lots of splits the hierarchy can get deeper than normal, if containers get closed. i3 does not flatten this hierarchy (atm). This means that the depth option does needs to ignore containers that only contain one child/node. This is done only incrementing the current depth when checking against the limit if the nodes of the parent container is bigger than 1.

Syphdias · 2022-10-30T16:55:33Z

So I think I fixed it by ignoring containers with only one child container in them and disregard them as depth. Depth changed a bit:

depth 0 is unlimited
depth 1 is control the first split
depth 2 is control until the seconds split (former depth limit of 1)

nwg-piotr · 2022-10-30T19:46:41Z

Correct me if I'm wrong.

-l 0 does nothing visible;
-l 1 spoils autotiling;
-l 2 does what we want;
-l 3 and more causes behavior expected, but probably unwanted by anyone.

Right? If so, shouldn't we just have a boolean flag, e.g. -m for "Mimic master/stack layout" (or -l for "Limit autotiling"), and constant DEPTH_LIMIT = 2?

Syphdias · 2022-10-30T22:15:33Z

Correct me if I'm wrong.

Nope, I totally agree. Unless your aspect ratio is different maybe. I tried flipping my 16:9 monitor to vertical and there 3 might make sense – not to me though.

We could provide both options and imply -l 2 with -m. We'd need to answer the question to what happens to -l 3 -m though. Which one wins or do they conflicz and exit 1?

Do you use a master/stack layout on sway? If yes, is a sway feature or do you use scripts/tooling? I thought about making a script to swap containers in a certain way but that is out of the scope of this repo.

nwg-piotr · 2022-10-30T22:23:18Z

Do you use a master/stack layout on sway?

Nope, but I liked it on DWL. In contrary to DWL itself. 🤣

I do believe in the KISS rule. Let's not multiply arguments that do the same job. If you find the current -l argument better, I'm all right with it. My nwg-shell config will just disallow values other than 2, exactly as I did it for testing purposes:

Whoever uses the original autotiling script, will decide on their own.

Syphdias · 2022-10-30T22:31:52Z

My worry with -m would be that for certain layouts it might not imitate master/stack but still work as expected. For example you could turn your first split manually to be a splitv and autotiling would then splith in both containers (infinity with -l 2, and differently with default). Another example might be #29 where the first two splits are horizotal due to window geometry I assume.

So I'd say -l and the recommendation to try 2 as a value and that 1 does not really make sense.

nwg-piotr · 2022-10-30T22:33:43Z

OK. Could you also update README.md? I'm working on a small fix to another repo at the moment.

Syphdias · 2022-10-30T22:50:38Z

Absolutely, I'll get to it tomorrow.

updating help
updating README

nwg-piotr · 2022-10-30T22:56:55Z

Thanks!

Hope you'll give sway & nwg-shell a chance one day. The project needs valuable contributors.

The help and README did not go into details what exactly the --limit option does. Since it can be used to imitate a master-stack layout, and can enable tabbed or stacking layouts to some degree, this adds the explanation in the section talking about this limitation. It softens the limitations of autotiling while still stressing that tabbed and stacked layouts may behave unexpectedly and are not supported by autotiling.

nwg-piotr · 2022-10-31T11:27:57Z

Thank you! I'll merge later tonight.

Syphdias · 2022-10-31T11:36:08Z

Hope you'll give sway & nwg-shell a chance one day. The project needs valuable contributors.

This genuinely makes me interested in giving Wayland a go, but if I remember correctly the problem with Wayland was nvidia drivers (bought while my daily driver was still Windows).

Syphdias · 2022-10-31T11:37:04Z

Thank you! I'll merge later tonight.

No rush, I rephrased the section about not working with tabbed/stacked quite a bit. Please nitpick away.

nwg-piotr · 2022-10-31T11:37:48Z

Dunno what the nvidia status nowadays is. My hardware is all-AMD.

Syphdias · 2022-10-31T13:58:56Z

Dunno what the nvidia status nowadays is. My hardware is all-AMD.

Pretty happy with my Ryzen 7 from a few years back. My next GPU will be red (or blue – one can hope). But currently not in need of a better one, (un)fortunately. But I am looking forward to a new one.

Just a quick Google for the state of things in 2022:

As far as the GBM and wlroots support goes, my experience is that it is not up to the point of being properly used. Better than before, but not good enough.
— https://blog.devgenius.io/wayland-and-nvidia-in-2022-2f0407fb34f4

They did not even try to game on it and it failed the test.

nwg-piotr · 2022-10-31T20:26:24Z

Actually my previous laptop, back in 2015-2020, was an Intel/Nvidia hybrid, and sway worked well on it. The best test is what you check yourself.

nwg-piotr · 2022-10-31T20:45:08Z

About to merge.

Final reflection: for my use case (1 vertical and 2 horizontal displays), using -l 3 would make sense if applied to the vertical display only.

Maybe it's worth of further consideration: a possibility to use different depth level on different displays.

Syphdias · 2022-11-01T11:11:19Z

I think the easiest might be to add --outputs like --workspaces and you define as many exec_always as you have displays. Not sure if that is compatible with your gtk interface for configuring it.

nwg-piotr · 2022-11-01T22:08:42Z

Not sure if that is compatible with your gtk interface for configuring it.

Haha! If I managed to integrate swaync and gtklock, this is going to be be an easy task. ;)

nwg-piotr · 2022-11-03T00:01:57Z

Well, I've just realized that the version I use in nwg-shell allows a single instance only. :/

nwg-piotr · 2022-11-03T02:15:29Z

For my specific use case it's possible to store the "output-name": limit dictionary in the nwg-shell-config settings file, so I won't need arguments. It could look something like this: https://github.com/nwg-piotr/nwg-shell-config/blob/limit-per-output/nwg_shell_config/autotiling.py

nwg-piotr · 2022-11-03T02:36:10Z

Syphdias · 2022-11-08T12:51:34Z

Would there be use for something like: limit of 3 on output x with workspace 1 3 5, but limit of 3 on output y with any workspace.
Hm, I don't know. Just trying to add to the idea. It would be hard to do via CLI if you only want one process. Something like xrandr arguments comes to mind where the order matters (if you are not familiar with it: xrandr --output DP-1 --mode 2560x1440 --output DP-2 --mode 1920x1080). Also, probably hard to do with argparse.

It would split the arguments in two groups: one for filtering (output, workspace, maybe more) and one for attributes (limit, ratio, etc.). This would add complexity which I am not a fan of. You can already do all this by running multiple commands.

nwg-piotr · 2022-11-08T12:54:52Z

It's going to look like this on my side:

But I use no arguments, just dictionaries stored in a json file.

Syphdias · 2022-11-08T22:37:25Z

I know. But if it was possible through arguments you could use the @ symbol.

Off topic:
I tried to rewrite the nested loop and replace it with "gates" also ran black to unify formatting. What do you think?
https://github.com/Syphdias/autotiling/blob/code-style/autotiling/main.py

nwg-piotr · 2022-11-08T23:17:51Z

Well, I use PyCharm, so it will reformat it back on my next Ctrl+Alt+L.

Syphdias · 2022-11-09T08:36:16Z

Black runs on save in my vscode 😅
I can leave the formatting alone.

Mainly check out the logic rewrites

nwg-piotr · 2022-11-09T11:10:41Z

Mainly check out the logic rewrites

Looks good while reading. Needs to be tested carefully, anyway.

If it comes to formatting: I'm strongly against reformatting

def switch_splitting(i3, e, debug, outputs, workspaces, depth_limit):

to

def switch_splitting(
        i3, e, debug, outputs, workspaces, depth_limit, splitwidth, splitheight, splitratio
):

It's 1 line -> 3 lines. The more code I see w/o scrolling - the better.

Syphdias · 2022-11-09T12:28:15Z

Ignore the formatting. It is done by black which is quite opinionated. The only reason it got touched is the line length. Again, Ignore the formatting. I just didn't turn it off for this.

I'll wait for the open PR to be done. If you like the sieve/gate structure, I can open I PR for the logic changes and migrate them one by one to make it obvious that the behavior is still the same.

nwg-piotr · 2022-11-09T12:31:35Z

I'll wait for the open PR to be done.

Yes, please. There's still a thing I'd like to be considered there.

HirschBerge · 2022-11-13T06:11:54Z

Not sure if this is strictly related, but on ultrawide screens, tiles go three across before tiling correctly
but if I force my resolution to 1920x1080 it will tile correctly

Syphdias · 2022-11-13T12:18:41Z

This behaves as expected and what is also reported by #29. autotiling does not alternate between splits, it decides based on the ratio how to split. On your ultrawide after opening two windows the windows will still be wider than tall. This means that the split will will be horizontal.

There is an open PR #44 that will allow you to set the ratio when it will split.

HirschBerge · 2022-11-13T15:48:54Z

This behaves as expected and what is also reported by #29. autotiling does not alternate between splits, it decides based on the ratio how to split. On your ultrawide after opening two windows the windows will still be wider than tall. This means that the split will will be horizontal.

There is an open PR #44 that will allow you to set the ratio when it will split.

Thank you!

Syphdias force-pushed the limit-depth branch from 4f7d070 to efb8f29 Compare October 27, 2022 18:19

nwg-piotr merged commit 79f4696 into nwg-piotr:master Oct 31, 2022

Syphdias mentioned this pull request Nov 1, 2022

Add option to limit to outputs #42

Merged

Syphdias deleted the limit-depth branch November 1, 2022 19:55

Add depth limit as argument for how often to autotile #41

Add depth limit as argument for how often to autotile #41

Conversation

Syphdias commented Oct 27, 2022 • edited Loading

nwg-piotr commented Oct 27, 2022

Syphdias commented Oct 29, 2022

nwg-piotr commented Oct 29, 2022 • edited Loading

Syphdias commented Oct 29, 2022

nwg-piotr commented Oct 29, 2022

Syphdias commented Oct 29, 2022 • edited Loading

nwg-piotr commented Oct 29, 2022

Syphdias commented Oct 29, 2022

nwg-piotr commented Oct 29, 2022 • edited Loading

nwg-piotr commented Oct 29, 2022 • edited Loading

Syphdias commented Oct 30, 2022

nwg-piotr commented Oct 30, 2022 • edited Loading

Syphdias commented Oct 30, 2022

nwg-piotr commented Oct 30, 2022 • edited Loading

Syphdias commented Oct 30, 2022

nwg-piotr commented Oct 30, 2022

Syphdias commented Oct 30, 2022 • edited Loading

nwg-piotr commented Oct 30, 2022

nwg-piotr commented Oct 31, 2022

Syphdias commented Oct 31, 2022

Syphdias commented Oct 31, 2022

nwg-piotr commented Oct 31, 2022

Syphdias commented Oct 31, 2022

nwg-piotr commented Oct 31, 2022 • edited Loading

nwg-piotr commented Oct 31, 2022 • edited Loading

Syphdias commented Nov 1, 2022 • edited Loading

nwg-piotr commented Nov 1, 2022 • edited Loading

nwg-piotr commented Nov 3, 2022

nwg-piotr commented Nov 3, 2022 • edited Loading

nwg-piotr commented Nov 3, 2022

Syphdias commented Nov 8, 2022

nwg-piotr commented Nov 8, 2022

Syphdias commented Nov 8, 2022

nwg-piotr commented Nov 8, 2022

Syphdias commented Nov 9, 2022

nwg-piotr commented Nov 9, 2022

Syphdias commented Nov 9, 2022

nwg-piotr commented Nov 9, 2022

HirschBerge commented Nov 13, 2022 • edited Loading

Syphdias commented Nov 13, 2022

HirschBerge commented Nov 13, 2022

Syphdias commented Oct 27, 2022 •

edited

Loading

nwg-piotr commented Oct 29, 2022 •

edited

Loading

Syphdias commented Oct 29, 2022 •

edited

Loading

nwg-piotr commented Oct 29, 2022 •

edited

Loading

nwg-piotr commented Oct 29, 2022 •

edited

Loading

nwg-piotr commented Oct 30, 2022 •

edited

Loading

nwg-piotr commented Oct 30, 2022 •

edited

Loading

Syphdias commented Oct 30, 2022 •

edited

Loading

nwg-piotr commented Oct 31, 2022 •

edited

Loading

nwg-piotr commented Oct 31, 2022 •

edited

Loading

Syphdias commented Nov 1, 2022 •

edited

Loading

nwg-piotr commented Nov 1, 2022 •

edited

Loading

nwg-piotr commented Nov 3, 2022 •

edited

Loading

HirschBerge commented Nov 13, 2022 •

edited

Loading