Consider upstreaming plasma layout

[nazar-pc]

I've been using https://github.com/numirias/qtile-plasma for years, it is quite awesome!

However I have difficulties upgrading with breaking changes introduced in Qtile since I need to patch my local version of qtile-plasma with more and move fixes.

Please consider upstreaming such an awesome layout, it is basically an advanced version of Tile and the only layout I use 99.9% of the time.

[tych0]

We'd be happy to accept a pull request with the code :)

[nazar-pc]

Okay, I'll try to create one when I'm in the right mood

[jrwrigh]

I made a similar request a few months ago: #4487, but there was a pretty mixed reaction to it's inclusion in the main repository. Happy to see that @tych0 is interested in bringing it over, but I'm curious if @m-col or @elParaguayo have changed their thoughts.

I say this not to try stop this from happening, but just to make sure everyone's on-board with the idea before @nazar-pc (or whoever else) puts in the work to make a PR.

[elParaguayo]

@jrwrigh I'm getting a weird error and I'm curious if you see the same. Open qtile shell. (Assuming Plasma is the active layout) do cd layout and then mode_vertical(). Try opening a new window and see if you get an error message appear in the qtile shell session.

The code seems to work but I've no idea why I get an error message. Sorry, can't paste the message at the moment as I'm at work!

[jrwrigh]

TBH, I don't use qtile shell ever (maybe I've been using qtile wrong for years... haha).

But doing I think what you asked:

❯ qtile shell
 > cd layout
layout
layout > mode_vertical()
layout >

And then opening a new window (not within qtile shell), I do not get an error in the shell or in the log. Note I'm still on qtile 0.24.0 and am on my own fork which incorporates several fixes (not sure if you're going off of that or the main repo right now).

[elParaguayo]

OK. Thanks for checking.

[jrwrigh]

Curious, what was the error message?

Edit: To clarify above, I'm on my own fork of Plasma, not qtile itself.

[elParaguayo]

It's unrelated to the layout, it's a side effect of an earlier commit (@tych0 - I'll post in your PR about this)

[elParaguayo]

It does have some tests, though in 5 minutes of playing around with them I couldn't get it running. I'm not super familiar with pytest or the qtile testing framework (which it appears to be using).

Wow. There are a lot of tests here. Really helpful for learning what the code does. Slowly working my way through it still.

[jrwrigh]

You mean if it tries to add a fixed node as a child? ie.

assert(child.fixed)
parent.add_child(child)

[jrwrigh]

I think that is true to my experience. If I have a fixed width child node and move it to another group (which iiuc would call add_child on the new group's parent node), the fixed width is not respected in the new group.

I don't have much of an issue with that behavior. Most of the time if I have a fixed width window, it's because I've changed the size of a split from it's equispaced point. If I'm sending that to another group, I don't really care about the size in the new group.

[elParaguayo]

Consider this example:

>>> from libqtile.layout.plasma import Node
>>> Node.min_siize_default = 10
>>> root = Node("root", 0, 0, 100, 100)
>>> a = Node("A")
>>> b = Node("B")
>>> c = Node("C")
>>> root.add_node(a)
>>> root.add_node(b)
>>> print(f"{a.width=} {b.width=}")
a.width=50.0 b.width=50.0
>>> a.width = 60
>>> print(f"{a.width=} {b.width=}")
a.width=60 b.width=40.0
>>> print(f"{a.fixed=} {b.fixed=}")
a.fixed=True b.fixed=False
>>> root.add_node(c)
>>> print(f"{a.width=} {b.width=} {c.width=}")
a.width=56.66666666666666 b.width=21.66666666666667 c.width=21.66666666666667

I would still expect node A to have a width of 60 with B and C being 20 each. There's no reason for A to have been shrunk.

[elParaguayo]

Think I've fixed it.

EDIT: Hmm. I suppose it depends on what you think the appropriate behaviour is:

• Assume you've got a root that is 120 wide and a node with a minimum width of 10
• Add 2 nodes and they'll both be 60 wide but flexible
• Set the first to 80 and the second to 40. They're now both fixed with.
• Add a new node. QUESTION: how wide should that node be? There are two options
  • 40 (i.e. container width / 3) - but this makes a big change to the existing nodes which had a fixed size
  • 10 (i.e. the minimum width) - we have two fixed width nodes already so we should make as small a change as necessary

The layout seems to do the first but I think we should do the second.

[jrwrigh]

I'm fine with either tbh, but the 2nd option makes the most sense, at least in a vacuum. These are edge cases that I don't run
into very often.

[elParaguayo]

Draft PR: #4760

[elParaguayo]

This has now been merged.

[elParaguayo]

Closing as the layout has now been merged.