i3bar: protocol for workspace buttons #3818

xzfc · 2019-10-10T09:56:29Z

Proposal

I am proposing to add ws_command option to i3bar. It should be similar to status_command but for controlling workspace buttons instead of statusline blocks.

This would allow the user to define arbitrary complex customizations without introducing actual complexity into the i3bar code or i3bar configuration syntax.

Examples of customizations are:

Show empty workspaces feature request : show empty workspaces #2617
Hide certain workspaces Allow i3bar to not display some workspaces #2333
Groups of workspaces Feature request: add "groups of workspaces" #1808
Sort workspaces alphabetically sort i3bar workspaces alphabetically #3548

Examples of per-button customizations:

"min_width":32 (related to Minimal width for workspace buttons #3711)
"separator":true — personally I find it is useful to visually separate groups of workspaces.
"color":[...] — could be used to distinguish fake empty workspaces.

Diagram

Example configurations

bar {
	ws_command ~/bin/ws_command_default.sh
}

ws_command_default.sh - recreate the default i3bar behavior.

#!/bin/sh
i3-msg -t subscribe -m '["workspace"]' | {
	i3-msg -t get_workspaces;
	while read; do i3-msg -t get_workspaces; done;
}

ws_command_hide.sh - hide workspace named foo unless it is visible.

#!/bin/sh
i3-msg -t subscribe -m '["workspace"]' | {
	i3-msg -t get_workspaces;
	while read; do i3-msg -t get_workspaces; done;
} | jq -c '[ .[] | select(.name != "foo" or .visible) ]'

ws_command_show_empty.sh - show empty workspaces foo and bar on LVDS1 even if they do not exist at the moment.

#!/bin/sh
i3-msg -t subscribe -m '["workspace"]' | {
	i3-msg -t get_workspaces;
	while read; do i3-msg -t get_workspaces; done;
} | jq -c '
	def fake_ws(name): {
		num: -1, name: name,
		visible: false,
		focused: false,
		urgent: false,
		output: "LVDS1",
		"colors":["#191919", "#111111", "#444444"], # dimmed inactive_workspace
	};
	. + [ fake_ws("foo"), fake_ws("bar") ] | unique_by(.name)
'

The text was updated successfully, but these errors were encountered:

orestisfl · 2019-10-10T10:46:00Z

I like the idea and since we have the infrastructure to support IPC and JSON parsing this will not be very complex. @Airblader will have to approve though.

Btw, are you willing to implement this?

xzfc · 2019-10-10T10:55:27Z

Btw, are you willing to implement this?

Yep.

Airblader · 2019-10-10T11:31:55Z

Yes, I think this is a great, extensible solution. I'd say we for now start with a protocol that only covers what we currently need and leave additional features (such as separators) for follow-up work. So it would be good to define what that protocol could look like and then get the PR going.

Something we need to decide is whether coloring focused / urgent workspaces is left up to that interface as well or whether this will still be done by i3bar.

Airblader · 2019-10-10T11:33:30Z

Another thing, I would not support multiple protocols here like we do in the status command and go for JSON right away.

ammgws · 2020-05-28T15:19:39Z

@xzfc Just wondering if you made any progress on this?

ymolists · 2020-06-10T21:15:14Z

@xzfc I am trying to get clarifications around the design so please bear with me if i am missing some parts. The way i see it this is pushing the complexity to people who are customising i3 in a very significant way. Here is my use case :

bindsym $mod+F1 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 1 
bindsym $mod+F2 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 2
bindsym $mod+F3 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 3
bindsym $mod+F4 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 4
bindsym $mod+F5 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 5
bindsym $mod+F6 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 6
bindsym $mod+F7 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 7
bindsym $mod+F8 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 8
bindsym $mod+F9 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 9
bindsym $mod+F10 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 10
bindsym $mod+F11 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 11
bindsym $mod+F12 exec --no-startup-id /opt/i3/sbin/i3ws.sh --wsgrp 12

wsgroup is a grouping of wspaces with each having a unique id. Each func keys basically maps to one single ws group.

The logic to managing all these is handled by the script internally. The script does not store state outside of i3 for all the existing wsgroup at all. Instead it stores this metadata as part of the i3 wspace name so effectively the wspace name is <ws_id>:<ws_name>:<groupid>:<hide_suffix>
This script works because in my local i3 branch i added a way to hide a ws if the ws name has a fixed suffix as part of the name.

This proposal would make writing simple bindings/scripts like the one above i have more complicated because it has to be able to sync whit whatever ws_command_hide.sh script is called behind the scene.

Can someone tell me how i can make sure the binding i have above would work in the new scheme ? I might be missing something.

ymolists · 2020-06-11T01:53:08Z

I thought the i3ws.sh and ws_command_hide.sh had to be in sync and be coupled somehow. I am starting to see they actually do not need to. Starting to see the light !

xzfc · 2020-06-14T06:20:57Z

@ymolists
I see it this way:
Both i3ws.sh and ws_command_hide.sh have to be in sync in the implementation sense. That is, i3ws.sh should rely on the behavior of particular ws_command_hide.sh.

But the following much the same as what you describe.
They don't have state outside of i3. i3ws.sh would add some particular suffix to workspace names that have to be hidden; and ws_command_hide.sh would hide workspaces with this particular suffix.

RafalSkolasinski · 2020-07-17T09:19:03Z

Hi,

Is there any already existing mechanism that would allow me to hide specified workspaces on the bar?

I am very much looking towards that feature.

orestisfl · 2020-07-17T10:22:17Z

Hi,

Is there any already existing mechanism that would allow me to hide specified workspaces on the bar?

No, see the patches in #2333 for workarounds

ymolists · 2020-11-03T05:30:31Z

@xzfc Just wondering if you made any progress on this? I cant wait to start using an official supported way to hide workspaces.

Stardust-kyun · 2021-01-19T20:19:37Z

Would this only be available for i3bar, polybar's i3 module, and similar workspace indicators, or would it work with workspace indicators like the xworkspaces module for polybar? Not sure if I explained that well enough, but hopefully you can tell what I meant.

Airblader · 2021-01-19T20:20:55Z

This is purely i3bar, nothing else. Any other client is already free to display (or not display) whatever they choose.

Stardust-kyun · 2021-01-19T20:24:42Z

This is purely i3bar, nothing else. Any other client is already free to display (or not display) whatever they choose.

I actually forgot to specify what for! From what I understand, other clients like polybar are unable to display empty inactive workspaces because of i3 dynamically loading them, so I was wondering if the example of "show empty workspaces" would work with other clients.

orestisfl · 2021-01-19T20:31:10Z

The concept of "empty workspaces" does not exist in i3 since there are virtually infinite non-created workspaces. Third party bars can already choose to display whatever non-existent workspace they wish. The difference of the new protocol is that i3bar can do the same.

…

On Tue, Jan 19, 2021, 21:25 Stardust-kyun ***@***.***> wrote: This is purely i3bar, nothing else. Any other client is already free to display (or not display) whatever they choose. I actually forgot to specify what for! From what I understand, other clients like polybar are unable to display empty inactive workspaces because of i3 dynamically loading them, so I was wondering if the example of "show empty workspaces" would work with other clients. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#3818 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABMCZPVODOJMJDGGRLB6AYLS2XTBZANCNFSM4I7KS4JA> .

Airblader · 2021-01-19T20:31:57Z

Clients which strictly show workspaces based on EWMH will not show such workspaces since they don't exist and thus we don't list them in the hints. This feature won't change anything about that, no.

Stardust-kyun · 2021-01-19T20:48:24Z

Ah, I see. Do either of you know of any way that I can create dummy workspaces for xworkspaces or the like to somewhat create that?

Airblader · 2021-01-19T20:57:56Z

I don't even know that tool. Regardless, you can ask that question on the i3wm subreddit. Let's please keep the discussion here on topic.

Stardust-kyun · 2021-01-19T21:07:50Z

Alrighty, sorry for the confusion. Thanks!

MarcosCosmos · 2021-05-28T11:32:11Z

Clients which strictly show workspaces based on EWMH will not show such workspaces since they don't exist and thus we don't list them in the hints. This feature won't change anything about that, no.

I am confused here... Do you mean to imply that there is already a way for clients to query the configured (but not existent) workspaces known to i3, or, else that such a feature is out of scope?

I understand that erasing empty workspaces is long standing behaviour and I expect changing that would surely be a breaking change, but I get the (possibly mistaken) impression that you intend to make internal information available to i3-bar exclusively, which seems incongruous?

Third-party tools may be free to display anything they like but, and I am happy to be disproven: it seems to me that (perhaps like i3-bar) almost all, if not all, clients that aren't entirely manual/static in the content they display, are designed to assume that the WM has exclusive authority over what information can be be displayed in the layout, including workspaces. I would argue, in fact, that this is necessary, in order to display dynamic content (which is to say, all content, with how i3 is designed).

Airblader · 2021-05-28T11:35:00Z

There are no "configured workspaces". Workspaces are created when you switch to them and exist for as long as there are windows on them or they are active on an output. Otherwise they don't exist, and thus also cannot be queried. That's true both externally and internally – i3bar doesn't know about them either, they simply do not exist.

MarcosCosmos · 2021-05-28T11:36:29Z

I was referring to the "show empty workspaces" feature which I see on the OP, has that been excluded from this issue, then?

Airblader · 2021-05-28T11:41:50Z

Currently, i3bar pulls its information on which workspaces exist from i3, which is fed by its internal state. With this PR you have a protocol such that you can define yourself where this information comes from. This allows you to enrich the information you query from i3's IPC with non-existing workspaces, or replace it entirely, which means those workspaces would be shown in i3bar.

A feature like "show empty workspaces" cannot possibly be implemented, because empty workspaces don't exist in the first place. With this protocol you can choose how you want to implement it:

You could simply hard-code a list of workspaces to be shown in i3bar.
Alternatively, a list of hard-coded workspaces merged with those which exist based on i3's IPC response.
Or you could listen for new workspace events, build up a set of workspaces which existed but never remove from that list, which basically means your i3bar will show all workspaces which existed at some point.
…

MarcosCosmos · 2021-05-28T12:08:42Z

To clarify, I understand well enough that an empty workspace cannot exist in the sense that workspaces must be destroyed when they empty, as that is your convention.

However, it appears as though this PR would introduce an analogous or an adjacent concept, specific to i3, and only in respect to the information emitted via IPC, is that correct?

It seems like in your most recent comment, you are saying I would be able to feed artificial information into the IPC that other clients can then receive?

I do not use i3bar currently; I don't know it would suit my needs (this issue aside); But I am not actually aware of any client/bar that would allow me to configure a manual set of workspaces without managing all IPC interactions myself.

I recently switched to i3 to make use of it's mode switching, and found the lack of empty workspaces made it severely more time consuming to determine which, of the workspaces pinned to an output are free to use.

On that note, these pinned workspaces, I would argue, ARE in fact configured and, as far as client are concerned, differ only in that said configuration cannot be observed when the workspace is out-in-use.

The fact that this configuration is accessible only when in use is arguably part of the semantics of i3 and so, of course, any i3-aware client will naturally adopt the same model. The proposed IPC changes seem useful, but I don't actually see how they assist in the display of a pseudo-statically list of (potential) workspaces any more than not having these changes would? - It seems that it would not allow for any differentiation between a current and future workspace, without including third-party information tags, which standard clients would be unable to register.

In that sense it may actually introduce more problems than it solves.

Edit: it would probably be more useful to provide a feature that queries the output a non-existent workspace with a particular name has been configured to pin to, if any.

Airblader · 2021-05-28T12:11:42Z

It seems like in your most recent comment, you are saying I would be able to feed artificial information into the IPC that other clients can then receive?

No – this protocol does not feed the IPC. It feeds i3bar.

But I am not actually aware of any client/bar that would allow me to configure a manual set of workspaces without managing all IPC interactions myself.

I don't know of one either, but this is really outside our control. We will not implement a feature to keep empty workspaces, this is the most we want to do to support it, but it requires i3bar. For other clients you'll have to raise feature requests to allow specifying workspaces to be shown despite not existing.

MarcosCosmos · 2021-05-28T12:13:34Z

For other clients you'll have to raise feature requests to allow specifying workspaces to be shown despite not existing.

I made an edit to my last post that is relevant on this point:

Edit: it would probably be more useful to provide a feature that queries the output a non-existent workspace with a particular name has been configured to pin to, if any.

This pinning information is something that clients need to honour, to show the workspace on the correct bar instance; Without it, no client-side solution would be complete.

Airblader · 2021-05-28T12:15:51Z

I don't think I understand your edited-in sentence. Can you elaborate?

MarcosCosmos · 2021-05-28T12:20:27Z

Certainly. Consider, for example, the config line:

workspace "1:a" output HDMI1

This is persistent configuration information within i3 but exposed, to my knowledge, only when the workspace is in use.

It is, for the purposes of any IPC clients, ~~functionally equivalent~~ analogous to the presence of any empty workspace. I am not concerned in the slightest with what you expose via EWMH, but this output pinning is important information that I believe should be exposed via IPC.

Airblader · 2021-05-28T12:22:31Z

I think what you're looking for is something you will get from #4427.

MarcosCosmos · 2021-05-28T12:24:27Z

Yes, that would certainly suffice, thank you!

ymolists · 2021-06-30T17:15:23Z

I cant wait for this feature. Can someone summarise the state it is in now ? Thank you for all your efforts !

orestisfl · 2021-07-03T09:02:25Z

There is PR #4311 that mostly works but I need to handle some issues regarding reading input line-by-line. ETA unknown.

ymolists · 2022-08-20T03:51:02Z

@orestisfl i am still hopeful someone will review/complete that PR !!!

Closes #3818 (parent issue) Fixes #1808 Fixes #2333 Fixes #2617 Fixes #3548

Closes i3#3818 (parent issue) Fixes i3#1808 Fixes i3#2333 Fixes i3#2617 Fixes i3#3548

i3bot added the enhancement label Oct 10, 2019

orestisfl added the accepted Has been approved and is ok to start working on label Oct 10, 2019

orestisfl assigned xzfc Oct 10, 2019

orestisfl mentioned this issue Apr 5, 2020

WIP: Support layers for managing many workspaces #3986

Closed

3 tasks

orestisfl mentioned this issue Apr 23, 2020

Semantic Grouping of Workspaces in Layers/Groups #4024

Open

ammgws mentioned this issue Jun 2, 2020

Add a workspaces block greshake/i3status-rust#624

Closed

orestisfl mentioned this issue Jun 10, 2020

Allow i3bar to not display some workspaces #2333

Closed

orestisfl mentioned this issue Oct 11, 2020

Show empty workspaces Airblader/i3#117

Closed

Airblader added tentatively_accepted This feature has been previously accepted, but needs to be re-evaluated now. and removed accepted Has been approved and is ok to start working on labels Dec 31, 2020

orestisfl added accepted Has been approved and is ok to start working on requires-configuration This feature request requires new configuration and removed tentatively_accepted This feature has been previously accepted, but needs to be re-evaluated now. labels Jan 2, 2021

orestisfl assigned orestisfl and unassigned xzfc Jan 4, 2021

orestisfl mentioned this issue Jan 4, 2021

i3bar: Add protocol for workspace buttons #4311

Merged

10 tasks

ammgws mentioned this issue Jan 12, 2021

Workspaces block and Bindingmode block greshake/i3status-rust#980

Closed

orestisfl mentioned this issue Jan 12, 2021

Feature request: Configurable workspace namespacing #3885

Closed

ammgws mentioned this issue Jan 31, 2022

workspace underline greshake/i3status-rust#1418

Closed

orestisfl added a commit that referenced this issue Jan 22, 2023

i3bar: Add protocol for workspace buttons

ff30991

Closes #3818 (parent issue) Fixes #1808 Fixes #2333 Fixes #2617 Fixes #3548

orestisfl added a commit that referenced this issue Jan 22, 2023

i3bar: Add protocol for workspace buttons

ba1f40f

Closes #3818 (parent issue) Fixes #1808 Fixes #2333 Fixes #2617 Fixes #3548

orestisfl closed this as completed in #4311 Jan 22, 2023

orestisfl mentioned this issue Jan 22, 2023

i3bar: Support customizations in workspace buttons protocol #5387

Open

desmana pushed a commit to desmana/i3 that referenced this issue Feb 27, 2023

i3bar: Add protocol for workspace buttons

aeaf00f

Closes i3#3818 (parent issue) Fixes i3#1808 Fixes i3#2333 Fixes i3#2617 Fixes i3#3548

jbenden pushed a commit to jbenden/i3 that referenced this issue Jun 7, 2023

i3bar: Add protocol for workspace buttons

e710930

Closes i3#3818 (parent issue) Fixes i3#1808 Fixes i3#2333 Fixes i3#2617 Fixes i3#3548

orestisfl mentioned this issue Sep 21, 2023

Release 4.23 #5681

Closed

1 task

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

i3bar: protocol for workspace buttons #3818

i3bar: protocol for workspace buttons #3818

xzfc commented Oct 10, 2019

orestisfl commented Oct 10, 2019

xzfc commented Oct 10, 2019 via email

Airblader commented Oct 10, 2019

Airblader commented Oct 10, 2019

ammgws commented May 28, 2020

ymolists commented Jun 10, 2020 •

edited

ymolists commented Jun 11, 2020

xzfc commented Jun 14, 2020

RafalSkolasinski commented Jul 17, 2020

orestisfl commented Jul 17, 2020

ymolists commented Nov 3, 2020

Stardust-kyun commented Jan 19, 2021

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

orestisfl commented Jan 19, 2021 via email

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021

ymolists commented Jun 30, 2021

orestisfl commented Jul 3, 2021

ymolists commented Aug 20, 2022

i3bar: protocol for workspace buttons #3818

i3bar: protocol for workspace buttons #3818

Comments

xzfc commented Oct 10, 2019

Proposal

Diagram

Example configurations

orestisfl commented Oct 10, 2019

xzfc commented Oct 10, 2019 via email

Airblader commented Oct 10, 2019

Airblader commented Oct 10, 2019

ammgws commented May 28, 2020

ymolists commented Jun 10, 2020 • edited

ymolists commented Jun 11, 2020

xzfc commented Jun 14, 2020

RafalSkolasinski commented Jul 17, 2020

orestisfl commented Jul 17, 2020

ymolists commented Nov 3, 2020

Stardust-kyun commented Jan 19, 2021

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

orestisfl commented Jan 19, 2021 via email

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

Airblader commented Jan 19, 2021

Stardust-kyun commented Jan 19, 2021

MarcosCosmos commented May 28, 2021 • edited

Airblader commented May 28, 2021 • edited

MarcosCosmos commented May 28, 2021 • edited

Airblader commented May 28, 2021 • edited

MarcosCosmos commented May 28, 2021 • edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021 • edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021 • edited

Airblader commented May 28, 2021

MarcosCosmos commented May 28, 2021

ymolists commented Jun 30, 2021

orestisfl commented Jul 3, 2021

ymolists commented Aug 20, 2022

ymolists commented Jun 10, 2020 •

edited

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited

Airblader commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited

MarcosCosmos commented May 28, 2021 •

edited