Feature Request: Separate activity bucket per monitor

[hrobertson]

Thank you 2e3s for providing awatcher!

I have multiple monitors and often work actively using a window on each monitor. Switching focus back and forth between each monitor results in lots of short activity segments which is difficult to parse in the timeline.

Another scenario is when I have a video call open on one monitor but a window on the other monitor has focus. When reviewing my time later, it's very difficult to see how long I was in the video call despite it occupying a whole monitor for a decent amount of time.

I'd therefore love to have the (configuration) option to have each monitor (wl_output?) represented as it's own activity bucket.

Do you think this is feasible? If you'd be able to point me in the right direction I can look into contributing a PR.

Thanks

Edit:
I should have mentioned I'm using KDE Plasma.

[hrobertson]

I see the KWin script. I'll have a play with that.

Edit: I see I can get the client.output.name in the KWin script, so just a case of updating the watcher to support that and identifying if the output/monitor can be determined by the other watchers as well to maintain feature parity.

[2e3s]

I'd therefore love to have the (configuration) option to have each monitor (wl_output?) represented as it's own activity bucket.

Hello! Thank you for your suggestion, it sounds interesting! But unfortunately I think it is not feasible in most environments because the functionality the watcher is based on most environments provide: the concept of "active window". If you leave a window with a call on another monitor and work in the other, the window loses the focus and it's not active anymore. It can be on top of other windows, or there can be 2+ windows arranged side by side (e.g. tiled), equally visible. So the only thing which environments can provide consistently is the window in focus.

[hrobertson]

I'm still using the concept of the active window, but reporting on the last active window per-output and increasing the heartbeat pulse time (to a few minutes) so that switching back and forth within the pulsetime it registers as continuous activity.

You're right in that this is an attempt to address the problem of recording simultaneously visible windows and that problem isn't limited to multi-monitor setups, and this approach doesn't solve that problem in a generic way. I can imagine a solution for single monitor setups based on window coordinates but that would need to take a more nuanced approach to determining buckets. Discrete monitors makes it easy to assign activity to a finite number of buckets.

This solution seems to be working for me now. I'll see how I find it over the next few weeks.

[2e3s]

reporting on the last active window per-output and increasing the heartbeat pulse time (to a few minutes) so that switching back and forth within the pulsetime it registers as continuous activity.

Yes, it essentially smooths out your activity at the cost of precision. I had thought about running separate instances awatcher for 2 monitors, and what can be done internally to help with that, e.g. some environment variable with a monitor number, but it all stops at the idea of focusing. AFAIK, no DE has a concept of a focus per output (monitor, wl_output in Wayland). The only way you can make it working is running a virtual machine (QEMU/VBox/etc, some options are more seamless than others) and dedicating one monitor to it. A separate instance of OS will keep track of its window focus alright.

[hrobertson]

I have the precision as well - I have not replaced the original behaviour, only added additional per-monitor buckets.
When looking back at any given moment in the timeline I can see which window had focus, and which window last had focus on each monitor. Plus if focus remains on a single monitor for over 5 minutes the other monitor's timeline becomes blank from that time. With that additional context I am far better able to determine which project I was working on at any given time.

Personally I'm not too concerned about which window had focus. I care about being able to determine what I was working on.

[2e3s]

Ah, do I understand correctly that making a bucket name configurable with optionally a desktop number is what you are saying? Like one instance of the watcher sending the data to bucket user-linux-{$monitorNumber} where the variable is the monitor where the focused window is located?

[hrobertson]

Probably easier to show you what I've done: https://github.com/hrobertson/awatcher/tree/feat/output-buckets

This was my first touch of Rust and priority was just getting something working to experiment with hence the 300 second duration being hardcoded and only supporting kwin.

Here you can see third row is active window (the original behaviour). Bottom two rows are my two monitors with VSCode open on one and Gitkraken on the other.