search

ActivityWatch / aw-watcher-window

Cross-platform window watcher (for use with ActivityWatch)
Mozilla Public License 2.0
91 stars 53 forks source link

fix: ignore loginwindow events #88

Open iloveitaly opened 1 year ago

iloveitaly commented 1 year ago

workaround for #85. Yes, we should do the polling solution we outlined, but this has worked for me locally the last couple weeks and I don't have the time for anything else right now :)

could we get this in as a stopgap until someone has time for the correct solution?

iloveitaly commented 1 year ago

@ErikBjare friendly reminder here!

iloveitaly commented 1 year ago

@ErikBjare friendly reminder here!

ErikBjare commented 1 year ago

I'm not sure how I feel about this.

Kinda feels like sweeping the problem under the rug. Merging this PR would lead to otherwise unexplained gaps (unless you happen to look at logs).

I would prefer a workaround on the front-end, such as adding a "ignored pattern" feature (similar to the "always count as active pattern" feature). Which is a wanted feature anyway.

iloveitaly commented 1 year ago

@ErikBjare what do you think about "filling in" AFK events as outlined in #85? It feels like we should have a AFK timeline, even if the computer goes to sleep.

This would fix this issue without (a) adding the complexity of polling in the swift app and (b) without hacky exclusion rules.

iloveitaly commented 1 year ago

@ErikBjare friendly bump here! Curious your thoughts on the above Q.

theolundqvist commented 1 year ago

I'm not sure how I feel about this.

Kinda feels like sweeping the problem under the rug. Merging this PR would lead to otherwise unexplained gaps (unless you happen to look at logs).

I would prefer a workaround on the front-end, such as adding a "ignored pattern" feature (similar to the "always count as active pattern" feature). Which is a wanted feature anyway.

Hello! Is there any such feature? I would like to ignore the terminal window completely since I use the tmux watcher. Otherwise i get double the time each day. How is this usually solved when using editor watchers?

ErikBjare commented 1 year ago

@theolundqvist It's fairly trivial to implement (esp if you look at the audible-as-active PR for guidance). You could probably do it yourself! If not up your alley, I will try to get around to it Soon™.

Also, cool to see someone in the same city as me randomly on the internet. If you wanna grab lunch sometime, let me know :)

ErikBjare commented 1 year ago

@iloveitaly Filling in AFK events when the device is asleep/off is not something we want to do (we used to!), as it makes it very difficult to know:

Looking at this, I think that there might be an issue with the window watcher. If the computer is off (unable to pull the active window), it should not fill the time when the device was off.

The window watcher should not use a high pulsetime to fill these gaps, but instead rely on the poll-augmentation to expect a recent last event, which is updated with a smaller pulsetime.

theolundqvist commented 1 year ago

@ErikBjare thanks, I took another approach for now but I may check into that PR. I'am currently working on a fork of the aw-watcher-tmux by akohlbecker to make it work more like an editor watcher. I use neovim as an editor but wanted to track the time I am spending in the git repo no matter if I am in neovim or not. My goal is to show today's total time spent in the current git repo in the tmux statusbar. I think I will be able to get it done.

I plan to also include short breaks and focus shifts to browser etc as work in the repository but that will be done in post. I guess it is best to not augment the data directly in ActivityWatch.

Thanks for the invitation. I recognised the founders names as swedish but did not know that ActivityWatch was founded by LTH students, cheers!