#FlowRatio Tracker

A work sampling based time tracker. That runs in the browser with no additional anything. Try it out: https://xkqr.org/flowratio/

Here's some background: https://two-wrongs.com/flowratio-stochastic-time-tracking.html

#Usage

Open the web page, and allow notifications. If it says "PENDING", go do whatever you were going to do next. At some point in the future, you'll get a notification and it will ask you about what you are doing. Write down a short keyword that indicates what you are doing RIGHT NOW.

If you want to structure your keywords, you can use a colon to separate a subcategory from a main category, like "work:docs" for writing documentation at work.

Once you've done this a few times, you can click on report to see how you spend your time.

#Tips

You don't need to track everything you do in great detail! If you're curious e.g. how much time you spend in meetings, you can just tag "meetings" and "other" for a few weeks and you'll find out. Or if you want to increase the amount of time you spend on writing documentation, you can just tag "docs" and "other" for a while and try to make the number go up.

#Development

If you want to test your changes, open index.html in a browser and it should work. Maybe you have to go through a separate web server like python -m http.server 8001.

In development you might want faster feedback. Add ?dev=1 to the URL to get much more aggressive time constants. If you want to clear out the data because you have toyed around, load it with ?clear=1 – but remember to remove that before you refresh in case you collect data you want to keep.

#Todo

#Features, Analytics Application

• [x] Percent of time spent on each high-level category with error bars
• [x] Total time spent on each thing?, with error bars
• [ ] Timespan selector for above display – this takes figuring out how to re-render the flame graph without breaking everything
• [x] Allow selecting a category and the above happens for only entries in that
  • Even cooler would be a flamegraph – can I get that to happen? Could a flamegraph sort of supersede everything else I want to do? I think it might... EXCEPT it doesn't have error bars! How do I get error bars on that? Or should I indicate insecurity some other way?
• [x] Convert displays to pretty graphical plots
• [x] Migrate history display to analytics
• [x] Won't do: migrate CSV export/import to analytics. The reason is that I want analytics to be read-only, and the main application is the only thing that writes to the data.

#Bugs, Analytics Application

• [ ] Discoverability of the colon syntax is nonexistent

#Features, Main Application

I think these are the bare minimum needed to reach the dangerous dogfooding point.

• [x] Show notification when sampling
• [x] Randomise interval between samples
• [x] Export to JSON
• [x] Change export to CSV
• [x] Autodetect when running again after sleep, closed tab, etc. Add "off" tagged Off means flowratio was off, not anything else. This can be accomplished practically by storing the time for the next sample in localstorage and then if now is singificantly greater than nextping, it means the sample happened when flowratio was unable to bug the user, i.e. "off"
  • Autodetection in place, remains to be tested
• [x] Set sane default times
• [x] Stop clearing localstorage automatically

There are also some nice-to-haves.

• [x] Import from CSV
• [ ] Autocomplete when typing
• [ ] Nice design

I can sense at least one advanced feature unlikely to happen.

• [x] Allow user to edit the recent history of tags directly in the application. The user can do things like fix spelling errors by export/import. - But aaaactually, I fairly often get pings when I'm presenting things in meetings and it would be really convenient to be able to change the afk to a "work:meeting" tag after, so I'll add this feature anyway.
• [ ] Give user ability to clear out log– the user can do this by means of export/import.

#Bugs, Main Application

These bugs are apparent.

• [x] When AFK, it gets into an infinite loop somehow

• [x] Fan speed goes up when using – caught in rerendering loop or just tired computer?

  • I'm going to chalk this up to tired computer because I cannot see much CPU usage in about:performance nor in the profiler

• [x] Notification disappears after a brief while, despite requireInteraction: true – why is that? I'd prefer it to time out with the afk auto-submission.

  • I think this might just be the GNOME notification system, because on Windows it persists until the afk auto-submission.

• [x] Subsequent prompts triggered when afk should be auto-tagged afk, not flowratio. That said, this is probably a low-prio bug since under realistic settings it will rarely happen. (But it does happen all the time under development, so it's easy to reproduce!)

• [x] When my Windows laptop is closed, FlowRatio can generate both "off" and "afk" events, sometimes at the same time. Something something modern standby and how do I prevent it? - This happens also on Linux and I'm now thinking it's not that it does things while asleep, but rather that when FlowRatio catches up, everytime a new expired ping time is generated, the submission form races to auto-submit an AFK before the generation code submits an OFF. Sometimes the form wins handily and the OFF never comes in, sometimes the OFF wins and the AFK never comes in, and sometimes – and this was the clue – the AFK happens instants before the OFF and both are submitted – but never the other way around! - This seems to indeed have been the problem. I haven't gotten any double submissions now.

• [x] When my Windows laptop is closed, FlowRatio seems to generate an excessive amount of pings, on average every 32 minutes. This might not sound far from 45 minutes, buuut unless I'm doing statistics wrong, this should happen extreemely rarely by accident. - For some reason, this may have been fixed now as well. I'm now consistently getting ping counts within 1–2 standard deviations. Not sure why fixing the above problem fixed this too, or if I'm just been lucky since then. It's worked long enough that I'm going to call it fixed for now.

• [ ] Sometimes during night (when the Windows machine restarts, maybe?) ten tags from during the day are somehow copied onto the most recent off/afk samples. This is relatively easy to correct manually (export, change the tags to off, import again) but it's annoying and confusing!

  My best guess is that it has something to do with the in-line sample
  editing functionality, that it somehow saves the value of the text inputs
  on top of the just-generated off pings, when maybe focus automatically
  moves away from those boxes for some reason?
  
  It does indicate that I should be more careful about using "unfocus as
  save" for this sort of thing. Maybe I should add an explicit save button
  just to rule this out as the cause.
  

These bugs are latent.

• [ ] Set some sort of mutex in local storage that prevents two flowratio tabs from being active at the same time. Maybe just have a random id generated on page load, and store that as the active tab in localstorage. Each second, read from localstorage and ensure this tab is still active. Otherwise blank out and ask user to refresh to reset to this tab.
• [x] What happens if one desires to adjust lambda while using flowratio? Maybe the lambda rate needs to be stored along with the sample time everywhere. (That lets us translate the sample into "1/lambda minutes of work")
• [ ] If flowratio has been off for a long time it will sit in a blocking loop to generate all the "off" samples. It might be possible to generate a single "off" sample with a fat lambda instead. How big of a problem is it? Someone starting up flowratio after a year with a lambda of 45 minutes will have 12,000 "off" samples generated. I'm going to say that this is not going to be a problem for at least six months.
• [ ] Changes to currentTime re-renders the entire application. With small history a-re-render-a-second is fine but I should break the time-dependent parts out from the ones that don't care about time. And/or memoize the history so at least not that re-renders every second but only when it changes.

This is a fun refactoring project I might play with at some point.

• [ ] Convert from oldschool setTimeout etc to moderner async/await code?