triplekill / watchme

Window usage data collector & analyzer for Microsoft Windows

watchme

Window usage data collector and analyzer script for MS Windows. AKA an "active window logger" or "active window tracker"

Written by Jonathan Foote ([e-mail])

Description

watchme.py is a Python script for MS Windows that polls to collect information about the active window. The script can be controlled via a the task tray interface it adds the system tray after it is launched. In addition to tracking window information, it also includes some basic data analysis capabilities.

WARNING!

This script will collect data about what you do on your system and store it in plaintext to your local filesystem. USE THIS SCRIPT AT YOUR OWN RISK -- there is no warranty, and I take NO responsibility for what you do with it, or anything that happens to you as a result of using it.

Usage

Run watchme.py via python from the commandline:

python watchme.py

Or, to "install" it, create a batch script to run it with pythonw and add a shortcut to the batch script to your Startup folder.

pythonw watchme.py

To see if it is working, try this (assuming you have cygwin, git shell, or tail otherwise installed):

tail -f data/\ windows.csv

... as you click around you should see raw window activity info logged to your shell

Dependencies

Requires pywin32: http://sourceforge.net/projects/pywin32/

Background

This is a toy application that I occassionally work on for fun, largely as a pastime when I end up having to watch a chick flick or reality TV :). Please feel free to hack it up.

Design notes

Numbers

---

This script logs activity to CSV files. I've been running it on a machine that gets moderate to heavy use every day for the past 308 days (as of 9/8/2013), and the CSV files take up less than 100MB of disk space. From the outset I figured the current implementation of window logging would take up a little over 100MB per year, and based on data so far I think that is accurate.

Polling

---

This script uses polling to grab window activity. I usually try to avoid polling in favor of event-driven design, but after reading a bit on methods for logging window activity (and implementing some tests) I went with polling for these reasons: 1) I had to build a DLL to support handling win API callbacks, which complicated the build. 2) The win 32 API calls that I was playing with didn't cover all of the events I needed -- certain events, like minimizing a window, didn't trigger callbacks. 3) According to a 2012 (or was it 2011?) blog post the team from "time cockpit", who do this for a living, use polling too, so at a minimum it probably will be usable (even if it is not the best solution).

TODOs

• Daemonize & survive Windows sleeping (works with python.exe, but not pythonw.exe)
• Add idle time logging
• Gather licenses for included open source code
• Add license to text to readme and warn about privacy implications
• Clean up and comment code
• Add unit tests
• Add install/egg
• Beef up readme to include example and internals info
• Make icon
• Publish to github
• Add support for exe name processing to analyzer
• Add support for idle_time/window_time processing to analyzer
• Add additional analysis: histrograms, search by exe name, etc.
• Implement lock/some sort of singleton to prevent running proc twice
• Add ML :)
• Add support for aggregating data across machines (for VM use)
• Add support for other desktop managers
• Become a rich philanthropist, etc.

Licenses

See LICENSE.txt for details on the licenses of this project and included open source code