popylar : track software use with Google Analytics

Popylar is a small Python package that allows tracking execution of Python code (such as imports) via the Google Analytics API. For example, once you have set up a google analytics account, you could track the import of your package by adding this to the __init__.py file:

# In package __init__.py
import popylar
popylar.track_event(google_analytics_id, 'import', 'import_mypackage')

With this setup, every time the package is imported it will be tracked on Google analytics.

Why this Package?

Tracking the use of open software is, in general, very difficult, and with many open source projects being supported by private and public funding agencies, a means of more quantitative reporting would be beneficial. Current tracking options exist, but are relatively limited:

• PyPI tracks downloads, but cannot track when packages are installed with downstream package managers like conda, brew, apt, and others.
• Software hosting services like GitHub provide information about project forks, but no information about number of downloads or clones.
• Other means of tracking (mailing list traffic, bug report traffic, StackOverflow questions) are helpful but difficult to quantify and compare between projects.

More fine-grained user tracking is accepted practice for blogs and other websites, and popylar seeks to add that capability to software packages.

Wait... is this really OK?

Short answer: we're not sure. We agree that this feels a bit icky at first-blush, and we are open to feedback about the ethical issues or potential abuse of such a setup. That being said, the information tracked by popylar is identical to that tracked by any site using Google Analytics, including likely your favorite independent blogs, and popylar solves a real problem faced by many developers of the open source tools you probably depend on.

What information is tracked?

The tracked information is exactly that stored by Google Analytics when you visit a website, with the addition of a few pieces of metadata that can be configured by the package. In particular, the only identifying information provided by the Python package is a randomly-generated unique ID.

Can a user opt-out?

Yes. If a user runs popylar.opt_out(), popylar will never track usage for that system/user, and this opt-out state is preserved between sessions.

Can I still use packages off-line?

Yes. If no internet connection is detected, no information will be reported by popylar and the script will continue to execute as normal.

Where can I give feedback or offer ideas?

If you would like to weigh-in on this idea, please use Popylar's Github Issues. We'd love to hear your thoughts, particularly on any ethical or technical issues we may be overlooking.