Jacob Kaplan-Moss

Measuring the Django Community:

The Django community in 2012

I wrote this post in 2012, more than 12 years ago. It may be very out of date, partially or totally incorrect. I may even no longer agree with this, or might approach things differently if I wrote this post today. I rarely edit posts after writing them, but if I have there'll be a note at the bottom about what I changed and why. If something in this post is actively harmful or dangerous please get in touch and I'll fix it.

In 2007, and again in 2009, I made an attempt to measure the size of the Django community. By popular request — okay, a couple people asked for it, whatever — let’s do this thing again.

Users

In 2007 and 2009, I shared three ways of looking at how many people are using Django: hits to the website, downloads of the Django tarball, and sites listed as “using Django.”

So, here’s an overview of users, some notes on interpreting these numbers follow:

Metric200720092012
Website hits per month1m4.7m6.0m
Downloads per day1,0001,0002,500
Powered sites5002,700

Notes:

  • Similar to 2009, website hits per month are split roughly 60/40 between the docs (docs.djangoproject.com) and the main website (www.djangoproject.com).

  • Download numbers have jumped a lot, but there’s some data to unpack there:

    First, I’m now reporting downloads from PyPI (as collated by Crate) – about 1,500 per day – as well as direct downloads – another 1,000. PyPI wasn’t gathering stats in 2009 (or if it was I didn’t know how to find them), so that accounts for some of the jump.

    On top of that, the last few years have seen a dramatic rise in the use of virtualenv. This means that instead of installing Django once and using it for a bunch of projects, people are probably installing Django multiple times per machine.

  • I’m dropping the last metric, “powered sites”: we haven’t had an official “who’s using Django” list since roughly 2008, and djangosites.org only records a small fraction of sites using Django. Since this is no longer a number I can gather with any degree of accuracy, I’m going to skip it.

To me, these numbers show something like a 25% increase in Django’s user base since 2009. This fits my gut feeling: we saw explosive growth between 2007 and 2009, but since then the rate’s slowed somewhat as Django’s moved from an “early adoptors” tool to an established player in the web development ecosystem.

One new thing in 2012 is that suddenly there’s a nearly endless list of success stories. In 2009 we had a handful of high-profile Django users, but today the high-profile uses of Django read like a Who’s Who of the Internet. Check this list out: AMD, Canonical, Discovery, Disqus, HP, IBM, Instagram, Intel, Lexis-Nexis, the Library of Congress, Mozilla, NASA, National Geographic, the New York Times, Orbitz, PBS, Pinterest, Rdio, VMWare, Walt Disney, and the Washington Post.

Not bad for a few nerds hacking in a basement in Lawrence, Kansas, eh?

Community

I recorded two numbers in 2007 and 2009: membership on django-users and membership on django-developers. To those numbers, I’ll one more: number of Django-related packages on PyPI. I have no historical numbers of that, but the next time I do one of these I can update it.

Let’s see…

Metric200720092012
django-users members4,50016,00021,700
django-developers members2,0005,1007,000
Packages on PyPI2,000

Notes:

  • Mailing list membership numbers are measured same as last time. I still think Google Groups is broken for lists of this size.
  • The “packages” number comes from looking at the number of packages tagged Framework :: Django on PyPI. This likely understates the number a bit – not everyone remembers to use the framework tag.

Again, still seeing growth, though not as extreme as for the ‘07 - ‘09 period.

One interesting tidbit: PyPI currently lists about 20,000 packages, so this means that Django accounts for about 10% of the activity on PyPI!

Contributors

In past years, I recorded three numbers: people listed in AUTHORS, number of people with some level of commit access, and full core committers. We’ve seen lots of growth here, too:

Metric200720092012
AUTHORS160447524
Partial committers103633
Full committers51528

Since 2009 we’ve gotten more aggressive in giving out the commit bit. Things stalled out a bunch around 2010, and about a year or so ago we started correcting that. In particular, we’ve often been giving out full commit access instead of partial. This is reflected in the number of partial committers not increasing (it fluctuates because add/remove GSoC students) but the number of full committers going up a bunch.

(We actually have a few other candidates in the pipeline, so I expect this number to go up even more soon!)

Update, March 7

Indeed; we’re adding two new full committers this week. Hooray!

This was fun. We should do it more often!