To celebrate the release ofPython 3.7.3 on25th March 2019, and with undernine months left for Python 2, here’s some statistics showing how much different Python versions have been used over the past three years.

Here’s the pip installs for all packages from thePython Package Index (PyPI), between January 2016 and March 2019:

pip

The package installer

six

Python 2 and 3 compatibility library

NumPy

Scientific computing library

pytest

Testing framework

Coverage.py

Code coverage testing

Pillow

Imaging library

Django

Web framework

Matplotlib

2D plotting library

Flake8

Linter

Pylint

Linter

pylast

Interface to Last.fm

How

Statistics were collected usingpypi-trends.py, a wrapper aroundpypinfo to fetch all monthly downloads from the PyPI database on Google BigQuery and save them as JSON files. Data was downloaded over several days as getting all months uses up a lot of free BigQuery quota. Thenjsons2csv.py plots a chart usingmatplotlib. Raw JSON data is in therepo.

See also

• Data Driven Decisions Using PyPI Download Statistics
• Python version share over time, 1 (January 2016 — June 2018)
• Python version share over time, 2 (January 2016 — October 2018)
• Python version share over time, 3 (January 2016 — December 2018)
• PyPI Stats: See package download data for the past 180 days, without needing to sign up for BigQuery
• pypistats: A command-line tool to access data from PyPI Stats