The trek from Python 2 to Python 3 has been drawn-out, arduous and fraught with perils. How close are our dear Knights developers to all reaching the long sought glory of Python 3?
PIP Downloads
Let’s first jump into what is being used the most currently. This data examines fifteen different libraries downloaded via PIP for a particular Python version. We are only including 2.7 and 3.4+, the Python Versions that are currently supported.
The libraries analyzed are ones that have over 10K stars on github and have been downloaded via PIP. The contenders are: celery, django, flask, ipython, keras, mitmproxy, numpy, pandas, python-box, requests, scrapy, selenium, tensorflow, and tornado. (To be fair, numpy and python-box didn’t have 10K stars, but I used them in the script to make these graphics, so gave them some spotlight too.)
As of January 2019, Python 3 downloads are eclipsing Python 2 by over 20% with Python 3.6 bringing over 39% of it, almost directly matching Python 2.7’s total.
That is good, but not great news. Thankfully Python 2 won’t just stop working at the end of this year, but those are rookie Python 3 numbers, we got to pump them up!
Of course, we have to remember this is a small subset of all downloads. Subsequently, pip downloads themselves don’t tell the whole tale, but this does give us an idea of how things of are going.
This is accomplished by using the PyPI BigQuery data and some SQL (adapted from Artem Golubin’s post about this from last year), then throwing it into matplotlib.
SELECT
SUBSTR(details.python, 0, 3) as python_version,
COUNT(*) as download_count
FROM
TABLE_DATE_RANGE(
[the-psf:pypi.downloads],
DATE_ADD(CURRENT_TIMESTAMP(), -30, "DAY"),
CURRENT_TIMESTAMP()
)
WHERE
details.installer.name='pip' and
file.project = 'requests' -- change project name here
GROUP BY
python_version
ORDER BY
download_count DESC
LIMIT 100
Library Brawl: Who’s the Python 3 champs?
In this head to head, we are going to compare two similar libraries, and see how they are doing on the switch to Python 3.
Web Frameworks
The first two up are very popular web frameworks to develop in, Flask and Django.
It’s a dead heat! Both libraries are doing well at attracting developers with a fresh mindset.
Machine Learning
The most popular github package by far was tensorflow with over a hundred thousand stars. Here it’s paired against it’s younger brother keras, which actually depends on it (or other AI tools) to operate.
Machine learning needs to teach it’s developers how to update! It’s a sad day for AI.
Hacker vs Web Scraper
Okay, not really directly comparable tools with a man-in-the-middle proxy and a web scraper, but it’s still an interesting match up.
With this duo I was surprised they didn’t have a higher correlation. I was honestly expecting the mitm tool to have less Python 3 love, as a lot of “hacker” tools depend on the broken way Python 2 handles strings vs unicode, thus are hard to update.
Good job hackers, always keep your tool belt fresh! Scrapers….scrape it together.
Data Science
The last head to head is for the data scientists out there, and you got science in your name and numbers in your veins, you should be at the bleeding edge of tech!
Ouch, yinz need to get with the times.
Python Version Developers Use More Often
This is some hard to gather data as an individual, so I’m going to have to cheat and just base this information off JetBrain’s yearly state of the ecosystem reports from 2017 and 2018.
In 2017, 53% of devs reported using Python 3 as their main language, which went up 22% in 2018 to 75%. Based on those two points of data, we can come to a crystal clear, no doubt conclusion to how many developers will be using Python 3 as their main language in 2019.
That’s right, based on the past two year trend, 97% of developers should be using Python 3 in 2019.
Okay, well, maybe not. But I personal expect that number to be over 90% by the time Python 2 is EOL, which is excellent news.
Operating System Default Language
OSes have a fun time of being in the cross hairs of everyone from desktop to server users, trying to figure out the right combo of what’s best for their users and for their own technology stack going forward. Every major Linux distribution agrees Python 3 is the way to the future and they will need to change over. The hard part is deciding when it will impact the users least and best for their own release cycle. This has caused lots of headaches over the years. So where do we stand now?
| OS | Python Version |
| Windows 10 | None |
| OSX 10.8 | 2.7 |
| Debian 9 | 2.7 |
| RedHat 8* | 3.6 |
| Fedora 29 | 3.7 |
| Ubuntu 19.04* | 3.7 |
(* denotes upcoming releases this year)
Windows has the easy stance of just saying “do it yourself” and Mac is, as usual, not bothering to innovate and just hum along until it breaks. Thankfully most Linux distros, which power the internet, are either already updated or updating this year. I haven’t seen for sure that Debian 10 will be released with Python 3 or that it w ill be out before year’s end, but I would be surprised if either were not true. Then there’s Arch linux. Arch has had Python 3 as the standard for almost as long as it existed, good boy!
Are we ready?
In all honesty, we are. We are far more prepared for this than the financial sector was ready for Y2K, and we all survived that. Moreover, there are always going to be code bases that can’t update to the latest version easily, but that’s true across the entire software development world. That and the fact the Python Software Foundation has given an extended eleven years which has allowed for even the slowest of companies to have ample time to migrate to Python 3.
Python 3 everywhere? Bring it on!
