As in Python download stats for March 2019, I recently did some analysis of the python.org download logs. I used better tooling this time, so I have some more interesting numbers, and I’m also happy to run more queries if people have ideas (though all I have access to is essentially the URL).
I filtered out a few obvious scrapers, but it made little difference. They showed up with a high absolute number of hits, but evenly distributed across every single file on the server. The badly behaved downloaders from last year seem to be gone.
Category | Downloads | Users | Downloads/user |
---|---|---|---|
Windows | 15,999,056 | 3,941,325 | 4.06 |
Source | 5,297,204 | 832,034 | 6.37 |
macOS | 1,022,698 | 309,514 | 3.3 |
Sig | 770,146 | 58,952 | 13.06 |
Docs | 430,758 | 63,257 | 6.81 |
Other | 18,984 | 1208 | 15.72 |
RPM | 16,076 | 573 | 28.06 |
Definitions:
-
Windows any of the
.exe
,.msi
or.zip
packages, not counting sub-parts of an install (so only the 2.7 MSIs, essentially) -
macOS any
.dmg
or.pkg
file - Source any of the archives, except those containing docs
-
Docs those containing docs
-
Sig any
.gpg
,.asc
or.md5
files -
RPM some old
.rpm
files that are still floating around - Other anything else
Operating Systems
As with last year, this is heavily biased towards Windows, as python.org is the primary source for most users. Though this year, the Microsoft Store is an alternative - according to my dashboard, there were 226,272 downloads in the same time period. So quite small (1.4%) compared to python.org.
Versions
This chart is all downloads relating to a particular version (based on the directory name).
Version | Downloads | Users |
---|---|---|
3.9 | 193,347 | 32,902 |
3.8 | 10,173,285 | 2,945,632 |
3.7 | 6,449,794 | 1,611,958 |
3.6 | 2,207,577 | 475,685 |
2.7 | 2,121,831 | 469,376 |
Hope this is informative or interesting to people. I’m happy to take questions or requests for other pivots.