Recently I asked a Python library maintainer if he could upload a new tagged version of his library to PyPI (https://pypi.org/). He stated that PyPI is not so important nowadays [1].
Is is true? Do people (i.e., you) indeed tend not to use PyPI? If so, what do you use instead and why?
The requirements.txt
apparently allows you to specify the git repository directly, skipping PyPI by using git+https://github.com/<repo>
directly. Though I can't think of any particular advantage over using PyPI.
Are there any graphs to support the statement? Should the statement be true I would expect the amount of downloads from PyPI (per interval) would be decreasing in time. PyPI publishes this kind of data to Google BigQuery. Checked now and the number of downloads are as follows:
202009 1565855136 -- the query was run on 11th of September 2020
202008 5155068175
202007 5409386519
202006 5211181171
202005 5108756961
202004 4812648839
202003 4670947975
202002 4067963794
202001 4155726766
201912 3867376444
201911 3845881964
201910 3922992929
201909 3492788322
201908 3374679723
201907 3338326277
201906 2998812162
201905 3028973146
201904 2436032402
201903 2732697164
Obtained using this query:
SELECT
SUBSTR(_TABLE_SUFFIX, 1, 6) AS `month`,
COUNT(1) AS num_downloads
FROM `the-psf.pypi.downloads*`
WHERE
_TABLE_SUFFIX BETWEEN FORMAT_DATE(
'%Y%m01', DATE_SUB(CURRENT_DATE(), INTERVAL 18 MONTH))
AND FORMAT_DATE('%Y%m%d', CURRENT_DATE())
GROUP BY `month`
ORDER BY `month` DESC
I am aware that this question might not have a correct answer and thus might not be following all the standards of questions asked at StackOverflow. However I can't think of better place to ask on this.
PyPI is not so important nowadays Is is true?
Not at all. The advantages of publishing on PyPI:
One can publish wheels at PyPI so pip install package
downloads a platform-specific wheel.
PyPI is hosted at a CDN (content delivery network) so downloading from PyPI is fast.
3a. pip
caches downloaded packages so it doesn't re-download packages for every virtual environment. It caches by their original URLs and the URLs for PyPI are stable.
3b. pip
very badly caches cloned repositories. Usually it re-clones a-new the entire repository.