[Matplotlib-users] value "requires_dist" of Matplotlib sets to null

Thomas Caswell tcaswell at gmail.com
Sun Sep 19 17:18:21 EDT 2021


@Semyon: please join the mailing list
https://mail.python.org/mailman/listinfo/matplotlib-users so you can post
un-moderate.

This appears to be a long standing issue with setuptools / pypi [1].
Currently setuptools does not add the required meta-data to the sdist [2]
which if we upload the sdist first sets the singular (and immutable)
metadata from that artifact (which is missing `'requires_dist'` ).  The
subsequently uploaded wheels carry their own meta-data which (I believe) is
what the solver actually uses to ensure that the run-time dependencies
(`'install_requires'` to setuptools) are installed.

I am not clear if this meta-data is used for anything or what the practical
consequences of it being wrong are (other than messing up school
projects).  I was able to find a number of issues on github noting that the
meta-data was missing, but no mention of any consequences.

I would suggest looking at the dependency graph extracted from one of the
down-stream packagers (such as conda-forge, fedora, arch linux, debian,
...).  I expect those graphs to be more correct (within their set of build
artifacts) and to also correctly capture the non-Python dependencies of
Matplotlib.

Tom

[1] https://github.com/pypa/warehouse/issues/6231
[2] https://github.com/pypa/setuptools/issues/1805


On Sun, Sep 19, 2021 at 5:00 PM Семен Есаев <semen.esaev at yandex.ru> wrote:

> Hello,
> When requested to https://pypi.org/pypi/matplotlib/json, the dictionary
> with the key "requires_dist" is set to null. But the page
> https://matplotlib.org/3.1.1/users/installing.html#dependencies has a
> list with dependencies.
>
> I would be grateful if you correct this error or explain how to get
> package dependencies correctly without requests to json, I encountered this
> error because I got an task at the university with visualizing package
> dependencies. And when I query matplotlib, I get an empty graph, although
> there should be dependencies.
>
> Thank you, Semyon Esaev
> _______________________________________________
> Matplotlib-users mailing list
> Matplotlib-users at python.org
> https://mail.python.org/mailman/listinfo/matplotlib-users
>


-- 
Thomas Caswell
tcaswell at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.python.org/pipermail/matplotlib-users/attachments/20210919/cdfbb963/attachment.html>


More information about the Matplotlib-users mailing list