[Biopython] When to drop Python 3.6 support?

[Peter Cock]

Dear Biopythoneers,

We had various delays due in part to key people's availability, but
six months after the last email in this thread, we have now released
Biopython 1.79.

Biopython 1.79 is planned to be our final release to support Python
3.6, so if there are any last minute pleas to keep supporting it,
please reply within the next week.

Otherwise we will drop testing Python 3.6 on the main repository and
continuous integration, and raise an error trying to install under
Python 3.6.

Thank you,

Peter

On Tue, Dec 8, 2020 at 10:27 AM Peter Cock <p.j.a.cock at googlemail.com> wrote:
>
> Thank you for your comments Stéphane,
>
> Ubuntu LTS releases (and similar) are the only use case I could think
> of where people might still want to use Python 3.6, and I
> believe/assume they are a minority now. I may be wrong. Relatively few
> people have replied to this discussion, and you are the first doing
> so. Are you actually using the system provided Python 3.6 here?
>
> NumPy is our only required dependency, and they have now dropped
> support for Python 3.6. We don't currently require an especially new
> version of NumPy, so this in itself would not stop us supporting
> Python 3.6, but it would stop us using new functionality in numpy as
> desired.
>
> You asked about conda, the institute where I work, and many local
> universities are using conda on their clusters nowadays, and likely
> servers too. The big advantage here is giving users easy access to
> install tools on their own (using bioconda and conda-forge as well as
> the default channel for packages of relevance), while the underlying
> cluster OS remains with whatever older tools and libraries it comes
> with (including an old version of Python which would not be used).
> This separation and duplication also avoids version conflicts (e.g.
> between two different users of the cluster, or indeed using conda
> environments different tools anyone might wish to have installed at
> the same time). In contrast, while DebianMed for Debian/Ubuntu (and
> similar groups for other Linux distributions) do an excellent job of
> keeping bioinformatics packages up to date, this is only available to
> local systems administrators. i.e. Fine for a self managed Linux
> desktop or server (which is how I used Debian/Ubuntu for years), but
> not a multi-user system.
>
> Docker too has a role, especially in deployment. Unless you are trying
> to combine other tools requiring an older base image / older Python,
> then I don't see Biopython dropping Python 3.6 support being a
> barrier. But not being a Docker user, I may be missing something?
>
> Would anyone else like to speak in favour of keeping Python 3.6 support longer?
>
> Peter
>
> On Tue, Dec 8, 2020 at 9:54 AM Téletchéa Stéphane
> <stephane.teletchea at univ-nantes.fr> wrote:
> >
> > Le 10/11/2020 à 10:50, Peter Cock a écrit :
> > > I note that Python 3.6 has now been dropped from NumPy 1.20. It would be
> > > good to start preparing to drop Python 3.6 support in Biopython. According
> > > tohttps://www.python.org/downloads/  and also PEP494, Python will end
> > > support for Python 3.6 in December 2021, about a year away.
> > >
> > > Are there any major platforms (e.g. popular Linux distributions) where the
> > > default is Python 3.6 where our users might be inconvenienced? Personally
> > > I switched to using conda (with conda-forge and bioconda) a while back, and
> > > their default Python is quite up to date - so I don't usually worry about the
> > > system provided Python anymore.
> > Hi all,
> >
> > Ubuntu for one, in its 18.04 LTS (10 years, up to April 2028) is still
> > using python 3.6(.9)
> > as its official python version, and will probably default to this one
> > for "a moment":
> > https://wiki.ubuntu.com/Python/Python36Transition
> >
> > Out of numpy, is there any requirements in biopython "forcing" this
> > migration?
> >
> > As I see it for now, Python3 is still evolving rapidly, so getting "the
> > latest one"
> > may not seem a good idea if you want to break the community's work :-/
> >
> > And for conda, except if you are a Windows user, I don't get the point
> > of using
> > a whole new system in a directory (taking more space than the original
> > one...)
> > since when you want to combine multiple large collection software
> > (django + apache + rdkit + biopython in my case) conda is certainly not
> > the easiest
> > to use (better use docker in this situation).
> >
> > My 0.02 cents/pennies :-)
> >
> > Stéphane
> >
> >
> > --
> > Team Protein Design In Silico
> > UFIP, UMR 6286 CNRS, UFR Sciences et Techniques,
> > 2, rue de la Houssinière, Bât. 25, Nantes cedex 03, France
> > Tél : +33 251 125 636 - Fax : +33 251 125 632
> > http://www.ufip.univ-nantes.fr/ - http://www.steletch.org
> >
> > _______________________________________________
> > Biopython mailing list  -  Biopython at mailman.open-bio.org
> > https://mailman.open-bio.org/mailman/listinfo/biopython