Python 3 pledge

[mpharrigan]

Wanted to bring omnia's attention to the following "pledge"

http://python3statement.github.io/

I think it was brought forward by the IPython people. Specifically, they want scientific python packages to pledge to stop supporting python 2 (for free) after 2020 (python 2.7's eol)

1. I think this is a reasonable timeframe, and we might want to consider adding omnia to the list of pledged organizations. It would provide nice exposure
2. I've started discussions for msmbuilder and mdtraj to sign the pledge. I know several packages depend on mdtraj, so this isn't a decision to be made lightly. Please join the discussion there if you have opinions
3. I don't think anything in omnia depends on msmbuilder. I'll likely try to convince people to drop support before 2020 for msmbuilder. Please join the discussion there if you have opinions

[jchodera]

I'm actually +1 on the visibility bump from joining this.

Following the Python guidelines and terminating py2 support by 2020 sounds like a no-brainer. Before that---are there critical tools that are stopping support early that would make it difficult to go all the way to 2020?

[mpharrigan]

Before that---are there critical tools that are stopping support early that would make it difficult to go all the way to 2020?

IPython / jupyter is dropping support for python 2 soonish. I don't know how that effects us. Other than that, we could probably handle this on a package-by-package basis

[jchodera]

IPython / jupyter is dropping support for python 2 soonish.

Soon-ish looks like "mid-2019"---6 months early---according to the diagram.

[jchodera]

Other than that, we could probably handle this on a package-by-package basis

Could we just pick a target date omnia-wide and see if we can stick to it? Since there are interdependencies among packages, it might cause a lot more trouble to handle this on a package-by-package basis.

[mpharrigan]

IPython / jupyter is dropping support for python 2 soonish.

Soon-ish looks like "mid-2019"---6 months early---according to the diagram.

This is the last release that supports python 2. In < 6 months, the current version of ipython won't support python 3. So yes, they will continue to support an old version basically until 2020. At the same time, they're being pretty aggressive with their timeline for new releases

[jchodera]

In < 6 months, the current version of ipython won't support python 3.

Won't support python 3?

[jchodera]

Ah, I see:

How about this plan:

• Make sure we are py3-compatible by 2017.
• Drop py2 support in mid-2019 when IPython 5.x LTS py2 support ends.

[mpharrigan]

That sounds good to me. Curious to hear what the rest of omnia thinks

Make sure we are py3-compatible by 2017

Do we have an idea of how close to this we are? I've been operating under the sense that nearly everything is compatible with py3. We should gently prod omnia package maintainers to make this a priority.

[jchodera]

We should gently prod omnia package maintainers to make this a priority.

I think we need to start this process ASAP. Is there a way we can have our build framework complain when packages don't provide py3 versions, or somehow check this automatically and send out reminders periodically?

[mpharrigan]

Based on using github search for "skip" and examining the skip entries, these packages currently don't get built for py3

• assaytools
• schema
• subprocess32 (this is a backport for py2, so it doesn't count)
• pymol
• deepchem
• graphviz
• lomap
• ensembler
• pytram
• yank

[mpharrigan]

Is there a way we can have our build framework complain when packages don't provide py3 versions

We could open a PR that tries to build the omnia metapackage for python 3. Any dependencies it can't resolve will show up

[mpharrigan]

This issue is relevant #103

[jchodera]

Based on using github search for "skip" and examining the skip entries, these packages currently don't get built for py3

Thanks for this list! I think the end of 2016 is a realistic goal for most of this list. (No idea about pymol; ensembler might be EOLd and replaced with a new codebase for ensembler2.)

I think we want to both push for adding extra data to each recipe about package maintainers and then find some way to bug them periodically about the impending py3 support deadline (31 Dec 2016).

[jchodera]

We could open a PR that tries to build the omnia metapackage for python 3. Any dependencies it can't resolve will show up

Great idea!

[Lnaden]

As a reminder to all Omnia developers: The Dec 31, 2016 deadline for supporting Python 3 is approaching. How do we want to handle packages which are not up to python 3 compatibility yet? Near as I can tell there has not been any warning/info provided to developers about this since the middle of the year, and pushing for them to have py3 compatibility by year's end may not be reasonable at this point.

[peastman]

Which projects still don't support Python 3?

I went through the ones Matt listed earlier. Of those, schema, pymol, and yank no longer list [py3k] in their skip sections. Deepchem still lists it, but that can be removed since it now supports Python 3. That leaves assaytools, graphviz, lomap, ensembler, and pytram. What's the status of Python 3 support for those?

[marscher]

pytram is superceded by pyemma and its recipe can be removed in principle.

[franknoe]

I'd leave the repo, but we don't need it on omnia. Am 06/12/16 um 20:13 schrieb Martin K. Scherer:

…

pytram is superceded by pyemma and its recipe can be removed in principle. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#551 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGMeQnE7JEbE38YRwTY70FQaLqsZdct7ks5rFbPFgaJpZM4JIb38>.

--

---------------------------------------------- Prof. Dr. Frank Noe Head of Computational Molecular Biology group Freie Universitaet Berlin Phone: (+49) (0)30 838 75354 Web: research.franknoe.de Mail: Arnimallee 6, 14195 Berlin, Germany ----------------------------------------------

[jchodera]

From our end:

• we never got the pymol recipe fully working, so we can deprecate this recipe, especially since another one is available for osx and linux
• schema belongs to @Lnaden and @andrrizzi
• I'm supposed to clean up ensembler over the holidays. It's a mess right now.
• assaytools is already Python 3 compliant, so we just need to remove that skip section from the recipe. @sonyahanson: Can you do that? We should probably also cut an updated release of assaytools while we're at it.

[Lnaden]

Schema has been on Python 3 for a while now since we updated to 0.6.2.