python 2 or python 3?

[ajschumacher]

It may be time to go to 3. What do you think, @justmarkham? @sinanuozdemir?

[justmarkham]

No strong feelings. Here are my questions:

• Do we know if all of the Python modules we might want to use for the class work in 3?
• If someone is already using 2, and is required (for work or school) to keep working in 2, is it a pain to have both installed on the same machine?
• What are the benefits of moving to 3? (And I mean "real" benefits to students, not benefits like "upgrading is good because it helps to move the Python community to 3".)

[ajschumacher]

Good questions! I believe:

• What comes with Anaconda is here: http://docs.continuum.io/anaconda/pkg-docs.html - ugh; it looks like statsmodels isn't in their 3.4... Hmm.
• It isn't terribly hard with conda/virtualenv, I think. Of course it's still One More Thing.
• The benefit would be not having to later worry about moving to 3. People should seriously get on 3 one of these days. And of course not having to deal with Python 2 silliness and/or from __future__ import reasonable_behavior.

What do you think, @arahuja? @podopie? @datadave? What version(s) are you using in classes? In your work? (I'm still all 2.7, but I'm feeling like I should think about moving eventually, hence this discussion.)

[datadave]

I'm all for lazily sticking with 2.7. Python 2 is supported to 2020 http://www.i-programmer.info/news/216-python/7179-python-27-to-be-maintained-until-2020.html and apparently less than 2% of downloaded packages (still) on PYPI are Python 3, so for library compatibility's sake plus compatibility of other learning materials out there: I'm a 2.7 man.

Also, I love using "from future" because time travel.

I haven't looked into the major benefits from Python 3 though -- if you guys tell me Python 3 is worth it, then I'd be into checking out its new tricks and helping out the cause http://jakevdp.github.io/blog/2013/01/03/will-scientists-ever-move-to-python-3/.

On Sun, Jun 15, 2014 at 10:42 PM, Aaron Schumacher <notifications@github.com

wrote:

Good questions! I believe:

• What comes with Anaconda is here: http://docs.continuum.io/anaconda/pkg-docs.html - ugh; it looks like statsmodels isn't in their 3.4... Hmm.
• It isn't terribly hard with conda/virtualenv, I think. Of course it's still One More Thing.
• The benefit would be not having to later worry about moving to 3. People should seriously get on 3 one of these days. And of course not having to deal with Python 2 silliness and/or from future import reasonable_behavior.

What do you think, @arahuja https://github.com/arahuja? @podopie https://github.com/podopie? @datadave https://github.com/datadave? What version(s) are you using in classes? In your work? (I'm still all 2.7, but I'm feeling like I should think about moving eventually, hence this discussion.)

— Reply to this email directly or view it on GitHub https://github.com/ajschumacher/gadsdc/issues/67#issuecomment-46136914.

[ajschumacher]

Hmm that's helpful. I'm not overly moved by the 2% figure because most of pypi is garbage and the percentage of packages that matter is pretty high. The main advantage of 3 might be that you don't have to worry so much about text encodings because it does the right thing with unicode/utf-8. Not always an issue though. Maybe just for ease of everything we should stay with 2.7 One More Time?

[podopie]

+1 what @datadave said. I've had no strong urge to move towards 3, especially at work we still use 2.7 (app engine doesn't support 3 yet). half my code would be 3 and the rest 2.7... conflicting! :)

In Anaconda's defense the 3.3 packaging looks fantastic, as it includes roughly everything we have used in class.

Also in 2.7's defense, aside from the unicode thing, it's pretty solid.

[ajschumacher]

Yup sounding more and more like 2.7 is the place to (still) be... Probably easier lots of ways. Just out of curiosity, do you all use virtualenv in class (or conda equivalent)?

[justmarkham]

So if we are set on 2.7, we should say it explicitly on this page: https://github.com/ajschumacher/gadsdc/blob/master/00-preliminaries/tools.md

Are we decided?

[ajschumacher]

I think so!

[sinanuozdemir]

Agreed

Sinan Ozdemir http://about.me/sinanozdemir Professor of Mathematics / Computer Science Founder & CTO of Tier5 http://tier5.co M.A. in Theoretical Mathematics, Johns Hopkins University Mobile: +(609) 462-6706

On Thu, Jun 19, 2014 at 1:16 AM, Aaron Schumacher notifications@github.com wrote:

I think so!

— Reply to this email directly or view it on GitHub https://github.com/ajschumacher/gadsdc/issues/67#issuecomment-46524452.