Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Python 2 vs Python 3 for Computational Tasks

  1. Feb 15, 2013 #1
    I've decided to learn Python (having a background in C++ and Matlab), and I'm a little unsure about which version to get. I've read in a few places that Python 2.x is better because packages like numpy and scipy don't work with Python 3.x, but this seems to be out of date since as far as I can tell, these packages do support Python 3.x now. I plan to do mostly do computations (e.g. modeling/simulation) and then visualizations of those computations. Have most of the important packages for this type of thing been ported over to Python 3.x or am I safer going with Python 2.x?
  2. jcsd
  3. Feb 15, 2013 #2


    User Avatar
    Gold Member

    This may help you with your decision. In my lab we use Python 2 since it's more supported. Python 3 is a cleaner, better language, but Python 2 still has a lot of inertia.

    My advice would be to go with Python 3 if you can, keeping in mind much of the available Python code is still targeted at Python 2.

    If you do use Python 2, it's good practice to write code that is forward-compatible with Python 3. A simple example is the print functionality, which is a statement in Python 2, used as:
    Code (Text):

    [color=#008000][b]print[/b][/color] foo
    but a function in Python 3:
    Code (Text):

    The Python 3 functionality can be acheived by calling
    Code (Text):

    [color=#008000][b]from[/b][/color] [color=#0000FF][b]__future__[/b][/color] [color=#008000][b]import[/b][/color] print_function
    There are other changes between the versions which you can find online.
  4. Feb 16, 2013 #3
    That's funny...when I started to read the advice and thought the opposite was going to be recommended. Since it wasn't...I will recommend the opposite.

    Kind of learn Python 2 and 3 in a encompassing way so that you write Python 2 in a 3-compatible manner...that way, you know for sure that no module is going to stop you from doing your work; THEN, when all modules have been ported and 3 is as complete as 2 is today...your code should be easy to port...I thought at some point they were writing a program that would convert your Python 2 code to Python 3...I bet it works good when Python 2 has already been written with Python 3 in mind.
  5. Feb 16, 2013 #4


    User Avatar
    Science Advisor
    Homework Helper

    This is the sort of issue that keeps "not for profit" software as a minority interest, unfortunately.

    If the inventors of Python 2 have invented a "better" but incompatible language, good for them. But calling it "Python 3" and assuming that the rest of the Python community would instantly drop whatever else they were working on and port all their existing code was somewhere between "dumb" and "naive" IMO.

    As the old proverb says, sometimes "the best is the enemy of the good".
  6. Feb 16, 2013 #5


    User Avatar
    Gold Member

    There's 2to3, or something similar, but I don't know how well it works.
  7. Feb 16, 2013 #6
    Hhhhmmm....I don't know, AlephZero, I think I could argue that you may have it backwards.

    It is precisely because Python is open source and "not for profit" that thegreenlaser actually has a choice! Nobody is forcing s/he to pick one or the other or to "move forward" onto a specific version. If it was proprietary software, licensed, etc...s/he could very well have no choice but to start using the present version of a software regardless of base, popularity, bug-iness, etc. and simply because the previous one is no longer available.

    Anyway, that's just one of the things software freedom is about.
  8. Feb 16, 2013 #7


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    Er, that proverb usually refers to the situation where people reject using a good solution in favor searching for a mythical best solution -- not the situation where people are using something good and then something better comes along.
  9. Feb 16, 2013 #8
    By no means!!!!!.. Python 2 is the most common version for all uses from commercial software to scientific purposes and Python 3 is a future version and no one knows when it will be widely used, and most professional Python programmers use Python 2 these days.

    In addition, More versions of Python 2 are still released, what tells that Python 2 will not be out of date soon!!!!
  10. Feb 16, 2013 #9


    User Avatar
    Gold Member

    Python 2.7 will be the last major release in the 2.x series, which is going into what the PSF is calling an "extended maintenance period."

  11. Feb 17, 2013 #10
    Thanks everyone. I guess there's no real clear consensus on this issue, so I've decided to go with Python 2 just because of bigger library and better supporting documentation. As a casual programmer doing stuff for myself I'm guessing it won't make all that much of a difference in the end.
  12. Feb 17, 2013 #11


    User Avatar

    Staff: Mentor

    I would just check what are the major differences to keep them in mind while programming, so that the code would be easier to port in future (in case it will be necessary).
  13. May 7, 2013 #12
    Sorry for digging up an old thread, but I wanted to ask the same question, only I have no experience with programming at all (besides a little bit of Matlab), and I want to start by learning Python. I can't really tell from this thread which version is best, lol.

    On their website they say

    Since I'm just starting out, I suppose the 3rd-party stuff is not important, but from what I gather, version 2 is still the most commonly used, so which do I get?


    Edit: saying "if you don't know, try Python 3.3" sounds a little sneaky to me...
    I did read this page, and it's all Greek to me, no help at all.
  14. May 7, 2013 #13


    User Avatar

    Staff: Mentor

    Sounds quite logical to me, and I believe it to be the correct approach (even if my experience with Python is very limited).
  15. May 7, 2013 #14
    I was referring to the language they chose to use.
    "Here, try this..."
    "What is it?"
    "Just try it..."
    I would feel more assured if they said "If you don't know which version to use, download 3.3.1" or something like that, but that's being silly, I know.
  16. May 7, 2013 #15
    I second that.

    For my business, the most tragic victim of such a "drop" would be our usage of omniORBpy for CORBA prototyping and development scripting.

    No way.
    None at all.
  17. May 13, 2013 #16
    I have still found this thread of very little help. Accidentally I found this, from this page (under "Warnings for Beginners"):

    Unfortunately, when I read that, I had already downloaded Python 3 and started learning it with another book written for Python 2.

    I just wanted to add this for other beginners who can find no guidance on this topic.
  18. May 13, 2013 #17


    User Avatar

    Staff: Mentor

    Have you run into any problems because you use 3.3?
  19. May 15, 2013 #18
    The author of the book I'm using says it can be used by both Python 3 and 2 users, but some of the things in the book do not work in Python 3, so it seems he doesn't even use Python 3. It's quite a pain trawling the internet trying to see how things should be done in Python 3, otherwise you have to figure it out on your own, which is terrible for a complete beginner.

    Most of the help for beginners is actually written for other programmers, i.e. it's not actually comprehensible to real beginners. And most of the stuff on the web seems to be for Python 2 as well.

    That's the end of my rant, thank you, lol.
  20. May 15, 2013 #19
    Somewhat offtopic
    I think one of the odds "not for profit software" often suffers from is missing the right point for a platform freeze.

    Some features of higher versions may be handy, but more often than not, the alleged need for a platform feature is in fact a design problem.

    The most compelling case of recent days is

    I really would fancy an alternative to adobe development environment, esp because, being a Linux guy, I don't know know yet if I should mourn or applaud the enforced "transition" from official support to pepperflash-only.

    But compiling Lightspark would force so much need for updates and replacements on us, that I would hesitate to give a go for that - I simply cannot rule out side-effects when I replace core components of our current "Squeeze" platform.

    That's quite a pity, and esp because I somehow doubt that all feats of Lightspark's platform are really needed for it.

    Best regards, S.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook