8 ms·

Good. I've been getting into python a bit because i have an interest in datascience. I'm mostly a Java dev. I have to say the python2/3 divide is a real turn

Good. I've been getting into python a bit because i have an interest in datascience. I'm mostly a Java dev. I have to say the python2/3 divide is a real turn off. Many of the science libs want to use seem to be in 2.7 with no signs of moving.

viraptor 10y ago
At least the science libraries have a better reason for staying. They often require bigger fixes of the modules compiled from other languages. Still, it's good to see larger projects moving on.
jaipilot747 10y ago
Which libraries do you have in mind? Numpy, SciPy, Pandas and packages based off them all support Python 3.
- stevehiehn 10y ago
  It believe it was a digital signal processing library. I don't recall the name.
  - rbanffy 10y ago
    You used the word "many"...
    - stevehiehn 10y ago
      Ya, 'a few' is more accurate. I looked into pybrain which was 2.x. I understand that many people now use tensorflow. The point is it will be nice when more people normalize on 3. It did make me appreciate the backwards compatibility of Java.
joeyspn 10y ago
> Many of the science libs want to use seem to be in 2.7 with no signs of moving The most important scientific libraries have pledged to drop support before 2020, and are all python3-ready http://www.python3statement.org/ http://www.python3statement.org/
- lmm 10y ago
  Is there any hope of other libraries and OSes doing the same thing? I used to work on a GUI app in Python. I ported it to Python 3, then switched OSes for various reasons. 5 years on, on Ubuntu Xenial (so new I can't even use it in Travis, but that's a separate whine), I install pykdeuic4 and it's using Python 2. So I've basically abandoned that project for 5 years now, because every time I looked at it I thought "surely Python 3 will be here in a few months, I don't want to port backwards to Python 2". (Serious question: is there a PPA or anything I can use to get these things for Python 3? I need PyQwt as well as PyKDE)
  - rtpg 10y ago
    I've had similar issues with most GUI libraries. Seriously: if you can, try building your thing with tkinter. It's the standard lib, and will "work"
    - lmm 10y ago
      I'm not going to port to tkinter. I like KDE, I like Qt Designer, and I have a working GUI. If tkinter is the only thing that works in Python I'll abandon the language (I mostly work in Scala these days anyway) and find one with working Qt bindings.
      - 72deluxe 10y ago
        Just use C++ and your Qt bindings will never go away!
        lmm 10y ago
        Qt is a point in C++'s favour, but not enough to outweigh the downsides of the language.
        flukus 10y ago
        Could you write your UI in c++ with custom bindings and have the rest of the app in python? Qt's binding support has been it's biggest issue since forever.
sqrt28 10y ago
I am data scientist who uses both pythons and the only lib I had issues with was the first release of tensorflow (and that was fixed very soon). Otherwise I find it hard to believe that there exist any maintained libraries that do not support both versions.
Analog24 10y ago
I do a lot of data science/ML work and it is almost exclusively done with Python3 so I'm not sure what libraries you are referring to as I haven't had any issues.

viraptor 10y ago

At least the science libraries have a better reason for staying. They often require bigger fixes of the modules compiled from other languages. Still, it's good to see larger projects moving on.

jaipilot747 10y ago

Which libraries do you have in mind? Numpy, SciPy, Pandas and packages based off them all support Python 3.

stevehiehn 10y ago
It believe it was a digital signal processing library. I don't recall the name.
- rbanffy 10y ago
  You used the word "many"...
  - stevehiehn 10y ago
    Ya, 'a few' is more accurate. I looked into pybrain which was 2.x. I understand that many people now use tensorflow. The point is it will be nice when more people normalize on 3. It did make me appreciate the backwards compatibility of Java.

joeyspn 10y ago

> Many of the science libs want to use seem to be in 2.7 with no signs of moving The most important scientific libraries have pledged to drop support before 2020, and are all python3-ready http://www.python3statement.org/ http://www.python3statement.org/

lmm 10y ago
Is there any hope of other libraries and OSes doing the same thing? I used to work on a GUI app in Python. I ported it to Python 3, then switched OSes for various reasons. 5 years on, on Ubuntu Xenial (so new I can't even use it in Travis, but that's a separate whine), I install pykdeuic4 and it's using Python 2. So I've basically abandoned that project for 5 years now, because every time I looked at it I thought "surely Python 3 will be here in a few months, I don't want to port backwards to Python 2". (Serious question: is there a PPA or anything I can use to get these things for Python 3? I need PyQwt as well as PyKDE)
- rtpg 10y ago
  I've had similar issues with most GUI libraries. Seriously: if you can, try building your thing with tkinter. It's the standard lib, and will "work"
  - lmm 10y ago
    I'm not going to port to tkinter. I like KDE, I like Qt Designer, and I have a working GUI. If tkinter is the only thing that works in Python I'll abandon the language (I mostly work in Scala these days anyway) and find one with working Qt bindings.
    - 72deluxe 10y ago
      Just use C++ and your Qt bindings will never go away!
      - lmm 10y ago
        Qt is a point in C++'s favour, but not enough to outweigh the downsides of the language.
        flukus 10y ago
        Could you write your UI in c++ with custom bindings and have the rest of the app in python? Qt's binding support has been it's biggest issue since forever.

sqrt28 10y ago

I am data scientist who uses both pythons and the only lib I had issues with was the first release of tensorflow (and that was fixed very soon). Otherwise I find it hard to believe that there exist any maintained libraries that do not support both versions.

Analog24 10y ago

I do a lot of data science/ML work and it is almost exclusively done with Python3 so I'm not sure what libraries you are referring to as I haven't had any issues.