0

I am just about to install Numpy from source to include OpenBLAS. So I cloned the repo from Git with git clone https://github.com/numpy/numpy.git and then thought I'd see which branches are available.

On the Github page, there are only maintenance branches along with master. However, once in Terminal and in the cloned repo, if I use the command git checkout v then press TAB, I see there are many more - specifically there are the ones that correspond to the version I have installed – around 1.1.xx – and then there are some much higher, around 1.9.xx:

user@user:~/src/numpy$ git checkout v1.
Display all 109 possibilities? (y or n)

user@user:~/src/numpy$ git checkout v1.1
v1.1.0          v1.10.2rc1      v1.11.0b1       v1.11.2         v1.12.0rc1 
v1.10.0         v1.10.2rc2      v1.11.0b2       v1.11.2rc1      v1.12.0rc2 
v1.10.0b1       v1.10.3         v1.11.0b3       v1.11.3         v1.12.1 
v1.10.0.post2   v1.10.4         v1.11.0rc1      v1.1.1rc1       v1.12.1rc1 
v1.10.0rc2      v1.1.0rc1       v1.11.0rc2      v1.1.1rc2       
v1.10.1         v1.1.1          v1.11.1         v1.12.0         
v1.10.2         v1.11.0         v1.11.1rc1      v1.12.0b1       

user@user:~/src/numpy$ git checkout v1.9.
v1.9.0      v1.9.0b2    v1.9.1      v1.9.2      v1.9.3      
v1.9.0b1    v1.9.0rc1   v1.9.1rc1   v1.9.2rc1
  • What is going on here? Which do I install?
  • I notice that the higher versions seem to correlate to SciPy in some way. Are they the same things with different versions or must I be careful with dependencies? I found some information about their relationship, but it didn't help me in this instance.
Community
  • 1
  • 1
n1k31t4
  • 2,745
  • 2
  • 24
  • 38
  • 2
    "the version I have installed – around 1.1.xx" - your version is almost a decade out of date. The most recent version is 1.12.1. I would say to get the most recent version, but if you're on a version that out of date, you might run into compatibility issues with other things you're using when you try to upgrade. – user2357112 Apr 14 '17 at 21:29
  • For the latest scipy, the *minimum* version of numpy that is supported is 1.8.2. But if you are installing from scratch, you might as well install the latest released version of numpy, 1.12.1. (You may already know this, but just in case: the `rc#` and `b#` suffixes on the numpy versions indicate *release candidates* and *beta releases*, respectively. Don't install any of those.) – Warren Weckesser Apr 14 '17 at 21:36
  • 1
    You might also be aware of this, but some folks aren't: version 1.12.0 is newer than version 1.9.0. Don't think of the periods between the major, minor and point release numbers as decimal points. The periods are just separators of integer values. – Warren Weckesser Apr 14 '17 at 21:40
  • Thanks for your help - all is now clear! I have installed `v1.12.1` – n1k31t4 Apr 14 '17 at 22:48

0 Answers0