Python: stand-alone pandas library

[ZCodeNoob]

Is it possible for me to get stand alone version pandas to a user designated directory? 

The Linux machines I work on do not have internet access so I can't use pip. I just wanted to copy the libraries so I can use it. 

[Jeff Reback]

https://store.continuum.io/cshop/anaconda/

free distribution 

On Jun 12, 2015, at 4:34 PM, ZCodeNoob <jiey...@gmail.com> wrote:

Is it possible for me to get stand alone version pandas to a user designated directory? 

The Linux machines I work on do not have internet access so I can't use pip. I just wanted to copy the libraries so I can use it. 

--
You received this message because you are subscribed to the Google Groups "PyData" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pydata+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ZCodeNoob]

Thank you Jeff. I installed anaconda, which is great. But I still want to manually control what libraries I use from anaconda, and what I use from "official linux-installed version," and what I use from anaconda. 

Based on anaconda's instruction, if I set PATH=~/anaconda/, I would use all versions from anaconda. 

[Adam Hughes]

Anaconda will install libraries into a virtual environment anyway, so you can't really go back and forth between libraries installed on your OS's Python distribution and libraries installed through conda.  The easiest thing to do is to create a new virtual environment in anaconda and install Pandas and all its dependencies fresh.  Is this not suitable for your needs?

[John E]

I may not totally understand what you're doing but I'm also stuck behind a firewall and haven't figured out how to use a proxy server (or if IT allows it either), but you can grab all of the individual packages here:  http://repo.continuum.io/pkgs/index.html

I forget the syntax (it's not that hard to figure out though), but you can tell conda to then install any of those specific packages after you download them -- instead of having conda do it automagically over the web.  You just have to be careful to get the exact one that you need (right version for your OS, right python version, right numpy version, etc.).  You will also need to manually update dependencies as necessary (whereas conda over the web would do that automagically).

Anyway, it's absolutely possible to do what you are trying to do, although figuring out how may not be obvious.

[Ramesh Sampath]

You can always download miniconda that's just the conda package manager with Python and you can choose to install what you want.

I typically, download Miniconda (http://conda.pydata.org/miniconda.html), then create an environment and inside that - conda install numpy matplotlib pandas.  I also do conda install pip to install packages that are not yet in conda's main repo.

But I would suggest that you don't mix linux installed packages and conda installed packages.  I suggest that you leave linux level python alone and install and run all packages inside miniconda or anaconda or in an virtualenv environment.  Since Pandas only depends on Numpy, python-dateutil and pytz, you can certainly pip install pandas if you have gcc, gfortran and other dependencies.  Unfortunately, pre-built wheels doesn't exist in pypi for linux distributions.  If this is the first time deploying this in Linux, I would agree that you should use anaconda or miniconda at the minimum.

Thanks,

Ramesh

-