Raspberry Pi-Python: Install Pandas on Python 3.5.

Question

1. I have a Raspberry 3 Model B with the latest version of Raspbian (installed with Noobs 2.3.0).
2. I have successfully installed Python 3.5.2 using the instructions found in this post: http://bohdan-danishevsky.blogspot.com.es/2015/10/building-python-35-on-raspberry-pi-2.html
3. I am trying to install Pandas (particularly, version 0.18 or higher), on that version of Python (not 3.4).

4. I have tried pip install, but can't make it point to python 3.5.2 even uninstalling version 3.4. I have also tried

   sudo apt-get install python3-pandas
   

   with no luck...

I am quite desperate. Is there no way to do this? I have searched for multiple solutions in this site and still nothing.

Does anyone know how to do it?
Thanks

Answer 1

I'd been searching for the same thing: a way to "install" the current version of pandas on a Raspberry Pi 3. Using apt-get to install it on the RPi pulls an outdated version of pandas from the Raspbian repository.

On GitHub. I found two solutions for installing the current version of pandas (and Python/packages in general) on the RPi 3: one solution involves building and installing pandas directly on the Raspberry Pi, and the other solution involves installing a current Raspberry Pi version of conda (called BerryConda) on the Raspberry Pi and then downloading the current version of pandas as an RPi package from the Anaconda Cloud.

Solution 1:

User kleinee on GitHub has created a script that will create a Jupyter notebook server on Raspberry Pi 2 and/or 3. Although, the intent of the script is to create a Jupyter notebook server, it can be modified easily to install only the current version of pandas on the RPi.

The script starts by downloading the most recent version of Python 3 (currently, 3.6.1), setting it up and installing it on the RPi. The script then goes on to download other parts of what he calls a "scientific stack," which includes the current version of pandas. He also provides another script that will check for updates to installed Python packages and automatically update them, as necessary.

You can run the entire script as is (to install current versions of Python, jupyter, pandas, numpy, scipy, matplotlib, etc.) or you can edit the script to install only the current version of pandas. I ran the script "as is" yesterday and it took about 4 hours to install Python and the included "stack" (to include pandas) on my RPi 3.

The advantage to this solution is that you are downloading the current version of Python and packages directly from the source, so you will always get the most recent version. The disadvantage is that you have to "build" them locally on the RPi, which will take time (4 hours, in my case, for about a dozen packages, including Python itself).

Solution 2:

User jjhelmus has created "BerryConda," a current RPi version of conda. He has also created Python packages (including the current version of pandas) that he makes available on the RPi channel at Anaconda Cloud. Using this solution, you can install the current version of pandas using BerryConda without having to compile it, like you do in solution 1.

The advantage to this solution is that you do not have to compile anything locally, so setup is much faster. A possible disadvantage is that BerryConda and the associated packages are built and maintained by someone not associated with Anaconda or Continuum Analytics who may or may not continue to do so in the future. Although, the BerryConda packages are currently up-to-date, there's no guarantee that will be the case going forward.