PYTHONPATH being unset

[Torres__Efrain]

Hi There!

Installation process for this Beta release is simple and clean, contrary to the pain that used to be the manual Numpy and Scipy recompilation using Intel's MKL libs, thanks for this!  

Now I am trying to test Intel Python Beta release in my own python stuff,  which I usually include in the PYTHONPATH . However this env var  is unset  when the `pythonvars.sh` is sourced:

if [ ! -z ${PYTHONPATH+x} ]; then
    echo "PYTHONPATH being unset. Old value: ${PYTHONPATH}"
    unset PYTHONPATH 
fi

if [ -z ${PATH+x} ]; then
    PATH="/opt/intel/intelpython27/bin"; export PATH
else
    PATH="/opt/intel/intelpython27/bin:${PATH}"; export PATH
fi

So, apparently python preinstalled packages are now included in PATH and not in PYTHONPATH, then, how can I include my python experimental packages?

Is there a clean way to re-set the PYTHONPATH such that the intel-python look on this dirs (and, in this way avoid all necessary changes in my .zshrc/.bashrc) ?   

Cheers

Efrain

[Todd_T_Intel]

Hello Efrain. Thank you for trying our Beta!

The recommendation when working with conda environments is to keep PYTHONHOME and PYTHONPATH empty, using the "activate" and "deactivate" scripts to switch environments by modifying only the PATH.

Perhaps the cleanest way to include your test packages with our distribution is to create an environment using conda env and install your modules into that environment (using any installation means you require).

If you wish to use PYTHONPATH to include additional modules, You should be fine to set PYTHONHOME to the Intel(R) Distribution for Python installation root, PYTHONPATH to any additional library paths, and prepend the "bin" directory to the PATH.

You can copy/modify pythonvars.sh to create your own environment setup script that works to your liking.

Best,
Todd