Hi folks,

I was helping at a Python workshop workshop yesterday and in general I see
learners launch the Anaconda Jupyter notebook from the Anaconda GUI instead
of the terminal.

One such user on Mac OSX had launched the notebook from the Anaconda GUI
but had import errors.  However, if I had him install the packages from
inside the notebook using "! pip install pandas", the installation
succeeded but the import errors continued.

This was a little surprising, because that workaround succeeded in the
past.  I'm assuming that the shell bang pip command inside the cell invokes
pip associated with the running python's PATH, but maybe that's a false
assumption and there is a more sensible to install packages for the running
notebook python version.

It's also possible that the Anaconda GUI had selected a virtual environment
and that pip installed the package globally instead of into the environment.

Has anyone run into something similar and what is the correct way you would
go about installing a package from a running notebook?  Unfortunately I'm
mostly familiar with pip / pipenv instead of conda and don't often use
notebooks in my work.

I know that I could have also had the user not use the Anaconda GUI and run
the jupyter notebook from the terminal, but in the past I've also
encountered situations where users have multiple Anaconda and Miniconda
installations where one has to mangle the ~/.bash_profile and was therefore
hesitant to go down that road and opted to instead work within the context
of using the Anaconda GUI that the user is already familiar with.

Pariksheet

------------------------------------------
The Carpentries: discuss
Permalink: 
https://carpentries.topicbox.com/groups/discuss/T72d9652a490be7d3-Me32abccb4262ecc1eb389d93
Delivery options: https://carpentries.topicbox.com/groups/discuss/subscription

Reply via email to