Dear matplotlib users,
I've encountered an issue with 3d plotting that has me completely stumped.
Since a recent update that added ipympl, any plot that uses projection='3d'
is completely incorrect, even after subsequently removing the package. For
example, running the sample parametric line plot code from the mplot3d
tutorial located at
produces the following image
Each of the plots looks vaguely like the points being used to generate the
plot are sent to random locations, and clearly the axis is upside down --
possibly an error when computing the projection? 2D plotting works as
expected, as do other aspects of the package. Any assistance or suggestions
(including other places to ask) would be helpful.
Details: I have a group of users working in a JupyterHub (0.9.6) environment
on a local Ubuntu 17.10 server. The hub uses SystemUserSpawner to create
local Docker containers running JupyterLab (0.35.6) environments for
individual users. The containers run Ubuntu 17.10, Python 3.6.7 and
matplotlib 3.0.3. Plotting was working fine until I recently updated the
Docker images to include ipympl to allow the users to make interactive 3d
Once the issue above became apparent, I rolled back the dockerfile to the
previous, working version (rebuilt using --nocache), but the problem
persists. I have built a bare-bones version of the container which includes
only the OS packages, python, the current conda versions of notebook,
jupyterhub, jupyterlab, and matplotlib, and the problem persists. This
minimal Dockerfile is available here:
Because it persists across container rebuilds, my best guess is that adding
ipympl created or modified a file in the users' home directories, as these
are the only directories visible to the containers.
I'm happy to provide more detailed information, but don't want to dump
things that won't be useful.
Thanks for your attention,