Closed nick-youngblut closed 3 years ago
Hey @nick-youngblut thank you for reaching out.
That problem is not related to this package directly. Indeed, the conda environment are made available as kernel thanks to nb_conda_kernels. Then the default JupyterLab launcher is building all those cards.
There have been discussion and some propositions to improve that Launcher. But so far none of it made it to the code (see #3795 or #5953).
One trick enabled by nb_conda_kernels
is to modify the default name. It could help you: see https://github.com/jupyterlab/jupyterlab/issues/3795#issuecomment-455669749
I haven't yet switched from Jupyter to Jupyter Lab, mainly due to this issue. I assumed that a simple fix would be implemented by now (eg., alternative formatting as a list of kernels instead of a tiled array), but I guess the user still has to hack the naming as is proposed by @fcollonval (thanks for the suggestion! but it's not a straight forward way of solving what should be a minor UI issue).
Even with truncating the names (via jupyter lab --CondaKernelSpecManager.name_format="'{1}'"
; see https://github.com/jupyterlab/jupyterlab/issues/3795#issuecomment-463743040), I still can tell some kernels apart from one another (see pic). I'm surprised that no one has simply provided an option for a simple list instead of tiles (buttons). btw, I'm using Jupyter Lab 2.2.8
There is now an alternative launcher with list view to address this issue: https://github.com/fcollonval/jlab-enhanced-launcher
Moreover the name_format
configuration of nb_conda_kernels
has been extended to support more customization (starting from v2.3.1):
https://github.com/Anaconda-Platform/nb_conda_kernels#configuration
Description
For conda envs installed in custom locations, the path can be too long to fully display in the launcher (see the attached pic). In my case, many of the conda env "tiles" in the launcher look exactly the same. It would be nice if the user could change the format to just a list of paths instead of the default "tile" view.
Context