Closed cganley1 closed 5 months ago
I would second what @msanchezst just said, but what really concerns me as that dynamically installing the code undercuts one of the premises of the hub images... that they are stable. Would moving the solo pip install to the end of Dockerfile.custom help? That undercuts the frozen requirements some since they come from pip-tools which would not see this later install, but I think it may still work using both floating and frozen requirements since it would occur after all the env-xxx stuff. It would not be as stable as a pure env-xxx build, but once built the image itself would be stable.
New commit moves installation to dockerfile instead of post-start-hook.
New commit creates jdaviz kernel and updates environment variables.
Included xarray as additional roman-cal package, Added post-start-hook to install jdaviz. Upon further testing, jdaviz worked with test notebooks if it was installed as the last package in the kernel, hence the use of the post-start-hook. This will create issues capturing frozen requirements as dependencies are updated over time.