jupyter-docker-hello-world
Activate npstripout in your local repository to make sure that you do not commit huge notebooks to the repository:
pip install --upgrade nbstripout
nbstripout --install
nbstripout --install --attributes .gitattributes
Build the docker image:
pushd build-docker-base-image && make && popd
Run the docker image:
./jnw.sh # jupyter notebook wrapper
And connect with your browser to: http://localhost:8888
For using PyCharm with your docker container see Docker: Using Docker as a Remote Interpreter
Test it locally outside of docker via a python/conda environment:
conda env create -n jupyter-docker-hello-world -f build-docker-base-image/environment.yml
# conda env update -n jupyter-docker-hello-world -f build-docker-base-image/environment.yml
# conda remove --name jupyter-docker-hello-world --all
source activate jupyter-docker-hello-world
# install the other missing packages as described in the Dockerfile
source deactivate
conda info --envs # or: conda env list
conda list -n jupyter-docker-hello-world
conda env export > environment_.yml
If you want to work with this new environment in a jupyter notebook it is not as easy as I would have wished. You need to install a new kernel in jupyter. More details can be found here and here, but in short it works like this:
source activate jupyter-docker-hello-world
pip install ipykernel
ipython kernel install --user --name=jupyter-docker-hello-world
jupyter notebook # start the notebook and change the kernel for a given notebook to the newly visible 'jupyter-docker-hello-world'
Pay attention to .condarc for the channels it contains:
$ cat ~/.condarc
channels:
- pyviz
- conda-forge
- defaults
Check on the command line if a module is installed:
python -c "import watermark"
See notes-on-anaconda for some pitfalls, especially around not being able to specify channels on a per dependency level.