Dockerfile

FROM jupyter/base-notebook:2022-06-06 AS builder

# Copy the repository into the container
COPY --chown=${NB_UID} . /opt/afwizard

# Install Conda environment
RUN conda env update -n base --file /opt/afwizard/environment.yml

# Build and install the project
RUN conda run -n base python -m pip install --no-deps /opt/afwizard

# Thoroughly clean up the conda environment to save space
RUN conda clean --force-pkgs-dirs -a -q -y

# Second stage of multistage build
FROM jupyter/base-notebook:2022-06-06

# Make JupyterLab the default for this application
ENV JUPYTER_ENABLE_LAB=yes

# I don't see why we need a work subdirectory in home
RUN rm -rf ${HOME}/work

# Copy the artifacts from stage 0 that we need
COPY --from=builder /opt/conda /opt/conda

# Ensure that the conda provided PROJ database is found. This
# is somewhat not possible in the base environment. There is
# this discussion which allowed me to solve the issue although I
# think it is a horrible situation and definitely a bug:
# https://github.com/conda-forge/geopandas-feedstock/issues/63
ENV PROJ_LIB="/opt/conda/share/proj"

# Ensure that the Jupyter notebooks are located in the home directory
RUN copy_afwizard_notebooks ${HOME}/