Simplify image vastly, get rid of R

Dockerfile

@@ -1,61 +1,6 @@
-FROM buildpack-deps:focal-scm
+FROM quay.io/jupyter/scipy-notebook:2023-12-08
 
-ENV LC_ALL en_US.UTF-8
-ENV LANG en_US.UTF-8
-ENV LANGUAGE en_US.UTF-8
-ENV DEBIAN_FRONTEND=noninteractive
-ENV NB_USER jovyan
-ENV NB_UID 1000
-ENV SHELL /bin/bash
-
-ENV CONDA_DIR /opt/conda
-ENV R_LIBS_USER /opt/r
-
-# Explicitly add littler to PATH
-# See https://github.com/conda-forge/r-littler-feedstock/issues/6
-ENV PATH ${CONDA_DIR}/lib/R/library/littler/bin:${CONDA_DIR}/bin:$PATH
-
-RUN adduser --disabled-password --gecos "Default Jupyter user" ${NB_USER}
-
-# Create user owned R libs dir
-# This lets users temporarily install packages
-RUN mkdir -p ${R_LIBS_USER} && chown ${NB_USER}:${NB_USER} ${R_LIBS_USER}
-
-# Install these without 'recommended' packages to keep image smaller.
-# Useful utils that folks sort of take for granted
-RUN apt-get update -qq --yes && \
-    apt-get install --yes --no-install-recommends -qq \
-        htop \
-        less \
-        man \
-        nano \
-        screen \
-        tar \
-        tmux \
-        wget \
-        vim \
-        locales > /dev/null
-
-RUN echo "${LC_ALL} UTF-8" > /etc/locale.gen && \
-    locale-gen
-
-# Install R packages
-# Our pre-built R packages from rspm are built against system libs in focal
-# rstan takes forever to compile from source, and needs libnodejs
-# We don't want R 4.1 yet - the graphics protocol version it has is incompatible
-# with the version of RStudio we use. So we pin R to 4.0.5
-# littler, r-cran-mgcv, r-cran-survival, r-cran-matrix are specific packages needed
-# for apt to actually install the correct version of R 4.0.5
-ENV R_VERSION=4.0.5-1.2004.0
-RUN apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
-RUN echo "deb https://cloud.r-project.org/bin/linux/ubuntu focal-cran40/" > /etc/apt/sources.list.d/cran.list
-RUN apt-get update -qq --yes > /dev/null && \
-    apt-get install --yes --no-install-recommends \
-    r-base-core=${R_VERSION} \
-    r-base-dev=${R_VERSION} \
-    r-cran-littler=0.3.11-1.2004.0 \
-    nodejs \
-    npm
+USER root
 
 # Install desktop packages
 RUN apt-get update -qq --yes > /dev/null && \
@@ -72,135 +17,16 @@ RUN apt-get update -qq --yes > /dev/null && \
         openjdk-8-jre \
         libreoffice > /dev/null
 
-# for nbconvert & notebook-to-pdf
-# See https://github.com/utoronto-2i2c/jupyterhub-deploy/issues/35 as well
-RUN apt-get update -qq --yes && \
-    apt-get install --yes -qq \
-        pandoc \
-        texlive-xetex \
-        texlive-fonts-recommended \
-        texlive-science \
-        dvipng \
-        ghostscript \
-        cm-super \
-        libx11-xcb1 \
-        libxtst6 \
-        libxrandr2 \
-        libasound2 \
-        libpangocairo-1.0-0 \
-        libatk1.0-0 \
-        libatk-bridge2.0-0 \
-        libgtk-3-0 \
-        libnss3 \
-        libxss1 \
-        > /dev/null
-
-# for kableExtra & partitions & prettydoc & RSelenium & scholar & showtext & stopwords & copula
-# See https://github.com/utoronto-2i2c/jupyterhub-deploy/issues/25
-RUN apt-get update -qq --yes && \
-    apt-get install --yes -qq \
-        libcurl4-openssl-dev \
-        libxml2-dev \
-        imagemagick \
-        libmagick++-dev \
-        libgmp3-dev \
-        libglpk-dev \
-        libfreetype6-dev \
-        libgit2-dev \
-        libgsl0-dev \
-        > /dev/null
-
-WORKDIR /home/jovyan
-
-COPY install-mambaforge.bash /tmp/install-mambaforge.bash
-RUN /tmp/install-mambaforge.bash
-
-# Needed by RStudio
-RUN apt-get update -qq --yes && \
-    apt-get install --yes --no-install-recommends -qq \
-        psmisc \
-        sudo \
-        libapparmor1 \
-        lsb-release \
-        libclang-dev  > /dev/null
-
-# 1.3.959 is latest version that works with jupyter-rsession-proxy
-# See https://github.com/jupyterhub/jupyter-rsession-proxy/issues/93#issuecomment-725874693
-ENV RSTUDIO_URL https://download2.rstudio.org/server/bionic/amd64/rstudio-server-1.3.959-amd64.deb
-RUN curl --silent --location --fail ${RSTUDIO_URL} > /tmp/rstudio.deb && \
-    dpkg -i /tmp/rstudio.deb && \
-    rm /tmp/rstudio.deb
-
-ENV SHINY_SERVER_URL https://download3.rstudio.org/ubuntu-14.04/x86_64/shiny-server-1.5.15.953-amd64.deb
-RUN curl --silent --location --fail ${SHINY_SERVER_URL} > /tmp/shiny-server.deb && \
-    dpkg -i /tmp/shiny-server.deb && \
-    rm /tmp/shiny-server.deb
-
-# Needed by many R libraries
-# Picked up from https://github.com/rocker-org/rocker/blob/9dc3e458d4e92a8f41ccd75687cd7e316e657cc0/r-rspm/focal/Dockerfile
-# libglpk40 for igraph
-# libzmq3-dev for IRKernel
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends \
-        libgdal26 \
-        libgdal-dev \
-        gdal-bin \
-        libgeos-3.8.0 \
-        libproj15 \
-        libudunits2-0 \
-        libxml2 \
-        libzmq3-dev \
-        libglpk40 > /dev/null
-
-# Needed by staplr R library
-RUN apt-get update && \
-    apt-get install -y --no-install-recommends \
-    default-jdk > /dev/null && \
-    R CMD javareconf
-
-
-# R_LIBS_USER is set by default in /etc/R/Renviron, which RStudio loads.
-# We uncomment the default, and set what we wanna - so it picks up
-# the packages we install. Without this, RStudio doesn't see the packages
-# that R does.
-# Stolen from https://github.com/jupyterhub/repo2docker/blob/6a07a48b2df48168685bb0f993d2a12bd86e23bf/repo2docker/buildpacks/r.py
-# To try fight https://community.rstudio.com/t/timedatectl-had-status-1/72060/5 ,
-# which shows up sometimes when trying to install packages that want the TZ
-# timedatectl expects systemd running, which isn't true in our containers
-ENV TZ America/Toronto
-RUN sed -i -e '/^R_LIBS_USER=/s/^/#/' /etc/R/Renviron && \
-    echo "R_LIBS_USER=${R_LIBS_USER}" >> /etc/R/Renviron && \
-    echo "TZ=${TZ}" >> /etc/R/Renviron
-
 USER ${NB_USER}
 
-# Set CRAN mirror to rspm before we install anything
-COPY Rprofile.site /usr/lib/R/etc/Rprofile.site
-# RStudio needs its own config
-COPY rsession.conf /etc/rstudio/rsession.conf
-
-
-# Install R packages, cleanup temp package download location
-COPY install.R /tmp/install.R
-RUN /tmp/install.R && \
-	rm -rf /tmp/downloaded_packages/ /tmp/*.rds
-
-# Install bioconductor separately, as it seems to fail often
-COPY install-bioconductor.R /tmp/install-bioconductor.R
-RUN /tmp/install-bioconductor.R && \
-	rm -rf /tmp/downloaded_packages/ /tmp/*.rds
-
 COPY environment.yml /tmp/
 
 RUN mamba env update -p ${CONDA_DIR} -f /tmp/environment.yml && mamba clean -afy
+
 # For https://2i2c.freshdesk.com/a/tickets/187
 # If we don't set `NLTK_DATA`, the data gets downloaded onto $HOME, which
 # isn't available when the image is loaded onto JupyterHub.
 # So we download alongside our packages.
 # Note that textblob.download_corpora just calls nltk to download corpora
 ENV NLTK_DATA ${CONDA_DIR}/nltk_data
-RUN mkdir -p ${NLTK_DATA} && python -m textblob.download_corpora
-
-# Install IRKernel
-RUN r -e "install.packages('IRkernel', version='1.1.1')" && \
-    r -e "IRkernel::installspec(prefix='${CONDA_DIR}')"
+RUN mkdir -p ${NLTK_DATA} && python -m textblob.download_corpora

environment.yml

@@ -5,47 +5,30 @@ dependencies:
   # for jupyter-desktop-server
   - websockify
 
-  # language runtimes
-  - python=3.10.*
-
-  - numpy==1.24.*
-  - scipy==1.10.*
-  - matplotlib==3.6.*
-  - pandas==1.5.*
-  - seaborn==0.12.*
-  - sympy==1.11.*
-  - networkx==2.8.*
-  - scikit-learn==1.2.*
-  - beautifulsoup4==4.11.*
-  - statsmodels==0.13.*
+  - networkx==3.2.*
   - nltk==3.8.*
-  - spacy==3.4.*
-  - geopandas==0.12.*
-  - geopy==2.3.0
+  - spacy==3.7.*
+  - geopandas==0.14.*
+  - geopy==2.4.*
   # From https://github.com/utoronto-2i2c/jupyterhub-deploy/issues/29
   - astroquery==0.4.*
-  - astropy==5.2
-  - galpy==1.8.*
+  - astropy==5.3.*
+  - galpy==1.9.*
   - libgfortran==3.0.0
-  - tensorflow==2.10.0
-
+
   # Infrastructure things
-  - jupyterlab==3.5.* # Pinned to <3.6, until https://github.com/2i2c-org/infrastructure/issues/2244 is fixed
-  # Pinned until https://github.com/Jupyter-contrib/jupyter_nbextensions_configurator/pull/152 can land
-  # See https://github.com/2i2c-org/infrastructure/issues/2380
-  - nbclassic<0.5
-  - retrolab==0.3.*
-  - voila==0.3.*
-  - nbgitpuller==1.1.*
-  - jupyter-resource-usage==0.5.1
-  - jupytext==1.13.*
-  - RISE==5.7.1
-  - jupyter_contrib_nbextensions==0.7.0
-  - notebook==6.5.*
-  - jupyterhub==3.1.0
-  - jupyter-server-proxy==3.2.*
+  - nbgitpuller==1.2.*
+  - jupyter-resource-usage==0.7.*
+  - jupytext==1.15.*
+  - jupyter-server-proxy==4.*
 
-  - ipywidgets==8.0.4
+  # From https://2i2c.freshdesk.com/a/tickets/497
+  - contextily==1.4.*
+  - libpysal==4.9.*
+  - esda==2.5.*
+  - splot==1.1.*
+  - xlrd==2.0.*
+  - numba==0.57.*
 
   # python-graphviz installed on request by Scott Schwartz in
   # https://2i2c.freshdesk.com/a/tickets/697?note=80283596987
@@ -55,14 +38,15 @@ dependencies:
 
   - pip:
     # Infrastructure things that don't exist in conda-forge
-    - jupyter-desktop-server==0.1.3
+    - jupyter-remote-desktop-proxy==1.*
     - git-credential-helpers==0.2
-    - jupyter-rsession-proxy==1.4
-    - jupyter-shiny-proxy==1.1
     - jupyter-tree-download==1.0.1
 
-    - datascience==0.17.5
-    - otter-grader==4.3.4
+    # Infrastructure thing that doesn't support notebook 7
+    - jupyter_contrib_nbextensions==0.7.0
+
+    - datascience==0.17.*
+    - otter-grader==5.*
 
     # nbextensions
     - jupyterthemes==0.20.0
@@ -72,12 +56,12 @@ dependencies:
     - jax[cpu]
 
     # From https://2i2c.freshdesk.com/a/tickets/187
-    - textblob==0.17.1
+    - textblob==0.17.*
+
+    # Install this from pip because installing from conda conflicts with geopy / geopandas
+    # Re-evaluate once https://github.com/conda-forge/tensorflow-feedstock/pull/353 lands
+    - tensorflow==2.15.*
 
-    # From https://2i2c.freshdesk.com/a/tickets/497
-    - contextily==1.3.*
-    - libpysal==4.7.*
-    - esda==2.4.*
-    - splot==1.1.*
-    - xlrd==2.0.*
-    - numba==0.56.*
+    # Install from pip because it does not work from conda-forge atm
+    # https://github.com/conda-forge/rise-feedstock/issues/43 has information
+    - RISE==5.7.*