Minimal debian image with Python 3.6 and geo python tools

Dockerfile

FROM python:3.6-slim-stretch

ADD requirements.txt /tmp/requirements.txt

RUN apt-get update && \
    apt-get install -y \
        build-essential \
        make \
        gcc \
        locales \
        libgdal20 libgdal-dev && \
    python -m pip install numpy cython --no-binary numpy,cython && \
    python -m pip install \
        "rasterio>=1.0a12" fiona shapely \
        --pre --no-binary rasterio,fiona,shapely && \
    python -m pip install -r /tmp/requirements.txt && \
    python -m pip uninstall -y cython && \
    rm -r /root/.cache/pip && \
    apt-get remove -y --purge libgdal-dev make gcc build-essential && \
    apt-get autoremove -y && \
    rm -rf /var/lib/apt/lists/*

RUN dpkg-reconfigure locales && \
    locale-gen C.UTF-8 && \
    /usr/sbin/update-locale LANG=C.UTF-8

ENV LC_ALL C.UTF-8

CMD ["/bin/python"]

Makefile

default: image test notebook

.PHONY: image shell

image:
	docker build --tag slimpy:latest .

test:
	docker run -it --rm slimpy:latest python -c "import rasterio, fiona, shapely, numpy, sys; print('Python:', sys.version); print('Rasterio:', rasterio.__version__); print('Fiona:', fiona.__version__); print('Shapely:', shapely.__version__); print('Numpy', numpy.__version__);"

shell: image
	docker run -it --rm slimpy:latest /bin/bash

notebook:
	docker run -it --rm \
		-p 0.0.0.0:8888:8888 \
		--rm \
		--interactive \
		--tty \
		--volume $(shell pwd)/notebooks/:/notebooks \
		slimpy:latest  /bin/bash -c "cd /notebooks && jupyter notebook --ip=0.0.0.0 --allow-root"

requirements.txt

rasterio>=1.0a12
fiona
shapely
pandas
geopandas
pyproj
mapbox
rasterstats
mapboxgl
jupyter
mercantile
scipy
scikit-learn
seaborn
statsmodels
ggplot

@perrygeo Is there a way to install new dependencies without building the whole docker image? For instance, a new package comes at the end of the list in the requirements.txt file, but nothing else changed. Or this should be run from the Dockerfile, so the build will see that the req file is unchanged and doesn't install the existing layers?
(when working in Jupyter notebook, many times a new packages comes up as need, so the process takes a bit time to be completed)

Hi
Does it work in Heroku ?
regards

Thanks for this.

@Biaggio74 you could install each package as a separate docker RUN command, with packages more likely to need updating towards the end. What I'm going for in this docker image is a single RUN command - optimized for minimal container size.

For applications that need frequent python package updates, you could automate the build step by e.g. having Travis CI build and push the container on each commit. Ideally, if you were using this in production, you would have the versions explicitly pinned in the requirements.txt such that "upgrading" was a matter of pushing a change to this file.

Thank you! This was exactly what I needed. I made a fork that uses docker-compose in place of the Makefile, here: https://gist.github.com/johnniehard/90a7f4fc1b0701360f67ba77b9b50c7a