When running pip within docker, I've found out from many excellent SO answers how to cache downloading of packages, i.e. for example
# syntax=docker/dockerfile:experimental
...
RUN --mount=type=cache,mode=0755,id=pip,target=/root/.cache/pip \
pip3 install -r requirements.txt --target /root/.cache/pip
This caches the downloads appearing in requirements.txt, but not the installation itself. The problem is that if I update requirements.txt with a new package - assuming no shared dependencies change - one only saves on download time (which in this case is only about half the docker build time).
Now I could just add new packages one-by-one in the dockerfile, but then I have to circle back round for the inevitable squash into a single RUN line - the time hit is unavoidable if one wants to simplify the dockerfile and minimize the image size.
So, is there a way to speed up the build even further by caching more than just the package downloads?
Thanks as ever!
REFS:
Using a pip cache directory in docker builds
How to avoid reinstalling packages when building Docker image for Python projects?
