this layer is "big" (e.g. 19Mo in freecodecamp, 260Mo in gutenberg) in most scrapers
this layer takes time to build (many web resources to fetch at least)
Rebuilding this layer everytime code change looks like a waste of time (especially for the developer, but also in terms of computing resources) and storage (we re-upload the whole layer to ghcr.io everytime the branch is modified in a pull request and intermediate image versions might be kept on the developer machine).
In current Dockerfile, program and its dependencies are installed in the same layer: https://github.com/openzim/_python-bootstrap/blob/c02d44f76f0bdb85f78998ec0ffc13fb38b8a5a0/Dockerfile#L17-L19
This layer is rebuild everytime code, config or documentation are changed : https://github.com/openzim/_python-bootstrap/blob/c02d44f76f0bdb85f78998ec0ffc13fb38b8a5a0/Dockerfile#L13-L15
However:
freecodecamp, 260Mo ingutenberg) in most scrapersRebuilding this layer everytime code change looks like a waste of time (especially for the developer, but also in terms of computing resources) and storage (we re-upload the whole layer to ghcr.io everytime the branch is modified in a pull request and intermediate image versions might be kept on the developer machine).