Consider a more automated and cleaned-up assemble.sh script
@nisse have some opinions on how he would like to put together the docs.sto things:
- Files that need processing are checked in as FILENAME.ctrl (or similar)
- A script (assemble.sh) finds all FILENAME.ctrl files, and processes them appropriately to produce FILENAME. Ideally in a way that no additional configuration is needed in the assemble.sh script.
- FILENAME is never checked-in with git, and added to .gitignore
The advantages that were outlined in !22 (merged):
- The process will not delete the note that a file is copied from a different place; instead only new untracked files that can be listed in .gitignore will be added.
- The script doesn't need to explicitly list all the files to copy.
Might be reasonable, but @rgdd won't refactor right now due to lack of time and the fact that the current approach works well enough for the context of generating the site from a collection-release tarball (which has no notion of git). It is also not clear that .ctrl files is that neat for some of our pre-processing needs, like hugo.toml or components.md which is not simply "hugo metadata + .md document". It is also not necessarily a bad thing to have all files checked in -- ensures the hugo structure is always correct without having a setup that can run the assemble.sh script.
That said, if anyone thinks they can improve this and have the time, consider providing an MR.