Reduce memory usage significantly when pulling large channels all at once by
- Not loading entry script until it is being processed
- Deleting the entry memory once fully processed
- Disable yt-dlp from keeping entry info in memory since ytdl-sub reads it from file
A complete gutting of the internals of ytdl-sub to support functions in our variable syntax, in addition to being able to access a yt-dlp entry's .info.json fields using functions. Functionally, ytdl-sub should still look and behave the same from a user-perspective.
With so many lines of code changed (+8927, -2708), no doubt there will be new issues. Please make a GH issue or reach out on Discord if your config/subscriptions break in any way/shape/form.
Details on how to use function support will come soon in the form of proper documentation in our readthedocs.
For less popular sites, ytdl-sub's custom download archive was not working. This was because we weren't using the right extraction key (though it still worked for YouTube, Bandcamp, Soundcloud, and others). This bugfix should now exactly match yt-dlp's and never be a problem again.
Huge thanks to @Svagtlys for root-causing this 🤝
* actually working
* changed plugin format. Need unit tests, documentation, update e2e to filter
* refactor plugin name, matching now working
* description work in progress, need more intricate testing
* tests looks good, regenerate output
* e2e tested, need more failure tests, no dupe variable names, add docs
* unit tests passing again
* more fail tests
* updated with name, default, better docs
* sanitized vars created
