[Question] Generating Multiple Projects - Best Practices?

The work with projectReferences is great, especially with the addition to embedding sub projects in #570! With the project I work on, we currently split our app project from our frameworks project so as a result I have two separate project.yml files that each are used to generate the project for development.

Each time I checkout/pull, the regeneration flow is like this:

$ mint run yonaskolb/XcodeGen xcodegen --spec Frameworks/project.yml --project Frameworks/
$ mint run yonaskolb/XcodeGen xcodegen --spec App/project.yml --project App/
$ bundle exec pod install --clean-install

While this works, one thing that gets to me a little is the fact that I can't leverage the --use-cache argument along with postGenCommand to avoid regenerating the pods project (and workspace) in the event that there is no change.

I've seen a lot of recent tickets/discussions about people using project references so I'm wondering if there is a way that we could improve this kind of workflow to take advantage of caching capabilities?

I've seen the old discussion about generating workspaces in #168 however it doesn't look like there has been any movement on that, and maybe its a bit more complex than I thought in my comment at the end of that thread.

Another thing I was thinking was to potentially extend the projectReferences option even further with something like this:

projectReferences:
  Frameworks:
    path: Frameworks/Frameworks.xcodeproj
    spec: Frameworks/project.yml

In the event that a spec is declared, XcodeGen could use this as an indicator to generate reference projects also all in the same command so this would allow me to call just the following:

$ mint run yonaskolb/XcodeGen xcodegen --spec App/project.yml --project App/ --use-cache

And as a result, we could leverage the caching through all referenced specs and trigger the parent postGenCommand if anything changes with either the target project or any of its references.

Maybe this is a bit much, or maybe there are other plans on the horizon however I'm keen to get involved in working on a solution for this kind of thing so I'd appreciate any feedback or ideas/pointers as to how I (and others in the same situation) can achieve such behavior. Thanks!

Edit: Maybe another option could be to expose either a cache-check/validate command to detect if we need to run the generator (or somehow achieve similar from reading stdout in a more machine-readable way?) would be a simpler solution that allows people to chain commands and check the cache in their own scripts?

Start with the existing Frameworks/project.yml and App/project.yml workflow, then review the projectReferences, --use-cache, and postGenCommand entry points described in the issue. Define whether referenced specs should be generated in one command and how cache changes should trigger post-generation work; completion should support the stated single-command workflow or clearly document an alternative.