It's complicated: "they are copies, but". The buts:
If you use zero-install, then yes those files are duplicated as each repository will have it. For this reason zero-install is better suited at monorepos than projects with dozens of repositories
If you don't use zero-install, we still cache the archives into a global "mirror" before cloning them using the native clone operation (when supported, mostly OSX). For this reason you only pay the size cost once when relevant.
If you don't use zero-install and don't use OSX, you can enable the global cache mode which will cause Yarn to use the global mirror as datastore (in which case you only pay the size cost once no matter what).
Note that all this is about the 2.x; the 1.x had worse characteristics.
It's complicated: "they are copies, but". The buts:
If you use zero-install, then yes those files are duplicated as each repository will have it. For this reason zero-install is better suited at monorepos than projects with dozens of repositories
If you don't use zero-install, we still cache the archives into a global "mirror" before cloning them using the native clone operation (when supported, mostly OSX). For this reason you only pay the size cost once when relevant.
If you don't use zero-install and don't use OSX, you can enable the global cache mode which will cause Yarn to use the global mirror as datastore (in which case you only pay the size cost once no matter what).
Note that all this is about the 2.x; the 1.x had worse characteristics.
Interesting so I just stuff all repos into one then 😂 just kidding
Yea thanks for the summary I'll keep that in mind 🧠 Unfortunately I'm on macOS 😅
Are you still considering the hard link approach or is it too hard 🤔😅 haha sorry 😂