Refactor: Datasource classes

[rarkins]

I would like to consider class-based datasources to achieve better DRY than today plus some enhance features.

Currently datasources are somewhat declaration-based and the datasource index then handles things, e.g. caching, default registryUrls, etc. This type of stuff should remain although maybe much of the datasource index's logic will move into the base datasource class.

We definitely want error handling shared as much as possible.

Caching would be another aspect done automatically.

Something new: our http layer returns an auth variable today to indicate that the request was auth'd. In future I think we may return an auth fingerprint (e.g. sha256 the authorization header content and then use the first 5 chars). This would then be used in conjunction with caching, to allow private lookup caching.

Need to discuss how to do this. Start small and then enhance it?

[JamieMagee]

My initial attempt at converting cdnjs, clojure, and crate is in #6747. It needs rebasing, but it shows how we can achieve this datasource by datasource instead of having to do a 'big bang' approach.

[rarkins]

@JamieMagee I understand that PR better now. Perhaps worth refreshing and merging after @viceice refactors our modules to statically load instead of dynamic?

[viceice]

Are we really done here?

[JamieMagee]

There's some cleanup to do, but the majority of the work is now done.

[renovate-release]

🎉 This issue has been resolved in version 31.82.2 🎉

The release is available on:

• GitHub release
• 31.82.2

Your semantic-release bot 📦🚀