I added classpath scanning so that shadow-cljs doesn't rely on a central index to find closure JS files. There is a goog/deps.js file in the closure-library jar, listing all the files. But it is a .js file and I didn't want to parse that. This index is needed because there is no direct mapping of name -> file in closure. For CLJS we are strict about what the filename is for a given namespace, closure is not.

I did scan all the classpath, so that libraries in theory could also have those files. Questionable if anyone ever did that.

But if people are strict enough about the names regular closure JS files can still be found without that index.

I'll either remove classpath indexing entirely and use goog/deps.js instead. Or just scan the closure-library jars. Maybe let people opt in to parse specific classpath entries if needed?

Indexing the entire classpath actually can take quite a long time, so I added caching for this. Removing the indexing will also allow removal of the cache making things simpler overall.