I know #2156 is in flight for loading package.json in a standard way. I can update this to use that when it lands. Should just mean I replace my current getPackageContents implementation.

My plan was that once this is reviewed and in place we can either put it into the permissions tables and refresh canThis or pass it to the proxy to manage. If the proxy handles this, I'll refactor proxy.js into a class that takes the app name and permissions and adds checks before the api calls, each app would get their own instance of the proxy then.

@jgable , I'm about half a day away from updating #2156 as per Hannah's last comment. Just need a warning message as well. Expect this to be final in about 4 hours hopefully.

I'm gonna be honest. I don't want to use require-tree to parse the package.json. Just looking at it and I can't directly call parsePackageJson I'd have to call readAll. Also, that file could use some refactoring. The extend method should be using _.extend, and the way messages are shared globally across all the functions is kinda hacky. If we don't want many different ways of parsing a package.json, then we need to export a function that just does that, not the whole directory.

If you want I can put together another PR to clean up that file and then use it here.

@jgable Picked that up literally 5 minutes before you mentioned it :-)

@jgable is there something missing from this PR due to require-tree? The extend method was replaced, but readAll is still used.

I want to refactor that require-tree before I use it so I don't have to call readAll, just haven't had time yet. I'm hoping to get to it today.

Updated

Refactored to use parsePackageJson from require-tree. Still a little weird to pass messages around instead of just have it reject the deferred.

This is now on 003-data-updates

Still a little weird to pass messages around instead of just have it reject the deferred.

Why are we doing this? Is it to prevent require-tree from requiring errorHandling.js?

I think it's just a legacy implementation thing. I guess it's because more than one error could happen and didn't want to reject with an array of errors.

This has been remerged onto the apps-perms branch which is basically master with all the data changes on it.