Thanks for the heads up! If I understand correctly, is the idea of this PR (#194) that expressions will model more than just annotations? It looks like they're being ramped up to represent most python expressions (and represent annotations with more granularity). Does that sound right?

Yes, we will now parse attribute values, decorators and class bases as expressions as well (previously it was only for type annotations). Also, previously expressions were lists of strings and names, now they're actual instances of classes with proper attributes. For example an annotation or value which is a subscript will have a left attribute as well as a slice attribute. It makes it easier to access the slice part, while before we had to access elements using their index, which was extremely prone to index errors.