The code generator is creating properties for things like the ref attribute on format annotations, or the test attribute on asserts. These are not properties and the rules are different for them. E.g., for format properties, most of the time code will simply insist they have a value, want to know what it is, and it's an error if they are not defined. This uniform behavior is built into the getProperty that is called when you access one of these property members.

But for the non-property attributes, code will typically need to implement conditional behaviors. E.g., whether or not there is a ref attribute different things will happen.

So, since the XML Schema for format annotations doesn't distinguish these two kinds of attributes, the code-generator needs a special-case filter so that it either doesn't generate any code for these attributes, or generates different code for them (which will return an Option type so one can have the code behave differently.)