Features:

• Added synonym it for item. (need to decide on a final name for it!)
• Can omit the first parameter of any operation in a context where an implicit parameter is present. (extract [...], sum, CallToFunc, +10, etc, translates to it extract [...], it sum, CallToFunc(it), it+10)
• Conditions now have access to item/it to reference the instance being checked.
• Can directly access attributes of item/it wherever it exists, no need to write item ->.

Bug fixes (that I know of):

• addressed antlr warnings about ambiguous parsing paths.
• conditions are more robust. E.g., the following now runs fine.

type Foo
  a int (0..1)
  condition C:
    a exists

• nested inline functions are more robust.
• using item in a context where it doesn't exist is not allowed anymore.

and lots and lots of clean-ups.

Two particular challenging technicalities:

• Making the argument of a unary function call optional has as consequence that the parser cannot distinguish between variable references (RosettaCallableCall) and function calls (RosettaCallableWithArgsCall), because there might not be any parentheses. I therefore joint the concept into one: RosettaSymbolReference. This had a lot of impact in every part of the project.
• A question arises what to do with validation errors if the target of the error is an omitted parameter (e.g., omitting it in it sum where it is a list of Foo => should give a type error on it). Xtext by default just red-underlines the container (e.g., in this case sum would be red-underlined instead of it, which is okay for now), but Xsemantics filters out these validation errors. That needed a fix: see RetainXsemanticsIssuesOnGeneratedInputsFilter.