For more involved answers, how would you score code length for something like a binary calculator?
The length is the size of the save file
Unless there is another mechanism that can be used to reliably and consistently import a "program", the save file is the obvious choice. Yes, that means your score will probably be much higher than other languages. No, there's nothing wrong with that - competition within languages, remember? As commented, you can golf and/or generate the save files manually, which can save some bytes.
Plus, Java won't be the most verbose language anymore.