A decade ago, I wrote a TADS authoring tool. Part of that was a map drawer. I have recently had the desire to create an auto mapper.
In my opinion, parsing game output is not the way to do it.
It’s easy to parse"go north", but there are limitless ways to move location. If you are lucky, the interpreter will always output the name of your location first. And the first time that it doesn’t, your automapper is fritzed.
Similarly, items can pop in and out of existence and inventory in limitless ways.
Also, both locations, objects and NPCs can be renamed mid game.
My idea is to make minimal mods to the interpreter code. If it is modular, there will be core routines to create, destroy, move and rename (CRUD, in database terms). Mod those to broadcast info to a socket and let plugins decide what to do with it, including automapping.
What do you think? I would really like to discuss this with someone.
I first thought of parsing transcripts, as it wouldn’t matter which interpreter is used, but, as I said, there are too many problems.
A core update to the heart of the popular interpreters would allow plugins to be interpreter agnostic.