> We currently delegate all semantic operations to Metalua; on the Java side,
IntelliJ requires you to build an AST that contains every token in the file so
that you can reproduce the file exactly if given the AST built from it.
I'd be interested to see how I could incorporate Metalua into the process.
Metalua details are best discussed in the dedicated mailing list, but indeed, its ASTs contain enough information to recompose the original file, although not always trivially (the focus is put on ease of tree manipulations, not on easy and perfect faithfulness to the source string).
Moreover, we're currently changing some details of the Metalua parser and lexer, to suit LDT needs better. Most of those changes are intended to make the AST/source mapping tighter and easier to navigate.
