Experiment with lossless parser

[maxcountryman]

In order to eventually support comments, we'll need to use a parsing method that is lossless. Such an effort is already underway, but will require reworking formation.