Explorations in treating pretty-printing as a search problem
Pretty-printers have a bunch of heuristics around whitespace and width. These are usually burned into the library, perhaps parameterised by ribbon width or a maximum document length.
What happens when we instead encode all possible choices in our printer, then apply heuristics to guide the search?
Current answer: ┐(‘～` )┌
May eventually circle back into bidirectional parsing/printing.