Rigorous implementation

[lassik]

The current implementation is ad hoc and is not of based on a well-known pretty-printing algorithm from the literature. This makes it difficult to scale to more difficult jobs, line wrapping in particular.

It seems the usual route nowadays is:

• Use the idea in Wadler's A prettier printer.
• Use the eager implementation of Wadler's idea in Lindig's Strictly pretty.

This seems a good place for us to start.

[lassik]

There are more involved papers, but I expect they'd be too fancy for us and not offer much extra value. E.g.:

• A Pretty But Not Greedy Printer
• Correct-by-Construction Pretty-Printing
• Efficient simple pretty printing combinators

[lassik]

https://github.com/jyp/prettiest/blob/master/paper/blog.md