Hacker News Reader

There are many ways to avoid the boilerplate you mention. The deriving ppx extension is one of them. I personaly used my own poor's-man-deriving called ocamltarzan which I used in pfff to factorize boilerplate by using auto-generated visitors. See https://github.com/facebook/pfff/blob/master/lang_php/parsin...

View on HN

Yes, it looks much better than an ad hoc ML. As I said, tricks like Scrap Your Boilerplate are very useful, although they're still not as dense as Nanopass.

If a straightforward, depth-first visitor is relatively easy to infer, more complex strategies (e.g., the one you'd need to resolve the lexical scoping) are still a bit messy. And if your visitor is building a new AST out of an old one, things are getting much more tricky.

Reply on news.ycombinator.com