Nom is a fantastic library. I've never implemented a parser before and I found it relatively easy to pick up. Some of that is due to parser combinators in general, no doubt, but it was very satisfying to be able to compose a parser with nom.
This is much easier than trying to design the entire parser on a whiteboard/paper and implement it in one big go, I think.
I've written 4(?) different parsers with different major versions of nom, and each has been easier than the previous one.
The main thing I'd like to see improved related to it would be `cookie-factory`, its basically symetric serialization library.