A surprising enum size optimization in the Rust compiler

Yeah, there are all sorts of nifty little tricks about compressing the enums, Appel dedicates a huge chunk of section "4.1. Data representation" on it in his "Compiling with Continuations" book about the internals of the SML/NJ compiler for the Standard ML, and ultimately concludes that the returns are quickly getting really diminishing with those, and since datatypes can actually be abstract in ML (in pretty much the same way they can be in Rust), the applicability of those tricks is generally restricted to unexported, intramodular datatypes.

I like how Appel's book is still being referenced despite being first published in 1992!