This is, to me, an odd way to approach parsing. I get the impression the author is relatively inexperienced with Rust and the PL ideas it builds on.

A few notes:

* The AST would, I believe, be much simpler defined as an algebraic data types. It's not like the sqlite grammar is going to randomly grow new nodes that requires the extensibility their convoluted encoding requires. The encoding they uses looks like what someone familiar with OO, but not algebraic data types, would come up with.

* "Macros work different in most languages. However they are used for mostly the same reasons: code deduplication and less repetition." That could be said for any abstraction mechanism. E.g. functions. The defining features of macros is they run at compile-time.

* The work on parser combinators would be a good place to start to see how to structure parsing in a clean way.