Miss Dada  🏳️‍⚧️

I think I have a plan to move out of macros for nom. It’s a monstrous amount of work, but it could be backward compatible, ie I’d rewrite the macro internals so existing code still works

so, here's the idea: nom uses macros calling other macros, to generate a function's body. I like it because the generated code is more or less what I'd write manually, and somewhat easy to debug (clean stack traces, etc)

trait based solutions (everything implement a Parse trait with a parse() method and a lot of combinators) suffers from the same debugging issues as futures: it generates an object that will call a serie of Parse::parse() methods, rendering everything hard to follow

I’m currently looking at possible perf regressions in previous rustc versions. I don’t have the same behaviour between stable and nightly. And even stable is a lot slower than it should