After actually working with it a bit, I still think #rust might be one of the more reasonable popular languages out there, but the semantics of the `?` operator are stupid. It looks like the result of an oath to include it in the language, whatever the cost, between two drunken developers – one from Go one from Haskell.
@robryk I don't particularly like that it treats two types very specially at the level of the language – it would be better if it at least referred to a trait, which one could implement on any type. But I could understand if it was just that, maybe the convenience is worth making the language slightly uglier. What actually convinced me `?` is a mistake is the difference between `f()` and `Ok(f()?)`. As far as I know `?` is the only inherent feature of Rust that performs automatic coersion. Introducing this to the language cannot be worth it.
It would be completely fine as a macro – either `try!` as it used to be or a macro that rewrites `?`s in a function. The latter would preferably have an option for turning automatic coersion on/off, but even if it didn't I wouldn't care that much, as it wouldn't be a inherent feature of the language.
@timorl It's the most useful error propagation tool I've so far had my hands on; it is explicit and opt-in yet unobtrusive. What's there not to like?
@chrysn Kind of answered in the other toot, but in short:
1. Two types are treated very specially at the level of the language.
2. Afaik this is the only language feature that uses automatic coersion.
@timorl These two are odd, but https://github.com/rust-lang/rfcs/blob/master/text/3058-try-trait-v2.md seems to improve the previous MVP into something more aligned. There will still be some auto-into-ing left, but with a trait backing it it'll be smoother.
@chrysn Oh that is *massively* better than the current situation, I didn't know about this RFC. Even the auto-coersion doesn't bother me that much, since now it's a strange artifact of the `Result` implementation rather than an inherent feature of the language. Thanks!
@timorl Which part of its semantics and what do you usually want to do instead? (And what would you prefer something-like-? to do instead?)