My understanding is Rust compiles each crate in the same way C++ compiles each .cpp file, so if you stick everything in a single crate you get horrible compile times.
This does seem like a poor design decision by Rust to me, forcing people to break things into arbitrary crates just to get reasonable compile times.
It also seems like a disaster for incremental compilation.
> This does seem like a poor design decision by Rust to me, forcing people to break things into arbitrary crates just to get reasonable compile times.
Not necessarily, since codegen units can introduce intra-crate parallelism during compilation.
But in any case, as with basically everything there are tradeoffs involved in choosing compilation unit size. Making your compilation unit crate-sized also means that you have some more flexibility in your code organization (e.g., stuff can mutually depend on each other, which is itself potentially useful and/or harmful) and you don't run into other potential issues like the orphan rule. There are also potential impacts on optimization, though LTO muddy the picture. Codegen units are just the cherry on top.
There's almost certainly other things I'm forgetting and/or not knowledgeable about as well.
This does seem like a poor design decision by Rust to me, forcing people to break things into arbitrary crates just to get reasonable compile times.
It also seems like a disaster for incremental compilation.