Don't you think C/C++ are hard to learn too? C++ has a vast surface area and I think it's unlikely that most people who use it understand much of it. When I code in C, I feel like I have to be paranoid because it's so easy to forget a path that doesn't recover allocated resources. But if I'm not paranoid, I just leak the resources and don't care about it (ever, or until many months or weeks later when the problem's discovered). So the easiness is illusory IMO. It seems to me that all of those other resource leaks are just leaking too slowly for us to discover yet.
> Each project involved tons of compiler fighting that I now know how to avoid or solve.
As frustrating as it is, it's easier than struggling with heisenbugs. Those can easily eat up days trying to reproduce issues.
> Cyclic data structures require an Arena or Rc<Refcell<>>.
I ask because I used the latter recently, after struggling to create lifetimes that were acceptable to the compiler. I only found it when I searched for 'cyclic data structures rust' on Google.
It would have been much nicer if the compiler could have noticed that I was trying to create a cyclic data structure and then have informed me on the different ways of doing so. Perhaps a misunderstanding on my behalf, but the error messages that I previously received felt like they were sending me on a wild goose chase.
I still don't think writing Rust code is hard, but there are definitely improvements which could be made to the compiler error messages.
It's hard because now you're in charge of building a malloc(3) implementation. Depending how important this data structure is you might need to worry about 1. about fragmentation 2. free space managment 3. adjacency and cache effects
So you've traded on hard problem for another hard problem. Not something I'd call a win.
Trees, indexes (including bitmaps), buffer allocators and even just keeping state in streaming cursors (Volcano model). Are all pain in the ass in to implement.
In fact implementing a Cursor trait for like 10 different Cursor (SrcView, Join, Project, Compute, ...) became painful because each cursor struct has it's own complex lifetime params and those lifetimes were trying to leak into the main Cursor trait.
Right now I can't do that without resolving to runtime using RefCell. In my case the plans could be entirely materialized at compile time, so not really a win.
The situation won't get better until there's some kind of abstraction over lifetimes (HKT?).
But do you learn to code in a different paradigm where you don't struggle by iterating on a compilable and runnable piece of code, but rather struggle on iterating on a non-compilable piece of code and "once it compiles, it works"?
This will get shorter as tooling gets better this year.
The problem is that instead of a general technique for solving borrow checker issues, every issue has a different specific technique. Hash map matches (like the article) require the entry API. Multiple mutable pointers into a buffer require either using a typed Vec or Vec#split_off. Cyclic data structures require an Arena or Rc<Refcell<>>. String casting issues sometimes require .map(|x| &x). Global state requires lazy_static. Nice error handling requires error_chain. When you run into problems with "Send" you need a Mutex. The list goes on...
It's like pure lazy functional programming in Haskell in that unlike other languages you can't just pick it up based on previous knowledge and patterns. You need an entirely new toolbox of pure algorithms and data structures, or in this case borrow-safe algorithms and data structures.