It is interesting that neither the talk nor the questions from the audience rise the question of the error handling.
In Go returning an error implies that the caller will need to add 3 lines of the source for each call to propagate the error. This is really ugly. Rust managed to reduce that source overhead down to the single extra ? character while keeping all the benefits of explicit error handling. I wish Go gets something similar.
In my experience creating robust, usable code that responds properly to adverse conditions always requires tons of error handling. The closer to the problem you deal with it the more accurate your diagnostics can be. Otherwise it becomes very difficult to provide meaningful feedback to the user.
Nothing in your comment counters the GP’s statement that Rust manages to accomplish this with a single character, simultaneously avoiding needlessly obscuring the happy path and without the downsides of non-local error handling.
On a related note, Rust's hygenic macro system is simply amazing.
If you're only familiar with macros from C or C++ take a look. The basic idea is that your macro fictions don't use input text and output text. Instead they accept an Abstract Syntax Tree and return a known AS THE. That way the expansion plays well with the grammer.
It also manages separate variable scopes well. You only get bleed if you want it.
I used to shudder at the thought of macros but I'm convinced. Java or Go could use them reasonably I think.
I'm not a rust programer so my understand is probably a little wrong. I'm not intimately familiar with all of the details. Correct me if I'm mistaken!
>You can also write it Go style if you prefer; we started off with exactly that, then moved to a macro, then turned that macro into ?.
Interesting. Does the Rust book (you wrote), explain the difference between the Go style and the current Rust style, and explain how to use the Rust style (well)?
Asking because I've found this Some / None stuff (also there in other languages like F#) a little tricky to understand (although I have not spent a lot of time on it yet).
We talk about it a little, but the best resource on this topic by far is https://blog.burntsushi.net/rust-error-handling/ which works the whole way from first principles. It was in the first edition of the book but due to space requirements didn’t really fit in the second edition.
Thank you! You know what’s also “explicit”? Typing out assembly by hand. Somehow everything lower level than Go is totally fine to hide from the developer, but any ergonomics improvements at all and go would “no longer be simple and explicit”.
As far as I'm concerned 'explicit' just means that the behaviour under discussion is caused by some piece of source that talks about that behaviour and not something else.
Whether the bit of source is a single character or a paragraph of prose is about verbosity rather than explicitness as long as the behaviour under discussion is caused by a piece of source with precisely that purpose.
I don't know, most of the time, I add information to the error before propagating it:
if f, err := os.Open(path); err != nil {
return nil, fmt.Errorf("could not open path: %v", err)
}
This is the vast majority of my error management code in go. Is rust less verbose in this regard? AFAIK, the question mark operator is equivalent to `if err != nil { return nil, err }`.
In Go returning an error implies that the caller will need to add 3 lines of the source for each call to propagate the error. This is really ugly. Rust managed to reduce that source overhead down to the single extra ? character while keeping all the benefits of explicit error handling. I wish Go gets something similar.