Question on error-stack

[TimDiekmann]

Hi!

I was looking at libraries that use error-stack to see if there’s anything we can learn about what to improve. I came across this and noticed that you recently switched back from error-stack to anyhow. Was there anything specific that factored into that decision? We are constantly trying to improve the crate (and in the very near future we will also release [email protected]) and any feedback, good or bad, is very helpful, especially if you ran into problems!

Thanks a lot!

[SeaDve]

error-stack shake up error handling as it allows me to think of possible error paths more thoroughly and think of error messages in a by-level-of-abstraction way. For a bonus, it is easy to attach arbitrary objects in error.

However, I moved back to anyhow as most of the time, creating a by-level-of-abstraction error message is not very important and could have some development overhead, especially when thinking of an error struct/enum name or when to change context. Additionally, it could hinder readability (at least for me), especially if there are a lot of lines that early return an error that does not return a Report directly, having to call report() manually for each line.

In general error-stack is indeed helpful. What I like the most is how it prints the error messages and the line where the context was changed, but its benefit can also be a disadvantage if the by-level-of-abstraction errors are not very important. Nonetheless, I could see it being very useful on other projects.

[TimDiekmann]

Thank you very much for your feedback, very appreciative!