Adopt stacked error to snafu

[waynexia]

Hi @shepmaster! I'm the author of that post on hackernews. I'm thrilled to see snafu and this mechanism being discussed so much.

Snafu has no doubt helped us a lot in organizing complex error variants and saved time on debugging. I'm wondering if we can contribute it to the upstream (here) and let people use and evolve it together. From my side, I have two concrete motivations for doing this, (1) implementing it out of snafu requires some hacking code and (2) pushing it to a more common location can further accomplish the error stack, the current stack will become chaos when getting through different error libs from different dependencies.

This might not be a trivial change. I'd like to hear some suggestions and/or feedback from your side on whether this is viable and if so how can we get started ❤️

[Enet4]

Thank you for sharing your use case and experience in the use of Snafu!

I have been reading your blog post on & off between breaks in order to grasp what was developed on top of the library, and I'm probably still not quite there yet. It would be really helpful if you can explain it here in a few sentences.

From what I could gather (please correct me if I'm missing something in this rationale), the concept of stacked error is meant to be the StackedError trait with the corresponding procedure macro to implement it. The capabilities in this trait already exist in the library:

• next and last are methods for traversal of the error source chain. The built-in alternative is ErrorCompat::iter_chain, the output is usually a ChainCompat which implements Iterator.
• debug_fmt offers error reporting, which is already implemented in Report.

[si14]

I got back to that post, searched for a discussion here, so here I am a year later. My two pennies:

I suppose what would be nice to have is a "blessed" way to use the library in an app, rather than assembling the architecture described in the blogpost from the existing pieces. Report is a step in that direction, but (from what I understand) one difference is that Greptime's implementation requires location to be set in all cases, which allows a more uniform error reporting.

There was a recent announcement of another library that defaults to a StackedError-like reporting, so it seems that there's a demand for it: https://old.reddit.com/r/rust/comments/1p245cs/announcing_rootcause_a_new_ergonomic_structured/

Also related: #501