Open d-e-s-o opened 2 months ago
Is there a reason you are using try_parse_from
, rather than parse
?
Is there a reason you are using
try_parse_from
, rather thanparse
?
Yep.
This isn't the only peculiarity about claps' error reporting that doesn't make it work well with policies like anyhow
. For example, you likely should condition your exit code based on er.use_stderr()
(not a great name) or else --help
is considered an error when its generally recognized as not one.
We need more information to better understand your use case to determine what, if anything, we should be done about this.
This has very little if not nothing to do with anyhow
from what I can tell.
My use case is using regular error reporting paths as every other crate does and have a single program exit path. If every crate thinks it knows best when to exit then I have no control over or idea of what is going on. The same is true (and, to the best I can tell, accepted practice) for panics: you can unwrap or panic on invariants and everything else (certainly everything controllable by a user) shouldn't panic, but report a handleable error.
I assume that is the reason why the try_parse_from
API exist in the first place, right?
Please complete the following tasks
Rust Version
rustc 1.79.0-nightly (129f3b996 2024-06-10) (gentoo)
Clap Version
4.5.4
Minimal reproducible code
https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=6221486eaf814c633157f1c9fdf27c12
Steps to reproduce the bug with the above code
cargo run
Actual Behaviour
Prints:
Expected Behaviour
Prints:
Additional Context
The "error:" prefix is duplicated. I find this behavior irritating. One prefix seems to be added by the standard library, the other by
clap
(I don't believeclap
is alone with such behavior)In my opinion the standard library's behavior is questionable -- it's just way too opinionated and this is a good example -- but I wanted to get your guys' take and have an issue to reference to make the case for changing it (which I suppose may be a hard sell irrespectively).
In general, though, it also feels as if just the typical wording of "failed to [...]" or similar would be sufficient for convey the issue. Meaning that there is no need for an "error:" at all, even from the
clap
side. It just depends too much on the context how the error is reported whether it makes sense to include such a prefix or not.Any opinions/comments/thoughts?
Debug Output
No response