-
-
Notifications
You must be signed in to change notification settings - Fork 3.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[bevy_ui/layout] Replace unwrap, expect, and panic with Result #12805
base: main
Are you sure you want to change the base?
[bevy_ui/layout] Replace unwrap, expect, and panic with Result #12805
Conversation
Remove the need to manually update GlobalTransform in bevy_ui/layout tests
7434453
to
b6e1f00
Compare
b6e1f00
to
f9c16ef
Compare
#[error("Unreachable")] | ||
/// Code paths that are deemed unreachable but avoid `unreachable!` to allow caller to handle and proceed | ||
Unreachable, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This... doesn't really make sense to me. Unreachable code by definition has no flow control paths leading to it. How is a caller expected to handle this error? Why is avoiding unreachable!()
even a goal? You can't get rid of all possible panics, std
is riddled with panicking API...
The same applies to internal state errors. If an internal invariant doesn't hold that is a bug, not an exception, and I don't understand how application code is expected to handle this. Panicking with an appropriate message is the most reasonable course of action in these scenarios IMO.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Well, you're not wrong, and I agree. I didn't mean all panics, just the ones created by unwrap and such. However, I'm clearly getting carried away by including Unreachable
in its scope. Thank you for calling me out on that early.
My original intentions were to bubble everything up to the caller (in this case, the ui_layout_system
), expecting it to return a Result
as well. Then that could just let unhandled errors trigger error logs (via ui_layout_system.map(error)
) instead of crashing. That way, it could maybe allow other UI entities to finish processing. But that increases the complexity as seen here.
You make another excellent point about "how can a caller be expected to handle InvalidState
?"
Context: Right now, the internal maps are exposed as pub(super)
. This could result in someone adding code outside the structs associated functions that interact directly with the internal map or the specific sequence of calls initiated by the calling system that creates an InvalidState
. For now, InvalidState
is a marker for something intended to be unreachable but not guaranteed. If the maps can be made private and the exposed API made to be infallible, then I think replacing InvalidState
back to panic!
or unreachable!
depending on the context would be appropriate.
There's not much expectation of actually handling InvalidState
as the caller. Not saying we should, but theoretically, we could attempt to recreate the UiSurface
for transient issues like a child UI node being promoted to the root UI node.
When I circle back around to this I'll
- remove
Unreachable
and revert tounreachable!
.
This is 5 of 5 iterative PR's that affect bevy_ui/layout
Blocked by [bevy_ui/layout] UiSurface update internal mappings #12804[bevy_ui/layout] Update internal mappings #16362Diff to parent PR
This is a very ambitious attempt to add error handling to
UiSurface
and theui_layout_system
(WIP) instead of panickingObjective
unwrap
,expect
,panic!
, withResult
unwrap
considered harmful #12660Solution
unwrap
,expect
,panic!
, withResult
UiSurfaceError
Changelog
Migration Guide
TBD - probably should be saved for 0.14