Fix TypeError when converting Rust encoding errors to Python #5668
Conversation
src/err/mod.rs
Outdated
| /// str::from_utf8(bytes).map_err(|e| e.into_with_bytes(bytes)) | ||
| /// # } | ||
| /// ``` | ||
| pub trait IntoPyErrWithBytes { |
There was a problem hiding this comment.
Rather that adding a new trait I would prefer if we just made a constructor on PyUnicodeDecodeError which returns the PyErr. Internally it could make use of PyUnicodeDecodeError::new_utf8 to construct it.
There was a problem hiding this comment.
I've moved the implementation to a new constructor for PyUnicodeDecodeError.
I don't want to call PyUnicodeDecodeError::new_utf8 since it is fallible and I don't want to have to worry about handling errors while already creating an error. I've kept the implementation mostly the same using a private Utf8ErrorWithBytes struct which implements PyErrArguments and calls PyUnicodeDecodeError::new_err under the hood.
Lemme know if this approach sounds reasonable to you, or you want any more changes
staticintlucas
force-pushed
the
main
branch
3 times, most recently
from
feddff7 to
a8b289c
Compare
|
Will add some tests for my changes tomorrow to make codecov happy |
2 participants
Fixes #5651.
Fixes the implementation of
PyErrArgumentsforstring::FromUtf8Errorandffi::IntoStringError.Removes the broken implementation of
PyErrArgumentsforstr::Utf8Error,string::FromUtf16Error, andchar::DecodeUtf16Error.Adds a convenience trait called
IntoPyErrWithBytesforstr::Utf8Errorto allow easy conversion toPyErr. The implementation ofFrom<str::Utf8Error>forPyErrwas being used in a few places internally in PyO3, so this felt like the best solution. And unlike the UTF-16 errors,str::Utf8Errorcontains everything needed to construct a PythonUnicodeDecodeErrorexcept the source bytes. I exposed this API publicly since it's probably a pretty common use case (at least compared to UTF-16), but let me know if you want any changes (especially regarding API naming).