`cbor2.CBORDecodeError` should be a subclass of `ValueError`

[vytas7]

Things to check first

• I have searched the existing issues and didn't find my bug already reported there

• I have checked that my bug is still present in the latest release

cbor2 version

6.0.1

Python version

3.12

What happened?

With the recently released cbor2 6.0.x, cbor2.CBORDecodeError is not a subclass of ValueError.

When using data serialization libraries like the stdlib's json, it is a common convention that .loads(...) or similar raises a subclass of ValueError (the passed value was incorrect). This used to work fine in cbor2 as well until the very recent 6.0.x releases.

How can we reproduce the bug?

Using older cbor2 (5.x), this would have worked just fine:

try:
    cbor2.loads(b'\xde\xad\xbe\xef')
except ValueError:
    print('Validation failed')

On 6.0.x, the decoding error is not caught as ValueError.

[agronholm]

Why are you reporting this as a bug? This is documented behavior. And this is a new major version which, in semver conventions, allows for breaking changes.

[agronholm]

Also, I've never heard of such a convention. Why didn't you bring this up during the release candidate phase that lasted for a month?

[vytas7]

Hey @agronholm, and thanks for the prompt response.
It is totally fair that you can make breaking changes in a SemVer major version; I was more appealing to the unwritten convention that may indeed not exist. The majority of data deserialization and validation libraries do follow it though, as mentioned, including the stdlib's json.

I don't use cbor2 myself per se (although I am potentially looking into it), we use it as an example of WebSocket binary media handler tests in Falcon. So it never gets a chance to test pre-releases.

[agronholm]

For the record, CBORDecodeError was never a subclass of ValueError. The now-eliminated CBORDecodeValueError was. It was never clear when to raise which, so I figured it would make it simpler for downstream users to only deal with a single exception class.

[vytas7]

Feel free to remove the bug label, but then it is my suggestion for improvement 🙂
From the popular libraries I know (like various JSON implementations, MessagePack, YAML, etc), only msgspec was not inheriting from ValueError, but that was changed in the recently released 0.21.0.

[agronholm]

Inheriting from multiple exception classes is tricky with PyO3 so it's not clear how feasible this is.