[Python-Dev] unicode Exception messages in py2.7 (original) (raw)

Steven D'Aprano steve at pearwood.info
Fri Nov 15 00:58:42 CET 2013

On Thu, Nov 14, 2013 at 04:55:19PM -0500, Tres Seaver wrote:

Fixing any bug is "changing behavior"; 2.7 is not frozen for bugfixes.

It's not a given that the current behaviour is a bug. Exception messages in 2 are byte-strings, not Unicode. Trying to use Unicode instead is not, as far as I can tell, supported behaviour.

If the exception message cannot be converted to a byte-string, suppressing the display of the message seems like perfectly reasonable behaviour to me:

py> class NoString: ... def str(self): ... raise ValueError ... py> msg = NoString py> msg = NoString() py> print msg Traceback (most recent call last): File "", line 1, in ? File "", line 3, in str ValueError py> raise TypeError(msg) Traceback (most recent call last): File "", line 1, in ? TypeErrorpy>

although it would be nice if a newline was used so the prompt was bumped to the next line.

The point is, I'm not convinced that this is a bug at all.

The real question is whether third-party code will break when the now-empty error messages appear with '?' littered through them?

This behaviour goes back to at least Python 2.4, the oldest version I have easy access to at the moment that includes Unicode. Given that this alleged bug has been around for so long, I don't think that it effects terribly many people. That implies that fixing it won't benefit many people either.

About the only things I can think of which might break would be doctests, but people expect those to break across third-dot releases of Python

Which people? I certainly don't expect doctests to break unless I've done something silly.

(one reason why I hate them). Exception repr is explicitly not part of any backward-compatibility guarantees in Python.

Do you have a link for that explicit non-guarantee from the docs please?

-- Steven

More information about the Python-Dev mailing list