Using builtin exception information is tricky as it 
consists of:
 a) the type of exception (easily accessible)
 b) the args attribute = a 1 element tuple with a string

1st example:

try:
    print foo
except NameError, e:
    print e.args
    symbol = e.args[0][17:-16]
    print symbols
==> 
("NameError: name 'foo' is not defined", )
foo

It would be nicer to have:
e.args = ("NameError: name 'foo' is not defined", "foo")
The first element being the current string for backward 
compatibilty.
=============================

2nd example:

try:
    (4).foo
except NameError, e:
    print e.args
==> ("'int' object has no attribute 'foo'",)

It would be nicer to have:
e.args = ("'int' object has no attribute 'foo'", 4, "foo")
Again, the first element being the current string for 
backward compatibilty.

=============================

Moreover, in the documentation about Exception, I read
"""Warning: Messages to exceptions are not part of the 
Python API. Their 
contents may change from one version of Python to the 
next without warning 
and should not be relied on by code which will run under 
multiple versions 
of the interpreter. """

So even args could not be relied upon !
But it also means that there is no need to be backward 
compatible (I am playing devil's advocate, backward 
compatibility is important !)

Seb

ps: There may be problems (that I am not aware) with
 a) an exception keeping references to other objects
 b) C API that can throw only exceptions with strings
 c) a specific advantage of having strings only in builtin 
exceptions

Logged In: YES 
user_id=80475

This looks like a good idea to me.

Logged In: YES 
user_id=89016

+1

This should probably be part of Brett's Py3000 exception PEP.

Candidates I can think of are:

KeyError(obj=..., key=...)
IndexError(obj=..., index=...)
IOError(file=..., code=...)
NameError(name=...)

Regenerating the message would be done with __str__().

I'm not sure how this would work on the level of the C API.

See also issue 5353.

Beware that making exceptions hold on arbitrary objects increases the
possibility of delayed collection and reference cycles, though.
(especially in trunk where the "current" exception can last after the
end of an except block)

Hi, I attempted a solution for this a while back

Attached is a (git diff based) patch for review

Changes can also been seen at: https://bitbucket.org/bloomberg/cpython/pull-request/2/http-bugspythonorg-issue1182143-add-name/diff
(but note the changes in Misc/NEWS and Tools/msi/msi.py are not related / somehow got pulled into my changes by mercurial)

Please review/let me know of any other process required. Thanks!

Add patch generated via mercurial

Heh, just noticed this issue passed its 10 year anniversary!

If someone has time to review my patch, that would be much appreciated. Thanks!

Reviewer please :) 

(or, advice on how I can get this to proceed, thx)

What does your patch implement?  It's not clear from the issue discussion that an API was decided on.

Thanks for the response.

In terms of python API changes - I exposed the 'name' attribute on UnboundLocalError and NameError exceptions. 

(This can be seen (& is verified) by the changes to test_exceptions.py exceptions test cases in the patch)

In terms of the wider changes in the patch - I factored UnboundLocalError and NameError to have specific native implementations, rather than been based off SimpleExtendsException. And thus able to accept the name of the variable causing the exception.

And I agree with your comment on the exact API never been agreed on. Since there are now native implementations of UnboundLocalError and NameError, I can happily change their implementation to expose whatever can be decided to be better (comments?!)

Thanks, George

Bump on this one please

I would suggest you to bring this up to python-dev (or perhaps python-ideas?).
Also note that Brett proposed something similar in #18162 and other related issues (#18156, #18163, #18165, #18166).