[Patches] Patch: AttributeError and NameError: second attempt.

[Nick Mathewson]

Guido van Rossum <guido@python.org> wrote:
>[Nick Mathewson wrote:]
>>               When you throw an AttributeError on an Object now, you
>> need to generate a new string containing the Object's type and the
>> missing attribute.  With my patch, even _this_ work was postponed until
>> you printed the exception.
>> 
>> Nick
>
>Good point.  The less work is done when the exception is raised, the
>better.  (I think that since the traceback is available separately it
>shouldn't be necessary to reference the frame or the namespace from
>the exception.)

I agree on the frame and the namespace.  I can patch NameError
traceback printing to give these better error via tracebacks, either
directly or on-the-fly through an 'import warnings' mechanism.  BTW,
in addition to case, my latest patch also catches users who forget
that 'self.' should precede instance attributes; this is one of the
most common errors in comp.lang.python.

I have still have a problem, however.

My original patch handled AttributeError as well as NameError.  As
near as I can tell, there is no way to use tracebacks to find the
objects associated with AttributeErrors.  This means that in code like
          import string
          string.JOIN(....)
I have no way to tell that the 'JOIN' attribute was supposed to be
invoked on a string object.  In my original patch, I stored the
'string' object as an attribute of the exception, but if I can't do
that, I don't think I can do anything meaningful with AttributeError.

How should I proceed?

A-warning-mechanism-would-certainly-be-nice-ly y'rs,
-- 
Nick