Abstracting try..finally with generators (instead of macros)

[Andrew Dalke]

Beni Cherniavsky wrote:
> True, I forgot about for's fallback to `__getitem__`.

"Fallback"?  That's true now.  I'm getting too old, aren't I.  :)

> About the `sys` reference: I think that the class, being defined in the
> module (and referncing global variables) retains the module -- which in
> turn retains sys by having it imported -- so there shouldn't be a problem.
> Do I miss anything?

I .. don't think it keeps a reference to the module, only the name of
the module.  Else in the pre-garbage collection days there would be
a lot of cycles.  (I think?)

 >>> spam = Spam()
 >>> dir(spam)
['__doc__', '__module__']
 >>> spam.__module__
'__main__'
 >>> type(spam.__module__)
<type 'str'>
 >>> spam.__class__
<class __main__.Spam at 0x8103114>
 >>> dir(spam.__class__)
['__doc__', '__module__']
 >>> spam.__class__.__module__
'__main__'
 >>> type(spam.__class__.__module__)
<type 'str'>
 >>>

The problem is that objects in each module need to be deallocated
during shutdown.  How's that done?

I dug up Guido's essay on this from the 1.5 days
   http://www.python.org/doc/essays/cleanup.html
Don't know how true it is now.

It has

 > M5. Clear all remaining modules carefully, except __builtin__ and sys.

But it also has

 > M2. The three standard I/O files (sys.stdin, sys.stdout and
 > sys.stderr) are restored to their initial values (which are
 > saved as sys.__stdin__, sys.__stdout__ and sys.__stderr__,
 > respectively, when the interpreter starts). If any of the
 > initial values is unavailable, the corresponding object is
 > set to None. [New.]

So there's no need to worry - 'sys' isn't cleared until the
very end.

					Andrew
					dalke at dalkescientific.com