[ python-Patches-1727209 ] First steps towards new super (PEP 3135)

Patches item #1727209, was opened at 2007-05-28 22:30
Message generated for change (Comment added) made by gvanrossum
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=305470&aid=1727209&group_id=5470

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Core (C code)
Group: Python 3000
>Status: Closed
>Resolution: Accepted
Priority: 5
Private: No
Submitted By: Guido van Rossum (gvanrossum)
Assigned to: Guido van Rossum (gvanrossum)
>Summary: First steps towards new super (PEP 3135)

Initial Comment:
A half-assed start of an implementation for super.

This is lacking parser support; instead, for now, you have to have a keyword-only argument named super, e.g.

def foo(self, arg, *, super):
    return super.foo(arg) + 1

Note that the first version changes method lookup for bound method lookup but not yet for unbound method lookup.  E.g. if f is defined in class B, and class C derives from B, C().f returns a bound method object whose im_class attribute is set to B (as opposed to C as it used to be) but C.f returns an unbound method object with im_class set to C.  (I hope to this in a later version.)

----------------------------------------------------------------------

>Comment By: Guido van Rossum (gvanrossum)
Date: 2007-06-11 17:20

Message:
Logged In: YES 
user_id=6380
Originator: YES

I've checked this into the p3yk branch now.  Tim Delaney will update the
PEP.  Note that the PEP number has changed; it's 3135 now.

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-06-06 17:00

Message:
Logged In: YES 
user_id=6380
Originator: YES

Sorry, forgot to add test_super.py to the patch.
File Added: super2.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-06-06 16:57

Message:
Logged In: YES 
user_id=6380
Originator: YES

OK, here's today's final version (I hope). This fixes test_scope.py
(wasn't too hard after all) and updates the docstring for super(). 
Question to all: is writing super() acceptable?
File Added: super2.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-06-06 16:01

Message:
Logged In: YES 
user_id=6380
Originator: YES

Here's a better version that lets you use super() as a shorthand for
super(__class__, <firstarg>).  test_scope.py still fails though.
File Added: super2.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-06-06 14:53

Message:
Logged In: YES 
user_id=6380
Originator: YES

I'm starting a new series of patches, super2.diff. These use a completely
different tack, inserting a cell variable __class__ into the class shared
by all the methods and initialized to the class after it's constructed. 
Right now, this allows us to write super(__class__, self).method() -- I
want to evolve it so that eventually you can write super.method() and super
will do the lookup of __class__ and the first argument itself.

There's one failing test, testFreeVarInMethod in test_scope.py.  This
seems significant but I want to upload a checkpoint first.
File Added: super2.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-31 19:00

Message:
Logged In: YES 
user_id=6380
Originator: YES

This breaks all class methods (as Phillip Eby pointed out):

class C:
   @classmethod
   def cm(cls): return cls.__name__
class D(C): pass
print(D.cm(), D().cm())

This should print "D D" but with the patch it prints "C C".

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-31 07:45

Message:
Logged In: YES 
user_id=6380
Originator: YES

Oops, a last-minute change didn't work out.  This version is better.
File Added: super.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-31 07:24

Message:
Logged In: YES 
user_id=6380
Originator: YES

Here's a version that modifies the grammar so 'super' is a keyword. The
compiler automatically inserts a 'super' keyword-only argument into a
function's signature when the 'super' keyword is used in its body.  It
works!

This requires the p3yk branch at -r55692 or newer.
File Added: super.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-29 02:08

Message:
Logged In: YES 
user_id=6380
Originator: YES

Here's a new version.  Forget about sup.py, the demos are now in
Lib/test/test_super.py.

Only a few tests fail due to the semantic change (about six files have 1-2
failures each).  I think this is acceptable.

File Added: super.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-28 23:23

Message:
Logged In: YES 
user_id=6380
Originator: YES

And a new sup.py example demonstrating that the new patch works.
File Added: sup.py

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-28 23:22

Message:
Logged In: YES 
user_id=6380
Originator: YES

Here's a new version of the patch where C.f.im_class (in the above
example) correctly returns B.
File Added: super.diff

----------------------------------------------------------------------

Comment By: Guido van Rossum (gvanrossum)
Date: 2007-05-28 22:31

Message:
Logged In: YES 
user_id=6380
Originator: YES

Here is sup.py, a small demo of how this is supposed to work.
File Added: sup.py

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=305470&aid=1727209&group_id=5470