[Tutor] Standard way to append to a list? ['+=' is in-place concatenation]

[Danny Yoo]

On Wed, 16 Apr 2003, Danny Yoo wrote:

> Wait a minute... I can't seem to prove this by overriding the extend()
> method!
>
> ###
> >>> class mylist(list):
> ...     def extend(self, other):
> ...         print "I'm extend()!"
> ...         list.extend(self, other)
> ...
> >>> l = mylist([1, 2, 3])
> >>> l += [4]
> >>> l
> [1, 2, 3, 4]
> >>> l.extend([4])
> I'm extend()!
> ###
>
> Weird!  I'm using Python 2.2.1 here; can anyone confirm if this is
> happening with Python 2.2.2 or 2.3alpha?  Does this look like a bug to
> anyone else?  ... Checking SourceForge... nope, I don't see anything
> about this.


Hi everyone,


Hmmm... I played around with this a little more.  It does work if we
override __iadd__ (in-place adding):

###
>>> class mylist(list):
...     def __iadd__(self, other):
...         print "I'm __iadd__()"
...         return list.__iadd__(self, other)
...
>>> l = mylist([1, 2, 3])
>>> l += [4]
I'm __iadd__()
>>> l
[1, 2, 3, 4]
###

__iadd__ is briefly mentioned in the Python Standard Reference here:

    http://python.org/doc/current/ref/sequence-types.html



So the system is not totally broken.  What appears to be the issue is that
__iadd__() for lists is hardcoded to use the primitive C function
'listextend_internal()'.  I feel, though, that a list subclass should
really check to see if the extend() function has been redefined when it's
doing a '+=' operation, to maintain polymorphic behavior.  Dunno how that
will affect performance, though.


In any case, I'll send a bug report to SourceForge later tonight and see
how the implementors feel about this... *grin*