[Numpy-discussion] Ransom Proposals

[Tim Hochberg]

Travis Oliphant wrote:

> Tim Hochberg wrote:
>
>>> Charles R Harris wrote:
>>>
>>>     >>> l = list(a)
>>>     >>> l
>>>    [999, 1, 2, 3, 4, 5, 6, 7, 8]
>>>     >>> a
>>>    array([999,   1,   2,   3,   4,   5,   6,   7,   8])
>>>     >>> l += a
>>>     >>> l
>>>    array([1998,    2,    4,    6,    8,   10,   12,   14,   16])
>>>     >>> a
>>>    array([999,   1,   2,   3,   4,   5,   6,   7,   8])
>>
>>
>>
>>
>> Let me add that I think that this is pretty dubious, so if this is a 
>> new feature, perhaps we should revert it before it becomes entrenched.
>
>
>
> I don't think it's a new feature, but it's simply the result of
>
> l += a being translated to
>
> l = l + a  # lists don't have in-place add's
>
> Numeric has this behavior as well.


That's what I would expect, but Charles claimed it was resulted in an 
error of some sort, so I was wondering if it was new.

I was also wondering if we could disable it somehow by proper finagaling 
of the various add and iadd ops, but I can't really see how. However, I 
did stumble on this odd behaviour:

        class reporter(object):
            def __init__(self, name):
                self.name = name
            def __add__(self, other):
                print "add called on", self.name, other.name
                return NotImplemented
            def __radd__(self, other):
                print "radd called on", self.name, other.name
                return NotImplemented
            def __iadd__(self, other):
                print "iadd called on", self.name, other.name
                return NotImplemented   
        class reporter2(reporter):
            pass
        a = reporter("A")
        b = reporter2("B")
        a += b
        print a

    ==>

    iadd called on A B
    add called on A B
    radd called on B A
    iadd called on A B
    NotImplemented

Shouldn't that raise TypeError? Yes? No? I'll go enter it as a bug, but 
I want to make sure I'm not missing something stupid.


-tim