[Python-ideas] Make max() stable

[David Mertz]

On Tue, Jan 21, 2014 at 1:11 PM, David Mertz <mertz at gnosis.cx> wrote:

> Slightly related, here's an invariant that I've wished would hold for a
> decade, but isn't likely to, even in Python 4:
>
>   assert all(not x<y for x,y in zip(a,b)) if a==b else False
>

Ooops, I meant:

  assert all(not x<y for x,y in zip(a,b)) if a==b else True

But the point is that it fails where a==b because equal elements may not be
"not unequal."


But this is just a question of inequality versus identity and that sets and
> dictionaries are, IMO, too sloppy about that.  That is, they behave exactly
> as documented and as the BDFL has decreed, but I still feel uneasy about:
>

>   >>> a = {1, 1+0j, 2}
>   >>> b = {1+0j, 1, 2}
>   >>> a
>   {(1+0j), 2}
>   >>> b
>   {1, 2}
>   >>> a == b
>   True
>
>
> On Tue, Jan 21, 2014 at 1:02 PM, David Mertz <mertz at gnosis.cx> wrote:
>
>>  Imagine implementing min and max this way (ignoring key= and the
>>> possibility of a single iterable arg):
>>>
>>> lst = sorted((x,y))
>>> assert lst == [min(lst), max(lst)]
>>>
>>> will pass for any x and y.
>>>
>>
>> Well, that's not possible, of course, if one is willing to be slightly
>> perverse:
>>
>> >>> @total_ordering
>> ... class SomewhatOrdered(object):
>> ...     def __init__(self, val):
>> ...         self.val = val
>> ...     def __eq__(self, other):
>> ...         return self.val == other.val
>> ...     def __lt__(self, other):
>> ...         return (self.val, random()) < (other.val, random())
>> ...     def __repr__(self):
>> ...         return repr(self.val)
>> ...
>> >>> x, y, z = map(SomewhatOrdered, (1, 1.0, 2))
>>
>> But even if you were slightly less perverse than this, *sets* (and
>> set-like collections) return elements in indeterminate order which the
>> language does not guarantee.  In particular, I do not think we are promised
>> this holds:
>>
>>   assert tuple(a)==tuple(b) if a==b else False
>>
>> I can certainly construct a class where that won't hold (i.e. a set-like
>> class that iterates in a non-deterministic order; this need not even be
>> perverse, e.g. if it is 'AsyncResultsSet' that gets its data from I/O
>> source or parallel computations).
>>
>> I have a feeling I could find plain old Python sets that would fail that,
>> but I'm not sure about it.
>>
>> --
>> Keeping medicines from the bloodstreams of the sick; food
>> from the bellies of the hungry; books from the hands of the
>> uneducated; technology from the underdeveloped; and putting
>> advocates of freedom in prisons.  Intellectual property is
>> to the 21st century what the slave trade was to the 16th.
>>
>
>
>
> --
> Keeping medicines from the bloodstreams of the sick; food
> from the bellies of the hungry; books from the hands of the
> uneducated; technology from the underdeveloped; and putting
> advocates of freedom in prisons.  Intellectual property is
> to the 21st century what the slave trade was to the 16th.
>



-- 
Keeping medicines from the bloodstreams of the sick; food
from the bellies of the hungry; books from the hands of the
uneducated; technology from the underdeveloped; and putting
advocates of freedom in prisons.  Intellectual property is
to the 21st century what the slave trade was to the 16th.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-ideas/attachments/20140121/f431db4d/attachment.html>