Constraints on __sub__, __eq__, etc.

[Peter Otten]

Roald de Vries wrote:

> On Feb 18, 2010, at 5:28 PM, Stephen Hansen wrote:
>> On Thu, Feb 18, 2010 at 8:19 AM, Andrey Fedorov
>> <anfedorov at gmail.com>wrote:
>>> It seems intuitive to me that the magic methods for overriding the
>>> +, -, <, ==, >, etc. operators should have no sideffects on their
>>> operands. Also, that == should be commutative and transitive, that
>>> > and < should be transitive, and anti-commutative.
>>>
>>> Is this intuition written up in a PEP, or assumed to follow from
>>> the mathematical meanings?
>>
>> It may be intuitive to you, but its not true, written down anywhere,
>> nor assumed by the language, and the mathematical meaning of the
>> operators doesn't matter to Python. Python purposefully does not
>> enforce anything for these methods.
> 
> Still, it's clear that (for example) '==' is not just a normal
> function call. Look at this example (in ipython):
> 
>  >>> False == False == False
> True
>  >>> True == False == False
> False
>  >>> (True == False) == False
> True
> 
> Anybody knows how why this is so?

As Chris said

expr1 <op1> expr2 <op2> expr3 <op3> ...

is resolved as 

(expr1 <op1> expr2) and (expr2 <op2> expr3) and (expr3 <op3> ...

where each exprN is evaluated just once.


For this to become the obvious way you have to look at interval checks like

a < b < c

Peter