[Python-3000] Python 3000 Status Update (Long!)

[Nicko van Someren]

On 20 Jun 2007, at 15:44, Nick Coghlan wrote:

> Nicko van Someren wrote:
>> 	>>> sum(['a','b','c'])
>> 	Traceback (most recent call last):
>> 	  File "<stdin>", line 1, in <module>
>> 	TypeError: unsupported operand type(s) for +: 'int' and 'str'
>> 	>>> sum([["a"],[u'b'],[3]])
>> 	Traceback (most recent call last):
>> 	  File "<stdin>", line 1, in <module>
>> 	TypeError: unsupported operand type(s) for +: 'int' and 'list'
>
> You can already make the second example work properly by supplying  
> an appropriate starting value:
>
> >>> sum([["a"],[u'b'],[3]], [])
> ['a', u'b', 3]
>
> (and a similar call will also work for the new bytes type, as well  
> as other sequences)

The need to have an explicit 'start' value just seems wrong.  It's  
horribly inconsistent.  Things that can be added to integers work  
without initialisers but things that can be added to each other (for  
instance numbers in number fields or vectors in vector spaces) can  
not.  I think in most people's minds the 'sum' operation is like an  
evaluation of "+".join(...), you are sticking an addition operation  
between the elements of the list.  The need to have an explicit  
initial value means that sum() is not the sum function for anyone who  
does math in any sort of non-standard number space.

> Strings are explicitly disallowed (because Guido doesn't want a  
> second way to spell ''.join(seq), as far as I know):
>
> >>> sum(['a','b','c'], '')
> Traceback (most recent call last):
>   File "<stdin>", line 1, in <module>
> TypeError: sum() can't sum strings [use ''.join(seq) instead]

I can appreciate the value of TOOWTDI, and I appreciate that (in the  
absence of string concatenation by reference) the performance of  
string sum() would suck, but I still think that wilfully making  
things inconsistent in order to enforce TOOWTDI is going too far.

	Nicko