Re: (*) -> 1

[Óscar Fuentes]

Andreas Eder <a_eder_muc@web.de> writes:

> On Mi 18 Jan 2023 at 15:07, Óscar Fuentes <ofv@wanadoo.es> wrote:
>
>> Michael Heerdegen <michael_heerdegen@web.de> writes:
>>
>>> Óscar Fuentes <ofv@wanadoo.es> writes:
>>>
>>>> Even a programmer can't assume that associativity and commutativity
>>>> can be used on a long, finite, sum.
>>>
>>> As long as it's a sum, of course can a programmer assume that.
>>
>> Of course not! Otherwise your numerical simulation can go very wrong :-)
>>
>> (hint: numbers in computers are approximate, adding small numbers to
>> large numbers is a bad idea.)
>
> Only floating point numbers.

Which proves my point :-) And you say "only" as if floating point
numbers were rarely used.

> Integers are exact and in lisps like elisp
> there is no artificial size limit.

If your calculation consists on adding arbitrary precision integers,
*that's* a rarity. If your calculation adds fixed-width integers (which
includes platform integers), you definitely need to worry about order of
terms.