Re: How PostgreSQL's floating-point hurts everyone everywhere

Peter Eisentraut wrote:
>
> peter=# select 1.0/0.0;
> ERROR:  float8div: divide by zero error
>
> Division by zero is not an "error" in floating point arithmetic.

    No?

    Oh, then 7 = 5 because:

        Assuming              2a = b                         | * 2
                              4a = 2b                        | + 10a
                             14a = 10a + 2b                  | - 7b
                        14a - 7b = 10a - 5b                  | ()
                      7 (2a - b) = 5 (2a - b)                | / (2a - b)
                               7 = 5

    In  the  given  context,  you  should find the mistake pretty
    easy.  Maybe the message should be changed to

        ERROR:  floatdiv: divide by zero nonsense
                                         ^^^^^^^^

    because a division by zero results in nonsense? Or should  it
    just  return a random value? What is the result of a division
    by  zero?   It's  definitely  not  infinity,  as  the   above
    demonstrates!

    You  might  be looking from a managers PoV.  Managers usually
    use this kind of arithmetic to choose salaries.  Any engineer
    knows that

        work = power * time

    We all know that time is money and that power is knowlede. So

        work = knowledge * money

    and thus

                  work
        money = ---------
                knowledge

    Obviously, the lesser you  know,  the  more  money  you  get,
    independant  of  the  work  to  be  done. As soon as you know
    nothing (zero), any money you get for the work is nonsense!

    This applies well to the real world, so it makes  sense  from
    an OO PoV.  But in science, I think it's still an error.

    Since  PostgreSQL  is an ORDBMS (not an OODBMS), I think it's
    correct to report an error instead of returning some  random.

    :-)


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck@Yahoo.com #