Once were numbers...

[David Vaughan]

> On Sunday, March 31, 2002, at 02:10 PM, Ian Summerfield wrote:
>
>> I'd like to see a change though, something that allows me to define a
>> precision to n decimal places.
>
>
Sorry to be contrarian but in relation to this now extended thread my 
principal amazement is at the breathless amazement of the discussants 
over this issue.

I also own an HP49G, which might be considered the ant's pants in 
programmable graphic engineering calculators. It is not an Infinite 
Precision machine but is considered accurate to 12 digits (and uses 13 
or 14 internally to achieve this). It also has an "exact" mode in which 
it returns integer fractions rather than decimal fractions for answers 
(e.g. enter 5 / 2 and it returns 5/2, not 2.5).

If you want precision, temper your expectations with a little knowledge 
of how computers work and recognise that 15 or more digits of actual 
precision means deadly accurate answers to 12 decimal places for pretty 
much anything that might be expected in a language of this nature.

The posts which raise problems of equalities are easily fixed if you pay 
attention to the fact that you are supposed to be comparing to zero or 
one decimal place (in the examples published). Do those who bemoan this 
"inaccuracy" and its influence on the expectations of the 
over-impressionable ever worry too much about the fact that MS Excel has 
corresponding internal errors?  Amazingly so, of course. It only runs on 
the same 32-bit computer with the same OS(s).

Time for me to retreat under cover before the flying brickbats land.

cheers
David

[post script: not aimed at you, ian, or indeed anyone in particular. I 
just needed to grab a post as a hook for the reply]