If loss of precision is a problem, you should be using extended precision
numbers (or rationals) for your values.

And, of course, you should be careful to use operations whose results are
extended precision numbers (or rationals) for the values you are working
with.

For example:

   <.1r2+10^13x

10000000000000


Thanks,


-- 

Raul



On Sat, Aug 20, 2016 at 12:45 PM, Erling Hellenäs <[email protected]>
wrote:

> Hi all !
>
> A number of problems like loss of precision, overflow, division by zero,
> exceptions and space requirements are relevant in this discussion?
>
> https://erlhelinfotech.wordpress.com/2016/05/30/infinity-
> nan-na-overflowloss-of-significance/
>
> Cheers,
>
> Erling
>
>
> On 2016-08-20 16:15, Raul Miller wrote:
>
>> I believe that types are a necessary evil which should be neglected
>> wherever possible.
>>
>> Sure, ok, it can be fun to make it painful to work with values of
>> different
>> types. But that mostly gets in the way when you do not want to focus on
>> that particular issue.
>>
>> The underlying myth, I think, is the very human "make it someone else's
>> problem" sort of thing we do when we want to focus on one particular
>> thing.
>> But there's no silver bullet and mostly we just wind up trying to sweep
>> problems under the carpet when we do this, rather than actually solving
>> useful problems.
>>
>> What we are really, doing, though, is working with engineering tradeoffs.
>> It's a matter of what you want to sacrifice and what you want to develop.
>>
>> Anyways, I'd probably go with another language (one which has strong type
>> theory roots) if I wanted to mess with types.
>>
>> (Though they do come up, also, if I get into implementing the language.
>> Which, I suppose, serves to emphasize the point...)
>>
>> Thanks,
>>
>>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to