Menu
▾
▴
Re: [Maxima-discuss] A proposal to solve Re: float(exp(10000)) works, but float(exp(10000/3)) doesn't
|
From: Stavros M. (Σ. Μ. <mac...@al...> - 2016-01-29 21:26:42
|
tl;dr: Lots of clever ideas, but I suspect that they will confuse many more
users than they help.
On Fri, Jan 29, 2016 at 11:43 AM, Raymond Toy <toy...@gm...> wrote:
> >>>>> "Stavros" == Stavros Macrakis <(Σταῦρος Μακράκης)" <
> mac...@al...>> writes:
> Stavros> On Thu, Jan 28, 2016 at 7:04 PM, Raymond Toy <
> toy...@gm...> wrote:
>
...
> Stavros> we could define 0.55r0 (r for rational) and 0.55x0 (x the
> Stavros> Roman numeral 10). And then make the default be... which
>
> Using "r" has been proposed before. Nothing came of it, AFAICT. I
> actually like "r". Not quite sure what the difference between 0.55r0
>
> and 0.55x0 is, though.
>
"r" is A/B where A and B are integers, but the input/output form is
decimal. 0.55r0 = 11/20. 1/3.0r0 prints as 0.333333, but is represented
internally as exactly 1/3; "x" is A*10^B, so 0.55x0 = 5.5x-1 = 55.0x-2 =
55*10^-2
> Stavros> one? And how many new users would use these reasonably?
> Default for what?
>
The default for 0.55. Is that interpreted as 0.55x0? 0.55r0? 0.55e0?
Stavros> Some users would be very happy that sin(%pi/3.0) =>
> Stavros> sqrt(3)/2, others would be dismayed that it is not
> Stavros> 0.866.... Will users be pleased or dismayed when
> Stavros> sin(0.55r0) simplifies to ... sin(0.55r0) rather than to
> Stavros> 0.522687...?
>
> It just has to be explained.
A small matter of documentation.... The longer and more complex the
documentation, the less likely it will be read. Though I can imagine that
1% of our users will benefit from these sophisticated options, I suspect
that 90% will be oblivious to them, and 9% will be confused by them.
> Is there any language that already
>
> supports 0.55r0?
>
Not as far as I know.
> Sort of. When I enter 1.0, I mean exactly that. The floating point
> number closest to 1.0. It is NOT approximate. It may have come from
> other sources, but once I type that into maxima, the only reasonable
> way to proceed is to treat it as exactly that number. It's up to the
> user to keep track of the actual interpretation. Or use interval
> arithmetic to denote the approximation.
>
Yes, I agree with all that.
It's like if someone converts a 1 foot to 30.48 centimeters. The
> conversion is exact (I think), but is that measurement of 1 foot
> really accurate to 0.1 mm? Probably not, but there's no way to know
> without additional info.
Yes.
|
