Re: [AC21.5] Texinfo manual kilo-patch

"Stephen J. Turnbull" <[email protected]> writes:

>>>>>> "Hrvoje" == Hrvoje Niksic <[email protected]> writes:
>
>     Hrvoje> Think of it this way: any FP number can be converted to a
>     Hrvoje> rational without loss of precision, while the reverse
>     Hrvoje> doesn't hold.
>
> True, but you do lose information: all FP numbers by definition have
> limited, known precision, but conversion to a rational claims that
> you've acquired an infinite amount of precision from somewhere.

I don't think it claims that.  If you can convert the rational back to
the exact same float, no data is lost.

> Again, this conversion is something that some, probably most, users
> would appreciate, but others will not.

It sounds like something should break by doing it, and that there are
good reasons why nobody else seem to be doing it, but I can't think of
any.

> As Jeff points out, Common Lisp does not convert from float to
> rational without user intervention.  I think the Common Lisp
> arithmetic rules have been incredibly carefully thought out; we
> should follow them where possible.  Not to mention that conforming
> to a bad standard is better than inventing a good non-standard.

I'm not sure I agree on your last point, but I agree that following CL
is a good idea, especially where we're less than sure about what we're
doing.  Still, rational contagion instead of fp contagion seems worth
considering.