Re: [Groff] Floating point arithmetic in groff

[Miklos Somogyi]

On 01/02/2006, at 3:31 PM, Werner LEMBERG wrote:

> > When implementing a feature requires less work than a few
> > workarounds, isn't it time to implement the feature :-) ?
>
> Hehe.  Have a look at this ooooold patch which I've received six years
> ago from Abramo.
>
> ......
>
> If you find it useful, and someone invests time to update the code to
> the current CVS (*together* with updates to the various documentation
> files -- this takes much more time than applying the patch itself),
> I'll add it.
>
>
>     Werner

Being illiterate in C and its ilk I can only guess that this patch is 
about the evaluation of real expressions
wherever they may lurk in a groff document.

If so, I'm glad to see environment variables in it. I would add "atan2" 
to the list of functions, more useful than
atan.

Error message on div by 0 is fine, but overflow may also occur by 
dividing with some small
number as well. If I remember well, IEEE's NaN is where all the 
exponent bits are 1 and the fraction
bits are not all 0. This could be (should be??) checked and an error 
message given after the deed.

All this stuff brought back memories when I wrote a "fast" float 
package for the PDP-8, pre-float_hardware
days. Jeez, we were young then.

Miklos