On Wed, 2007-04-18 at 15:49 -0400, Stavros Macrakis wrote:
> plot2d( '(0,1+g)-g,[g,0,200],[nticks,1000]);
>
> The '(0,...) construction blocks simplification.
>
> Not by coincidence, this happens in the intervals [2^n-1,2^n]. With a
> different base, it would happen at [base^n-1,base^n]. Of course, with
> exact arithmetic, it doesn't happen.
>
> The only complaint I have about Maxima here is that its plotting
> routines aren't clever enough to realize that a seemingly isolated
> blip should be investigated further. But perhaps it is noticing that
> the blip is so tiny that it is probably a calculation error (which it
> is...).
Once again, there is no reason for that complaint in the most recent
version of plot2d. I had introduced the 10 significant digits for
another reason, before this thread on binary floating point, but it
happened to work for this case too :)
Jaime