On Wed, 2007-04-04 at 15:59 -0400, Stavros Macrakis wrote:
> I would guess this is in gnuplot. Try
>
> plot2d([discrete,[[0,1],[1,2],[2,1e308],[3,4],[4,5]]]);
> or
> plot2d([discrete,[[0,1],[1,2],[1e308,2],[3,4],[4,5]]]);
>
> There is no line connecting [1,2] and [3,4], and the x/y-limits aren't
> affected by the 1e308. On the other hand, with 1e300, the x/y limits
> are huge and the line *is* drawn.
I think both gnuplot and plot2d are being silly in this case.
1e308 is written in maxout.gnuplot as a 1 followed by 308 zeros :)
That huge line is simply ignored by gnuplot, leaving a blank line
in between the data, which means a gap in the plot.
I will fix this very easily by replacing the Lisp format
currently used to print 1e308 ("~g") by something more reasonable,
such as ",10g"
So far, it looks as if Gnuplot is acting reasonably.
However, once the 1 with 308 zeros is replaced by 1.0E308, gnuplot
will ignore that line again and show a gap once again. If plot2d did not
send all the points to maxout.gnuplot, but only those within a
range, at least one could go around that bug by setting a y_range.
Regards,
Jaime