The new test compares both X and Y coordinate. The illegal division by
zero error was caused by coordinate 14.47700000 which was truncated to
14.477 for one point and remained 14.47700000 for another point.
I agree that "abs(a-b) < EPSILON" is correct way to compare floats but
it seems simple a == b works for this case.
On 6/15/08, Martijn van Oosterhout <[EMAIL PROTECTED]> wrote:
> On Wed, Jun 11, 2008 at 8:14 PM, Dirk-Lüder Kreie <[EMAIL PROTECTED]> wrote:
> > Jiri Klement schrieb:
> >>
> >> I've run into the same error today. Following patch makes it work for me:
> >>
> >> Index: lines2curves.pl
> >> ===================================================================
> >> --- lines2curves.pl (revision 8167)
> >> +++ lines2curves.pl (working copy)
> >> @@ -147,8 +147,7 @@
> >> shift @$points_ref;
> >>
> >> foreach my $point_ref (@$points_ref) {
> >> - if ($point_ref->[0].$point_ref->[1] eq
> >> - $clean_points_ref->[-1][0].$clean_points_ref->[-1][1]) {
> >> + if (($point_ref->[0] == $clean_points_ref->[-1][0]) &&
> >> ($point_ref->[1] == $clean_points_ref->[-1][1])) {
> >> next;
> >> }
> >>
> >> I don't understand why numeric values were compared using eq operator.
> >> There are other places in lines2curves.pl comparing points this way.
> >
> > I don't understand either why this was converted into a string comparison,
> > but if this really fixes the issue (i.e. by making a comparison work that
> > wouldn't work when used as string comparison) I'm all for it.
>
>
> I suppose the string concatination is because it wants to compare the
> X and Y coordinates simultaneously whereas the new test only compares
> the X coordinates. My guess is the original problem was caused by two
> point close enough to cause a division by zero but far enugh (in the
> order the machine precision) to make the test fail.
>
> If there are more places comparing like this I'd suggest making a
> function ComparePoints like so:
>
> use constant EPSILON => 1e-6;
> sub ComparePoints
> {
> my($a,$b)[EMAIL PROTECTED];
> if( abs($a->[0] - $b->[0]) < EPSILON and
> abs($a->[1] - $b->[1]) < EPSILON )
> { return 1; }
> return 0;
>
> }
>
> Have a nice day,
> --
> Martijn van Oosterhout <[EMAIL PROTECTED]> http://svana.org/kleptog/
>
_______________________________________________
Tilesathome mailing list
[email protected]
http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/tilesathome
