Hi Ugo,
Yes. Surface rendering is based on filling in the quadrangles
(Gridded) or triangles (Irregular) of the Field's topology.
When the display logic detects that any points have been
removed, either because of missing values or SelectRange, then
the topology is invalid so it falls back on point rendering,
which does not require topology.
Currently, the only work-around is for the application to
reconstruct a new topology, either using the IrregularSet
constructors (slow for lots of points) or explicitly using
DelaunayCustom if the application has some information about
the new topology.
Note that iso-surfaces and iso-lines work a bit differently,
and will evaluate the validity of each topology cell. This
can be misleading, since there may be many valid points but
no valid cells (i.e., each cell may include at least one
invalid point). In this sense falling back on point rednering
is good, since it shows precisely which samples are valid.
We could add an option for surface rendering [i.e., for Fields
of the form ((x, y) -> z)] to render those cells that have
all points valid [similar to the way iso-surface works now].
I've been veryy busy lately but when I get a breather I will
try to do this.
Cheers,
Bill
p.s., NOTE that my email address has changed to
hibbard@xxxxxxxxxxxxxxxxx
The old email server is a VAX with fatal Y2K flaws.
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
hibbard@xxxxxxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html
