Best way to plot cross sections of discrete-valued grids

Ah, while typing this email I see Barry's contribution --
Baz, it's time that you take a closer look at what's already
in sp; we're duplicating work.

Here's my reply:

Scott, you can. Consider the following example with
random numbers (1:10) on a 3d 100x100x100 grid.

library(sp)
# create 3D grid
xyz = expand.grid(x = 1:100, y = 1:100, z = 1:100)
d = data.frame(xyz, v = sample(1:10, 1e6, replace = T))
gridded(d) = ~x+y+z
class(d)
summary(d)
# first point on line:
p1 = c(5,5,5)
# second point on line:
p2 = c(95,95,64)
rbind(p1,p2)
pts = sample.Line(rbind(p1,p2), 1000, "regular")
# select grid elements for each of 1000 points:
plot(d$v[overlay(d, pts)], type = 'l')


It "misuses" sample.Line, which was written as a spsample method
for Line objects; too bad that Line objects in sp are limited to exist
in two dimensions. Anyway, this seems to work (although I
obviously didn't just verify that the result was correct!!)

I have a 3-D grid where the values of the cells are a small number of
discrete values.  (The grid represents subsurface soil and geology
types.)  Contiguous regions of the same value are relatively few in
number and large in size in comparison to the whole domain.  I am
looking for advice on two things:

1)  What is the best way to sample the grid along any transect defined
by a line between (x1, y1) and (x2, y2), and plot the 2-D vertical cross
section?  I would like to sample not only in the orthogonal directions
defined by the grid but along any other compass bearing as well.
  

See above.

Lattice plotting capability is desired.

library(lattice)
xyplot(z~1:1000, data.frame(z = d$v[overlay(d, pts)]), type = 'l')

2)  Is it better to plot the same-valued regions as polygons instead of
discrete cells?  EPS plots using levelplot() tend to show the grid as
faint lines, and have large file sizes because data on each cell is
retained instead of data for single-valued regions that tend to be
relatively large.
  

Am I correct that the lines you mention are shown in ghostview/ghostscript?
If that is the case, they result from the gs/gv anti-aliasing 
mechanism--they are
not part of the postscript. Printing to paper will not show it.

If file size is a problem, try pdf and wrap the pdf e.g. using pdflatex, 
or directly
print to bitmap formats such as png.

HTH,
--
Edzer

Thanks for any help,

Scott Waichler, Senior Research Scientist
Pacific Northwest National Laboratory
scott.waichler _at_ pnl.gov
http://hydrology.pnl.gov

_______________________________________________
R-sig-Geo mailing list
R-sig-Geo at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-geo