[R-sig-Geo] merging polygons and IDs

[Roger Bivand]

On Sun, 11 Jan 2015, David Depew wrote:

> Dear list,
>
> given a set of polygons created from points with a specified buffer
> distance,
>
> library(sp)
> library(rgeos)
> box <- readWKT("POLYGON((0 0, 0 1000, 1000 1000, 1000 0, 0 0))")
> plot(box)
> set.seed(1)
> pts <- spsample(box, n=40, type="random")
> pols <- gBuffer(pts, byid=TRUE, width=50)
> plot(pols, add=TRUE)
>
>
> I'd like to merge polygons where the buffers overlap
> i.e.
> merg.poly=gUnaryUnion(pols)
> plot(merg.poly)
>
> but I'm having difficulty figuring out how to specify that the merged
> polygons retain some unique Id. Clearly in the package help, one would
> specify this by use of the "id" argument, however since the polygons
> started as points, overlapping points do not have the same ID. Suppose one
> approach is to define a common ID for overlapping polys before using
> gUnaryUnion, but I'm at a loss where to begin with that approach. Any
> advice would be greatly appreciated,

Thanks for a self-contained example!

Right, this isn't obvious, as you need to go to a list/graph 
representation of the intersects predicate, which takes us outside rgeos. 
We'll compare this with the equivalent for distance neighbours in spdep:

library(spdep)
dnb <- dnearneigh(pts, 0, 100) # 2*50 buffer
nc <- n.comp.nb(dnb) # find graph components
merg.poly <- gUnaryUnion(pols, nc$comp.id)
plot(merg.poly)

which jumps over the buffered objects. To use the buffered objects, do:

gI <- gIntersects(pols, pols, byid=TRUE, returnDense=FALSE) # list
for (i in seq(along=gI)) {gI[[i]] <- gI[[i]][-(which(gI[[i]] == i))];
  if (length(gI[[i]]) == 0) gI[[i]] <- 0L} # remove i<->i edges
class(gI) <- "nb"
attr(gI, "region.id") <- names(gI)
nc1 <- n.comp.nb(gI)
all.equal(nc, nc1) # same result

So what this is doing is having gIntersects return a list. Since i 
intersects i, we have to drop the i<->i edges, pretend that this is an nb 
object, and use the n.comp.nb() function to find the graph components. 
These two approaches give the same results. There is a new vignette in 
spdep starting to look at representations of neighbours as graph objects:

http://cran.r-project.org/web/packages/spdep/vignettes/nb_igraph.html

Hope this helps,

Roger

>
> Thanks,
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 91 00
e-mail: Roger.Bivand at nhh.no