识别R中的网格中的最近邻居（空间）

Question

I would like to create a square grid and identify the grid cells that border a set of other grid cell for which a binary variable takes a 1. In the following example, I would like to generate a vector of cell ids that border id g13 and g24:

require(sp)         
grid <- GridTopology(c(0,0), c(1,1), c(5,5))    
polys <- as(grid, "SpatialPolygons")
centroids <- coordinates(polys)
id <- names(polys)
tr <- ifelse(id == "g13" | id == "g24", 1, 0)       
ex <- SpatialPolygonsDataFrame(polys, data = data.frame(id = id, tr = tr, row.names = row.names(polys)))

plot(ex)
text(coordinates(polys), labels = row.names(polys))

Such that it outputs a vector for all matching g13 as (g7, g8, g9, g12, g14, g17, g18, g19) and one matching g24 as (g18, g19, g20, g23, g24, g25). Any and all thoughts greatly appreciated.

Answer 1

rgeos::gTouches is perfect for this:

library(rgeos)
adj <- gTouches(polys, polys[which(ex$tr==1)], byid=TRUE)
apply(adj, 1, which)

# $g13
#  g7  g8  g9 g12 g14 g17 g18 g19 
#   7   8   9  12  14  17  18  19 
# 
# $g24
# g18 g19 g20 g23 g25 
#  18  19  20  23  25 

And, because everyone loves pictures:

plot(ex, col=ifelse(seq_along(ex) %in% c(unlist(adj), which(ex$tr==1)), 'gray', NA))
text(coordinates(polys), labels=row.names(polys))