Construct neighbours list from polygon list
The function builds a neighbours list based on regions with contiguous boundaries, that is sharing one or more boundary point. The current function is in part interpreted and may run slowly for many regions or detailed boundaries, but from 0.2-16 should not fail because of lack of memory when single polygons are built of very many border coordinates.
poly2nb(pl, row.names = NULL, snap=sqrt(.Machine$double.eps), queen=TRUE, useC=TRUE, foundInBox=NULL)
pl |
list of polygons of class extending |
row.names |
character vector of region ids to be added to the neighbours list as attribute |
snap |
boundary points less than |
queen |
if TRUE, a single shared boundary point meets the contiguity condition, if FALSE, more than one shared point is required; note that more than one shared boundary point does not necessarily mean a shared boundary line |
useC |
default TRUE, doing the work loop in C, may be set to false to revert to R code calling two C functions in an |
foundInBox |
default NULL using R code, possibly parallelised if a snow cluster is available, otherwise a list of length |
A neighbours list with class nb. See card for details of “nb” objects.
From 0.5-8, the function includes faster bounding box indexing and other improvements contributed by Micah Altman. If a cluster is provided using set.ClusterOption, it will be used for finding candidate bounding box overlaps for exact testing for contiguity.
Until 1.1-7, sf polygons included both start and end points, so could erroneously report rook neighbourhood where only queen was present, see https://github.com/r-spatial/spdep/issues/50.
From 1.1-9 with sf 1.0-0, s2 is used in bounding box indexing internally when pl is in geographical coordinates. Because the topology engine of s2 differs from the use of GEOS for planar coordinates by sf, some output differences may be expected. When s2 is used, (use_s2=TRUE and dfMaxLength > 0) segmenting is done to add points to bounding box edges, making them more likely to intersect on the sphere, but currently increasing timing and memory use substantially. If segmenting not is done, but use_s2=TRUE, candidate neighbours may be missed and others exchanged. Since treating spherical geometries as planar is also questionable, it is not clear whether spherical contiguous polygon neighbours should simply follow neighbours found by treating the geometries as planar https://github.com/r-spatial/s2/issues/125#issuecomment-864403372. However, current advice is not to use s2 for finding contiguity neighbours.
Roger Bivand Roger.Bivand@nhh.no with contributions from Micah Altman
columbus <- st_read(system.file("shapes/columbus.shp", package="spData")[1], quiet=TRUE)
col.gal.nb <- read.gal(system.file("weights/columbus.gal", package="spData")[1])
coords <- st_coordinates(st_centroid(st_geometry(columbus)))
xx <- poly2nb(as(columbus, "Spatial"))
dxx <- diffnb(xx, col.gal.nb)
plot(st_geometry(columbus), border="grey")
plot(col.gal.nb, coords, add=TRUE)
plot(dxx, coords, add=TRUE, col="red")
title(main=paste("Differences (red) in Columbus GAL weights (black)",
"and polygon generated queen weights", sep="\n"), cex.main=0.6)
# poly2nb with sf sfc_MULTIPOLYGON objects
sf_xx <- poly2nb(columbus)
diffnb(sf_xx, xx)
sfc_xx <- poly2nb(st_geometry(columbus))
diffnb(sfc_xx, xx)
xxx <- poly2nb(as(columbus, "Spatial"), queen=FALSE)
dxxx <- diffnb(xxx, col.gal.nb)
plot(st_geometry(columbus), border = "grey")
plot(col.gal.nb, coords, add = TRUE)
plot(dxxx, coords, add = TRUE, col = "red")
title(main=paste("Differences (red) in Columbus GAL weights (black)",
"and polygon generated rook weights", sep="\n"), cex.main=0.6)
cards <- card(xx)
maxconts <- which(cards == max(cards))
if(length(maxconts) > 1) maxconts <- maxconts[1]
fg <- rep("grey", length(cards))
fg[maxconts] <- "red"
fg[xx[[maxconts]]] <- "green"
plot(st_geometry(columbus), col=fg)
title(main="Region with largest number of contiguities", cex.main=0.6)
nc.sids <- st_read(system.file("shapes/sids.shp", package="spData")[1], quiet=TRUE)
system.time(xxnb <- poly2nb(nc.sids))
system.time(xxnb <- poly2nb(as(nc.sids, "Spatial")))
plot(st_geometry(nc.sids))
plot(xxnb, coordinates(as(nc.sids, "Spatial")), add=TRUE, col="blue")
sq <- st_polygon(list(rbind(c(0,0), c(1,0), c(1,1), c(0,1), c(0,0))))
sq2 <- sq + c(0,1)
sq3 <- sq + c(1,0)
sq4 <- sq + c(1,1)
gm <- st_sfc(list(sq, sq2, sq3, sq4))
df <- st_as_sf(gm, id=1:4)
plot(st_geometry(df))
text(st_coordinates(st_centroid(gm)), as.character(df$id))
unclass(poly2nb(df, queen = FALSE))Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.