Proper way to rbind SpatialPolygonsDataFrames with identical polygon IDs?

Question

What's the proper R idiom for rbinding SPDFs together when the IDs overlap? Note that here (as is often the case) the IDs are basically meaningless so it's pretty annoying that I can't just make rbind ignore them....

library(sp)
library(UScensus2000)
library(UScensus2000tract)

data(state) # for state names
states <- gsub( " ", "_", tolower(state.name) )
datanames <- paste(states,"tract", sep=".")
data( list=datanames )
lst <- lapply(datanames,get)

nation <- do.call( rbind, lst )
Error in validObject(res) : 
  invalid class “SpatialPolygons” object: non-unique Polygons ID slot values

# This non-exported function designed to solve this doesn't seem to work any more.
d <- sp:::makeUniqueIDs( list(arizona.tract,delaware.tract) )
Error in slot(i, "ID") : 
  no slot of name "ID" for this object of class "SpatialPolygonsDataFrame"

Answer 1

This is an even simpler approach:

x <- rbind(x1, x2, x3, makeUniqueIDs = TRUE)  

Answer 2

ID's, slots, and apply-type functions. My top three least favorite things that are absolutely essential to everything I do. I thought I would respond just to generate more content on this topic.

The code below works, but it retains the "useless" ID values. Better code would take the time to parse things so that every tract had the state FIPS, county FIPS, and tract FIPS as its ID. Just a few more lines to make that happen, but since you don't care about ID's we will leave it out for now.

#Your Original Code
library(sp)
library(UScensus2000)
library(UScensus2000tract)

data(state) # for state names
states <- gsub( " ", "_", tolower(state.name) )
datanames <- paste(states,"tract", sep=".")
data( list=datanames )
lst <- lapply(datanames,get)

#All good up to here, but we need to create unique ID's before rbind

#Modified from Roger Bivand's response at:
# https://stat.ethz.ch/pipermail/r-sig-geo/2007-October/002701.html

#For posterity: We can access the ID in two ways:
class(alaska.tract)
getSlots(class(alaska.tract))
class(slot(alaska.tract, "polygons")[[1]])
getSlots(class(slot(alaska.tract, "polygons")[[1]]))

#So to get all ID's
sapply(slot(alaska.tract, "polygons"), function(x) slot(x, "ID"))
#or
rownames(as(alaska.tract, "data.frame"))
#These should be the same, but they are quite different...sigh. Doesn't matter for
#what follows though

#To make them uniform we can write a function using the spChFIDs function from sp:
makeUniform<-function(SPDF){
  pref<-substitute(SPDF)  #just putting the file name in front.
  newSPDF<-spChFIDs(SPDF,as.character(paste(pref,rownames(as(SPDF,"data.frame")),sep="_")))
  return(newSPDF)
}

#now to do this for all of our state files
newIDs<-lapply(lst,function(x) makeUniform(x))

#back to your code...
nation <- do.call( rbind, newIDs )

Answer 3

Alright, here's my solution. Suggestions welcome. I'll likely submit this as a patch to sp unless anyone sees any glaring omissions.

#' Get sp feature IDs
#' @aliases IDs IDs.default IDs.SpatialPolygonsDataFrame
#' @param x The object to get the IDs from
#' @param \dots Pass-alongs
#' @rdname IDs
IDs <- function(x,...) {
  UseMethod("IDs",x)
}
#' @method IDs default
#' @S3method IDs default
#' @rdname IDs
IDs.default <- function(x,...) {
  stop("Currently only SpatialPolygonsDataFrames are supported.")
}
#' @method IDs SpatialPolygonsDataFrame
#' @S3method IDs SpatialPolygonsDataFrame
#' @rdname IDs
IDs.SpatialPolygonsDataFrame <- function(x,...) {
  vapply(slot(x, "polygons"), function(x) slot(x, "ID"), "")
}

#' Assign sp feature IDs
#' @aliases IDs<- IDs.default<-
#' @param x The object to assign to
#' @param value The character vector to assign to the IDs
#' @rdname IDs<-
"IDs<-" <- function( x, value ) {
  UseMethod("IDs<-",x)
}
#' @method IDs<- SpatialPolygonsDataFrame
#' @S3method IDs<- SpatialPolygonsDataFrame
#' @rdname IDs<-
"IDs<-.SpatialPolygonsDataFrame" <- function( x, value) {
  spChFIDs(x,value)
}

#' rbind SpatialPolygonsDataFrames together, fixing IDs if duplicated
#' @param \dots SpatialPolygonsDataFrame(s) to rbind together
#' @param fix.duplicated.IDs Whether to de-duplicate polygon IDs or not
#' @return SpatialPolygonsDataFrame
#' @author Ari B. Friedman, with key functionality by csfowler on StackExchange
#' @method rbind.SpatialPolygonsDataFrame
#' @export rbind.SpatialPolygonsDataFrame
rbind.SpatialPolygonsDataFrame <- function(..., fix.duplicated.IDs=TRUE) {
  dots <- as.list(substitute(list(...)))[-1L]
  dots_names <- as.character(dots) # store names of objects passed in to ... so that we can use them to create unique IDs later on
  dots <- lapply(dots,eval)
  names(dots) <- NULL
  # Check IDs for duplicates and fix if indicated
  IDs_list <- lapply(dots,IDs)
  dups.sel <- duplicated(unlist(IDs_list))
  if( any(dups.sel) ) {
    if(fix.duplicated.IDs) {
      dups <- unique(unlist(IDs_list)[dups.sel])
      # Function that takes a SPDF, a string to prepend to the badID, and a character vector of bad IDs
      fixIDs <- function( x, prefix, badIDs ) {
        sel <-  IDs(x) %in% badIDs
        IDs(x)[sel] <- paste( prefix, IDs(x)[sel], sep="." )
        x
      }
      dots <- mapply(FUN=fixIDs , dots, dots_names, MoreArgs=list(badIDs=dups) )
    } else {
      stop("There are duplicated IDs, and fix.duplicated.IDs is not TRUE.")
    }
  }
  # One call to bind them all
  pl = do.call("rbind", lapply(dots, function(x) as(x, "SpatialPolygons")))
  df = do.call("rbind", lapply(dots, function(x) x@data))
  SpatialPolygonsDataFrame(pl, df)
}

Answer 4

I've appreciated the detail of other answers here and, building on them, the one-liner I've come to is below. Like OP, I don't care much about the meaning of the ID, but the following could also be adapted to embed a more informative ID too.

lst <- lapply(1:length(lst), function(i) spChFIDs(lst[[i]], paste0(as.character(i), '.', 1:length(lst[[i]]))))