trypar.R

library(SpatialPosition)

load("../LargeSpatialPostition/com.RData")


# création d'une grille fine (5km)
grid <- CreateGrid(w = com, resolution = 5000)

# Les calculs de matrices de distances sont (beaucoup) plus rapides quand les 
# objets spatiaux sont en WGS84 :
# Stocker la projection initiale
old.proj <- proj4string(com)
# passage en WGS84
com <- spTransform(com, "+init=epsg:4326")
grid <- spTransform(grid, "+init=epsg:4326") 

knownpts = com
unknownpts = grid
varname = "POPULATION"
typefct = "exponential" 
span = 20000
beta = 2

library(foreach)
library(doParallel)


#setup parallel backend to use 8 processors
cl <- makeCluster(3)
registerDoParallel(cl)

# sequence pour découper la grille
sequence <- unique(c(seq(1,nrow(grid), 500),nrow(grid)) )

# nombre d'itération 
lseq <- length(sequence)-1
strt0 <- Sys.time()
ls <- foreach(i = 1:lseq, .packages = c('SpatialPosition')) %dopar% {
  strt1 <- Sys.time()
  print(strt1 - strt0)
  mat <- CreateDistMatrix(knownpts = knownpts, 
                          unknownpts = unknownpts[sequence[i]:sequence[i+1],], 
                          bypassctrl = TRUE)
  st <- stewart(knownpts = knownpts, 
                unknownpts = unknownpts[sequence[i]:sequence[i+1],],
                matdist = mat,
                typefct = "exponential", 
                span = 20000, beta = 3, 
                var = "POPULATION")
}
strt1 <- Sys.time()
print(strt1 - strt0)
stopCluster(cl)