library(igraph)
setwd("~/Documents/Cascade Project/")
load('Results/hyper-lcc.RData')

vic_ids = which(V(hyp_lcc)$vic==TRUE)

edgeWeights = function(eis){return(c(hyp_lcc_edges$weight[eis],Inf,Inf)[1:3])}

dag_dat_all = data.frame(matrix(nrow=1,ncol=10))
hyp_lcc2 = remove.edge.attribute(hyp_lcc,'weight')
ei = 1
ptm=proc.time()
for (u in vic_ids){
  if ((which(vic_ids==u) %% 1000)==0)  print(which(vic_ids==u))
  tu    = hyp_lcc_verts$vic.day[u]
  u_spawn = hyp_lcc_verts$spawn.date[u]
  nbhd  = unlist(neighborhood(hyp_lcc,nodes=u,order=3)) # get nodes within neighborhood
  nbhd  = nbhd[-1] # don't want to include u in the neighborhood
  tvs   = hyp_lcc_verts$vic.day[nbhd]
  v_spawn = hyp_lcc_verts$spawn.date[nbhd]
  nbhd  = nbhd[tu>v_spawn & (is.na(tvs) | tu<tvs)]
  tvs   = hyp_lcc_verts$vic.day[nbhd]
  dists = as.numeric(shortest.paths(hyp_lcc2,u,nbhd))
  
  es = get.shortest.paths(hyp_lcc2,u,nbhd,output='epath')$epath
  weights = matrix(unlist(lapply(es,edgeWeights),use.names = F),ncol=3,byrow=T)
  
  #will be faster to pre-allocate and fill in rather than rbind each time
  dag_dat_all[ei:(ei+length(nbhd)-1),] = data.frame(rep(u,length(nbhd)), nbhd, 
                                                    rep(tu,length(nbhd)), tvs, dists, 
                                                    weights, u_spawn, v_spawn, row.names=NULL)
  ei = ei + length(nbhd)
}
print(proc.time()-ptm) #3.5 hours
colnames(dag_dat_all) = c('from','to','t1','t2','dist','w1','w2','w3','spawn1','spawn2')
rownames(dag_dat_all) = NULL

# dag_dat_all$spawn1 = hyp_lcc_verts$spawn.date[dag_dat_all$from]
# dag_dat_all$spawn2 = hyp_lcc_verts$spawn.date[dag_dat_all$to]

save(dag_dat_all, file='Results/dag_dat_all.RData')
write.csv(dag_dat_all, file='Results/dag_dat_all.csv')

dag_dat_vics = dag_dat_all[!is.na(dag_dat_all$t2),]
save(dag_dat_vics, file='Results/dag_dat_vics.RData')
write.csv(dag_dat_vics, file='Results/dag_dat_vics.csv')