# read the RDS above, to save as .csv file.

source('AutoSHUD/Rfunction/LDAS_UnitConvert.R')
write.tsd.custom <- function (x, file, append = F, quite = F, header = NULL) 
{
  mat = as.matrix(rbind(x))
  nr = nrow(x)
  nc = ncol(x)
  if (!quite) {
    message("Writing ", file)
  }
  tt = stats::time(x)
  tday = as.numeric(difftime(tt, tt[1], units = "days"))
  if (is.null(header)) {
    t0 = format(time(x)[1], "%Y%m%d")
    header = c(nr, nc + 1, t0)
  }
  dd = data.frame(Time_Day = tday, mat)
  write(header, file = file, ncolumns = length(header), append = append, 
        sep = "\t")
  write(colnames(dd), file = file, ncolumns = nc + 1, append = T, 
        sep = "\t")
  write(t(dd), file = file, ncolumns = nc + 1, append = T, 
        sep = "\t")
}

fun.GLDAS.RDS2csv <- function(xfg){
  fns = file.path(xfg$dir$predata, paste0(xfg$prjname,'-', xfg$years, '.RDS'))
  cns = c('Rainf_f_tavg', 'Tair_f_inst','Qair_f_inst',
          'Wind_f_inst', 'Swnet_tavg','Lwnet_tavg', 
          'Psurf_f_inst')
  forcnames = c( "Prcp_mm.d", "Temp_C", "RH_%", "Wind_m.s", "RN_w.m2" )
  
  nf=length(fns)
  for(i in 1:nf){
    x=readRDS(fns[i])  
    message(i,'/', nf, '\t', basename(fns[i]))
    y=x[,cns,]
    if(i==1){
      dat = y
    }else{
      dat=abind::abind(dat, y, along=3)
    }
  }
  dn = dimnames(dat)
  nd = dim(dat)
  xl = list()
  
  # TODDO: modify the time to extract correct time
  time.tag = dimnames(dat)[[3]]
  idx = which(!grepl(' ', time.tag))
  time.tag[idx] = paste0(time.tag[idx], '00:00:00')
  time = lubridate::ymd_hms(time.tag)
  # time = as.POSIXct(dimnames(dat)[[3]], format= "%Y%m%d%H")
  diff_seconds = as.numeric(difftime(time[[2]], time[[1]], units="hours")) * 3600
  diff_seconds

  for(i in 1:nd[1]){
    message(i,'/', nd[1], '\t', dn[[1]][i] )
    x = t( dat[i,,] )
    y=unitConvert.GLDAS(x, diff_seconds)
    xl[[i]]=as.xts(y, order.by=time)
  }
  nx=length(xl)
  sitename = dn[[1]]
  sitename
  fns=paste0(sitename, '.csv')
  xmean = NULL
  for(i in 1:nx){
    fn=fns[i]
    # write this in correct format
    # write.tsd.custom(xl[[i]], file.path(xfg$dir$forc, fn))
    write.tsd(xl[[i]], file.path(xfg$dir$forc, fn))
    if(all(is.na(xl[[i]]))){
    }else{
      if(is.null(xmean)){
        xmean = xl[[i]]
      }else{
        xmean = xmean + xl[[i]]
      }
    }
  }
  xmean = xmean / nx
  xmean[is.na(xmean)] <- -999
  if(all(is.na(xmean))){}else{
    png(file = file.path(xfg$dir$fig, paste0(xfg$prefix, '_GLDAS_TS.png')), height=7, width=11, res=200, unit='in')
    plot.zoo(xmean, main='GLDAS')
    dev.off()
  }
}