A function to back fill NAs in a vector with the last non-NA value

backFillNA.r

# A function to back fill NAs in a vector with the last non-NA value
# https://gist.github.com/garyfeng/27e7f8e406192a8cb33a

backFillNA<- function (x) {
  nas<- which(is.na(x))
  # trick from http://stackoverflow.com/questions/24837401/find-consecutive-values-in-vector-in-r
  naList<-split(nas, cumsum(c(1, diff(nas) != 1)))
  # get the backfill values
  valueList<-lapply(naList, function(nalist) {
    prev<- nalist[1]-1
    #print(nalist)
    if (prev<=0) prev <-1
    x[nalist]<-x[prev]
    return(x[nalist])
  })
  # now back fill
  # can't use unlist(), as it converts everything to a common class.
  # see http://stackoverflow.com/questions/13859905/returning-a-vector-of-class-posixct-with-vapply
  #x[unlist(naList)] <- unlist(valueList)
  x[unlist(naList)] <- do.call(c, valueList)
  return (x)
}

x<-c("A","B",NA,NA,"C",NA,NA,NA,NA,"D",NA,NA);
#> x
# [1] "A" "B" NA  NA  "C" NA  NA  NA  NA  "D" NA  NA 
#> backFillNA(x)
# [1] "A" "B" "B" "B" "C" "C" "C" "C" "C" "D" "D" "D" 

########################
# The following recursive solution is taken from:
# https://stat.ethz.ch/pipermail/r-help/2008-July/169199.html
# it uses Recall(), R's method for recursive calling, to do back filling
# It works well for short vectors, but a test on a vector of 20k elements with many NAs
# and the recursion failed after a long while.

rna <- function(z) { 
	y <- c(NA, head(z, -1))
	z <- ifelse(is.na(z), y, z)
	if (any(is.na(z))) Recall(z) else z 
}

#> x
# [1] "A" "B" NA  NA  "C" NA  NA  NA  NA  "D" NA  NA 
#> rna(x)
# [1] "A" "B" "B" "B" "C" "C" "C" "C" "C" "D" "D" "D" 

hmm... the performance of the non-recursive version may be a function of how many NAs there are -- more precisely, how many NA runs there are in the data. We need to test this.

c.f. the Pandas implementation of forward and backward fillNA at http://pandas.pydata.org/pandas-docs/stable/basics.html#basics-reindexing and http://pandas.pydata.org/pandas-docs/stable/missing_data.html#cleaning-filling-missing-data