Fill first (only one) NA with next non-NA value by group using dplyr/tidyr

I hava a problem, where I need to fill NA by group, but only one (the first) NA before some non-NA value (direction = "up"). I know the tidyr fill() function but do not know how to use it conditionally (so to fill only one NA above some non-NA value).

df <- data.frame(g=c(1,1,1,1,1,1, 2,2,2,2,2,2,2),v=c(NA,NA,5,NA,8,NA, 1,1,NA,2,NA,NA,3))

Data (with group "g" and values "v"):

   g  v
1  1 NA
2  1 NA
3  1  5
4  1 NA
5  1  8
6  1 NA
7  2  1
8  2  1
9  2 NA
10 2  2
11 2 NA
12 2 NA
13 2  3

should become…

   g  v
1  1 NA
2  1  5
3  1  5
4  1  8
5  1  8
6  1 NA
7  2  1
8  2  1
9  2  2
10 2  2
11 2 NA
12 2  3
13 2  3

>Solution :

You can always create your own solution:

fill_one_up <- \(x) {
  n <- length(x)
  if (n <= 1L) return(x)
  for (i in 2L:n) {
    if (!is.na(x[i]) && is.na(x[i-1L])) {
      x[i-1L] <- x[i]
    }
  }
  return(x)
}

df <- df |> mutate(v2 = fill_one_up(v), .by=g)

#    g  v v2
# 1  1 NA NA
# 2  1 NA  5
# 3  1  5  5
# 4  1 NA  8
# 5  1  8  8
# 6  1 NA NA
# 7  2  1  1
# 8  2  1  1
# 9  2 NA  2
# 10 2  2  2
# 11 2 NA NA
# 12 2 NA  3
# 13 2  3  3

If you need more speed you can translate the logic to Rcpp:

Rcpp::cppFunction("NumericVector fill_one_up_cpp(NumericVector x) {
  int n = x.size();
  if (n <= 1) return x;
  for (int i = 1; i < n; i++) {
    if (!NumericVector::is_na(x[i]) && NumericVector::is_na(x[i-1])) {
      x[i-1] = x[i];
    }
  }
  return x;
}")