Transforming values in R data frame using three conditions within groups

Question

I have the following data frame:

 Group   Number    Letter
    G1        1         A
    G1        NA        B
    G1        NA        X
    G1        NA        D
    G2        1         A
    G2        NA        B
    G2        NA        C
    G2        NA        X
    G3        1         A
    G3        2         B
    G3        3         C
    G3        NA        X

I want to transform this dataframe within each group using the following conditions:

1. If the letter in the Letter column equals “X”, then all preceding rows with NA in the Number column have their letter changed to “U”.
2. If the preceeding rows have a number in the Number column instead of NA, they stay the same.
3. Rows with X in the Letter column do not change.

This should yield a data frame as follows:

 Group   Number    Letter
    G1        1         A
    G1        NA        U
    G1        NA        X
    G1        NA        D
    G2        1         A
    G2        NA        U
    G2        NA        U
    G2        NA        X
    G3        1         A
    G3        2         B
    G3        3         C
    G3        NA        X

How can I do this in R, ideally using dplyr?

Answer 1

data.table solution. Now used lapply but there is probably a way to vectorize using the by argument of data.table

library(data.table)

#read data
dt <- structure(list(group = c("G1", "G1", "G1", "G1", "G2", "G2", 
"G2", "G2", "G3", "G3", "G3", "G3"), number = c(1, NA, NA, NA, 
1, NA, NA, NA, 1, 2, 3, NA), Letter = c("A", "U", "X", "D", "A", 
"U", "U", "X", "A", "B", "C", "X")), row.names = c(NA, -12L), class = "data.frame", index = integer(0)) |> as.data.table()

#add rows
dt[, row := 1:nrow(dt)]

#find rows with last observations of x
maxrows <- dt[Letter == "X", max(row), by = group]

#change values
lapply(maxrows$group, FUN = function(x){
  dt[group == x & row < maxrows[group == x]$V1 & Letter != "X" & is.na(number), Letter := "U"]
})

#remove row
dt[, row := NULL]