Replace penultimate non-empty row tail with NA in R

Question

I have a data frame that looks like this:

 df <- read.table(tc <- textConnection("
 var1    var2    var3    var4
 1       2       7      NA
 4       4       NA      6
 2       NA      3       NA                
 4       4       4       4              
 1       3      NA      NA"), header = TRUE); close(tc)

I want to replace the penultimate non-empty row tails with NAs to give me a new data frame that looks like this:

 df <- read.table(tc <- textConnection("
 var1    var2    var3    var4
 1       NA       7      NA
 4       NA      NA      6
 NA       NA      3       NA                
 4       4       NA       4              
 NA       3      NA      NA"), header = TRUE); close(tc)

I'm able to identify the location of the penultimate non-empty row tails by using the statement:

 penultimate <- function(x) head(tail(x[!is.na(x)], 2),1)

My problem is how to now change these to NAs.

Answer 1

We can use apply

penultimate <- function(x) replace(x, head(tail(which(!is.na(x)), 2),1), NA)
t(apply(df, 1, penultimate))
#      var1 var2 var3 var4
#[1,]    1   NA    7   NA
#[2,]    4   NA   NA    6
#[3,]   NA   NA    3   NA
#[4,]    4    4   NA    4
#[5,]   NA    3   NA   NA

Or we can use which with arr.ind=TRUE

i1 <- which(!is.na(df), arr.ind=TRUE)
df[i1[!!ave(i1[,2], i1[,1], FUN = function(x) rank(-x)==2),]] <- NA

---

If we need to use dplyr , the rowwise with do can get the expected output

library(dplyr)
df %>% 
   rowwise() %>%
   do(data.frame(t(penultimate(unlist(.)))))
#  var1  var2  var3  var4
#  <int> <int> <int> <int>
#1     1    NA     7    NA
#2     4    NA    NA     6
#3    NA    NA     3    NA
#4     4     4    NA     4
#5    NA     3    NA    NA

---

Or using data.table

library(data.table)
setDT(df)[, as.list(penultimate(unlist(.SD))), .(rn= 1:nrow(df))][, rn := NULL][]