cleanly paste columns that contain NAs [duplicate]

Answer 1

Here's a way in base R -

dat$col3 <- apply(dat, 1, function(x) paste0(na.omit(x), collapse = "; "))

   col1   col2          col3
1 stuff things stuff; things
2 stuff   <NA>         stuff
3 stuff things stuff; things
4  <NA> things        things
5  <NA>   <NA>              

Answer 2

We can use tidyverse approach. Loop through the rows with pmap, remove the NA elements with na.omit and paste the rows together

library(tidyverse)
dat %>% 
  mutate_all(as.character) %>% 
  mutate(col3 = pmap_chr(., ~ c(...) %>% 
                             na.omit %>%
                             paste(collapse="; ")))
# col1   col2          col3
#1 stuff things stuff; things
#2 stuff   <NA>         stuff
#3 stuff things stuff; things
#4  <NA> things        things
#5  <NA>   <NA>   

Or another option is

dat %>%
   mutate_all(as.character) %>%
  mutate(col3 = case_when(is.na(col1)|is.na(col2) ~ 
       coalesce(col1, col2), 
       TRUE ~ str_c(pmin(col1, col2, na.rm = TRUE), 
             pmax(col1, col2, na.rm = TRUE), sep="; ")))
#   col1   col2          col3
#1 stuff things stuff; things
#2 stuff   <NA>         stuff
#3 stuff things stuff; things
#4  <NA> things        things
#5  <NA>   <NA>          <NA>

---

Or using a vectorized approach with base R

i1 <- !!rowSums(!is.na(dat))
dat$col3[i1] <- gsub("^NA;|; NA", "", do.call(paste, c(dat[i1,], sep="; ")))
dat
#   col1   col2          col3
#1 stuff things stuff; things
#2 stuff   <NA>         stuff
#3 stuff things stuff; things
#4  <NA> things        things
#5  <NA>   <NA>          <NA>