R: How to 'aggregate' (or combine) character columns?

Question

I have a df with three columns. Every column has a character or NA and every row has only one character. As this example:

df <- data.frame(a=c("NA","NA","NA","NA","fruits","fruits","fruits","fruits","fruits","fruits"), 
                 b=c("NA","NA","veggies","veggies","NA","NA","NA","NA","NA","NA"),
                 c=c("nuts","nuts","NA","NA","NA","NA","NA","NA","NA","NA") )

I want to combine all three columns, to get this:

1     nuts
2     nuts
3  veggies
4  veggies
5   fruits
6   fruits
7   fruits
8   fruits
9   fruits
10  fruits

Using numeric values I would use aggregate with na.rm=TRUE . However, I don't have any idea how to do this with characters. Ideas? Thanks

Answer 1

We can use max.col after converting the string "NA" to real NA . We get the row/column index with max.col , extract the values and then convert tot data.frame .

is.na(df) <- df=='NA'
data.frame(var=df[cbind(1:nrow(df),max.col(!is.na(df)))])
#      var
#1     nuts
#2     nuts
#3  veggies
#4  veggies
#5   fruits
#6   fruits
#7   fruits
#8   fruits
#9   fruits
#10  fruits

---

Or another option would be

data.frame(var= df[cbind(1:nrow(df),(+!is.na(df)) %*% seq_along(df))])

Answer 2

To polish the ideas provided in comments, you can do this:

data.frame(var = apply(df, 1, function(x) paste(gsub("NA", "", x), collapse = "")) )

      var
1     nuts
2     nuts
3  veggies
4  veggies
5   fruits
6   fruits
7   fruits
8   fruits
9   fruits
10  fruits

Answer 3

The actual data situation may determine whether is is better or worse than the line-by-line method. Here's one way to get a print-out like what you specify:

> as.matrix( df[df!="NA"] )

Or probably better:

> cat( paste( "\n", df[ df!="NA" ] ) )

 fruits 
 fruits 
 fruits 
 fruits 
 fruits 
 fruits 
 veggies 
 veggies 
 nuts 
 nuts