r - Adding multiple columns in a dplyr mutate call

Question

I have a data frame with a dot-separated character column:

> set.seed(310366)
> tst = data.frame(x=1:10,y=paste(sample(c("FOO","BAR","BAZ"),10,TRUE),".",sample(c("foo","bar","baz"),10,TRUE),sep=""))
> tst
    x       y
1   1 BAR.baz
2   2 FOO.foo
3   3 BAZ.baz
4   4 BAZ.foo
5   5 BAZ.bar
6   6 FOO.baz
7   7 BAR.bar
8   8 BAZ.baz

and I want to split that column into two new columns containing the parts on either side of the dot. str_split_fixed from package stringr can do the job quite nicely. All my values are definitely two parts separated by a dot so I can do:

> require(stringr)
> str_split_fixed(tst$y,"\.",2)
      [,1]  [,2] 
 [1,] "BAR" "baz"
 [2,] "FOO" "foo"
 [3,] "BAZ" "baz"
 [4,] "BAZ" "foo"
 [5,] "BAZ" "bar"
 [6,] "FOO" "baz"
 [7,] "BAR" "bar"

Now I could just cbind that to my data frame but I thought I'd figure out how to do that in a dplyr pipeline. First I thought mutate could do it in one:

> tst %.% mutate(parts=str_split_fixed(y,"\.",2))
Error: wrong result size (20), expected 10 or 1

I can get mutate to do it in two:

> tst %.% mutate(part1=str_split_fixed(y,"\.",2)[,1], part2=str_split_fixed(y,"\.",2)[,2])
    x       y part1 part2
1   1 BAR.baz   BAR   baz
2   2 FOO.foo   FOO   foo
3   3 BAZ.baz   BAZ   baz
4   4 BAZ.foo   BAZ   foo
5   5 BAZ.bar   BAZ   bar
6   6 FOO.baz   FOO   baz

but that's running the string split twice.

"Best" I can do so far in a dplyr way is this (which I only discovered while writing this question...):

> tst %.% do(cbind(.,data.frame(parts=str_split_fixed(.$y,"\.",2))))
    x       y parts.1 parts.2
1   1 BAR.baz     BAR     baz
2   2 FOO.foo     FOO     foo
3   3 BAZ.baz     BAZ     baz
4   4 BAZ.foo     BAZ     foo
5   5 BAZ.bar     BAZ     bar

which isn't bad, but loses a lot of the readability of piped things in R. Is there a simple approach using mutate that I've missed?

Answer 1

You can use separate() from tidyr in combination with dplyr:

tst %>% separate(y, c("y1", "y2"), sep = "\.", remove=FALSE)

    x       y  y1  y2
1   1 BAR.baz BAR baz
2   2 FOO.foo FOO foo
3   3 BAZ.baz BAZ baz
4   4 BAZ.foo BAZ foo
5   5 BAZ.bar BAZ bar
6   6 FOO.baz FOO baz
7   7 BAR.bar BAR bar
8   8 BAZ.baz BAZ baz
9   9 FOO.bar FOO bar
10 10 BAR.foo BAR foo

Setting remove=TRUE will remove column y