purrr::pmap with dplyr::mutate

Question

I have a function which takes multiple inputs and creates multiple outputs. For example:

example_fun = function(a,b){
    x = a+b
    y = a-b
    return(list(x=x, y=y))
}

How can I use dplyr::mutate to evaluate this function on each row of a dataframe? Turn

df = expand.grid(a=c(7,8), b=c(9,10))

df
  a  b
1 7  9
2 8  9
3 7 10
4 8 10

into

  a  b  x  y
1 7  9 16 -2
2 8  9 17 -1
3 7 10 17 -3
4 8 10 18 -2

this following code almost accomplishes it:

df = df %>%
    mutate(outputs = pmap(list(a,b), example_fun)) %>%
    unnest()

df
  a  b outputs
1 7  9      16
2 7  9      -2
3 8  9      17
4 8  9      -1
5 7 10      17
6 7 10      -3
7 8 10      18
8 8 10      -2

Answer 1

Change things slightly:

example_fun = function(a, b) {
  x = a + b
  y = a - b
  return(data_frame(x = x, y = y)) #data_frame, not list
}

df <- data_frame(a = sample(1:5, 10, rep = TRUE), b = 11:20) #made my own test dataset

df %>%
  mutate(outputs = map2(a, b, example_fun)) %>% #I use map2 rather than pmap
  unnest()

Answer 2

We can do this using the original function in pmap by applying the function to get the output as a tibble and then bind_rows with the original dataset

df %>% 
   pmap_df(example_fun) %>%
   bind_cols(df, .)
#   a  b  x  y
#1 7  9 16 -2
#2 8  9 17 -1
#3 7 10 17 -3
#4 8 10 18 -2