Why value in `mutate` ifelse custom condition is treated as a vector

[maciekbanas]

I can not understand why the value in the example below is treated as a vector and not scalar, when I try to put a simple string condition:

example_data <- data.frame(
  col_1 = c("John Test", "bobtest", "John Test"),
  col_2 = c(NA, "Bob Test", NA)
)

has_many_words <- function(char) {
  length(stringr::str_split_1(char, " ")) > 1
}

dplyr::mutate(
  example_data,
  col_2 = ifelse(is.na(col_2) & has_many_words(col_1), col_1, col_2)
)

Results in error:

My question is if only way to approach mutate with customized condition is with map iterator? This works, though does not look nice:

dplyr::mutate(
  example_data,
  col_2 = purrr::map2(col_1, col_2, function(x, y) {
    ifelse(is.na(y) & has_many_words(x), x, y)
  })
)

Is there any other way, am I getting something wrong?

[joranE]

Both ifelse & dplyr::if_else expect a vector, not a scalar, condition; i.e. mutate doesn't do any "looping" for you over values in the column. So the expected use of ifelse or dplyr::if_else is that your condition is already vectorized, and hence the looping happens via that vectorization.

In this case I believe you just need to switch to str_split and use lengths() rather than length:

has_many_words <- function(char) {
  lengths(stringr::str_split(char, " ")) > 1
}

[DavisVaughan]

I think @joranE is right that the main issue is that your has_many_words() function isn't vectorized quite right