[SOLVED] Why replace_na with mutate doesn't work as expected?

Why replace_na with mutate doesn't work as expected?

Why:

data.frame(a=c(TRUE,NA)) %>% mutate(a=replace_na(FALSE))

returns

      a
1 FALSE
2 FALSE

and so set the entire column to the value instead of just the NA elements?

In this case mutate is supposed to call the "vector" version of replace_na, that is, it should be equivalent to:

> df <- data.frame(a=c(TRUE,NA))
> df$a <- replace_na(df$a, FALSE)

The docs for replace_na(data, replace) says:

If data is a vector, replace takes a single value. This single value replaces all of the missing values in the vector. replace will be cast to the type of data.

On a side note, and not related to this question,

data.frame(a=c(TRUE,NA)) %>% replace_na(list("a"=FALSE))

works as expected.

Solution

You are misusing mutate() by not giving the replace_na() function a column to operate on. Running replace_na(FALSE) simple returns FALSE and that explains your result. The third line of code in the reprex gives the result you want.

library(tidyverse)
data.frame(a=c(TRUE,NA)) %>% mutate(a= replace_na(FALSE))
#>       a
#> 1 FALSE
#> 2 FALSE

replace_na(FALSE)
#> [1] FALSE

data.frame(a=c(TRUE,NA)) %>% mutate(a= replace_na(a,FALSE))
#>       a
#> 1  TRUE
#> 2 FALSE

^{Created on 2024-11-13 with reprex v2.1.1}