Modify arbitrary column attributes using dplyr::mutate()

Question

I would like to modify column attributes inside of dplyr::mutate(). Here is some example data:

library(dplyr, warn.conflicts = FALSE)

v1 <- tibble(
  id = 1:5,
  visit = 1,
  x = 1:5,
  y = c(0, 0, 0, 1, 1)
)

The end result I would like to get is the same result I would get from:

attr(v1$id, "source") <- "Collected at v1"
attr(v1$visit, "source") <- "Collected at v1"
attr(v1$x, "source") <- "Collected at v1"
attr(v1$y, "source") <- "Collected at v1"

I know I can just use a for loop.

for (col in names(v1)) {
  attr(v1[[col]], "source") <- "Collected at v1"
}

However, I have reasons for wanting to do this inside of dplyr. I'm trying to get to something like this.

v1 %>% 
  mutate(
    across(
      .cols = everything(),
      .fns  = ~ function_to_update_attributes("source", "Collected at v1")
    )
  )

Right now, I can't even get this to work for a single variable. This is the closest I've gotten.

v1 %>% 
  mutate(
    id = `<-`(attr(.[["id"]], "source"), "Collected at v1")
  )

Which returns

# A tibble: 5 × 4
  id              visit     x     y
  <chr>           <dbl> <int> <dbl>
1 Collected at v1     1     1     0
2 Collected at v1     1     2     0
3 Collected at v1     1     3     0
4 Collected at v1     1     4     1
5 Collected at v1     1     5     1

😞 Any constructive feedback is welcome and appreciated!

This is also posted on RStudio Community at: https://community.rstudio.com/t/modify-arbitrary-column-attributes-using-dplyr-mutate/144502

Answer 1

We can directly apply the attributes on the vector with dplyr

library(dplyr)
v1 <- v1 %>%
    mutate(across(everything(), 
     ~ {attr(.x, "source") <- "Collected at v1"
    .x} ))

-output

> str(v1)
tibble [5 × 4] (S3: tbl_df/tbl/data.frame)
 $ id   : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ visit: num [1:5] 1 1 1 1 1
  ..- attr(*, "source")= chr "Collected at v1"
 $ x    : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ y    : num [1:5] 0 0 0 1 1
  ..- attr(*, "source")= chr "Collected at v1"

Or if we want to use access the column name use cur_column() and assign with <<-

v1 %>%
  mutate(across(everything(), 
  ~ {attr(v1[[cur_column()]], "source") <<- "Collected at v1"
   .x})) 

-output

> str(v1)
tibble [5 × 4] (S3: tbl_df/tbl/data.frame)
 $ id   : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ visit: num [1:5] 1 1 1 1 1
  ..- attr(*, "source")= chr "Collected at v1"
 $ x    : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ y    : num [1:5] 0 0 0 1 1
  ..- attr(*, "source")= chr "Collected at v1"

---

Or if we want to replicate the same behavior i.e. making use of the original data object name along with its columns names as in for loop, use reduce

library(purrr)
v1 <- reduce(names(v1), ~ {
         attr(.x[[.y]], "source") <- "Collected at v1"
             .x}, .init = v1)

-output

> str(v1)
tibble [5 × 4] (S3: tbl_df/tbl/data.frame)
 $ id   : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ visit: num [1:5] 1 1 1 1 1
  ..- attr(*, "source")= chr "Collected at v1"
 $ x    : int [1:5] 1 2 3 4 5
  ..- attr(*, "source")= chr "Collected at v1"
 $ y    : num [1:5] 0 0 0 1 1
  ..- attr(*, "source")= chr "Collected at v1"

Answer 2

You can do it directly in a mutate(). You need to use set_attr() from the magrittr package.

mt2 <- mtcars %>% mutate(cyl2 = magrittr::set_attr(cyl, "source", "collected at v1"))

for all variables

mt2 <- mtcars %>% mutate(across(everything(), 
                         .fns = function(x) magrittr::set_attr(x, "source", "collected at v1")))