Generators labelled() and labelled_spss() don’t support adding label attribute

[huftis]

The labelled() and labelled_spss() generator functions supports adding a labels attribute, but they don’t have a built-in mechanism for adding a label attribute. (The labelattribute hold a short descripton of the variable.) This despite the fact that haven actually has very good support for the label attribute, e.g. when both writing and reading SPSS files.

Suggestion: Add a label argument to the labelled() and labelled_spss() functions. This would add the argument content as a label attribute. Although the argument logically belongs next to the labels argument, it’s probably best to add it as the last argument of the functions, in case people have used argument position when calling the functions.

Here’s a reprex illustrating that 1) the generator functions have no arguments for specifying the label attribute, and 2) exporting and importing to the SPSS format does support the label attribute (and correctly loses ‘unknown’ attributes):

library(haven)
library(tibble)

# No label argument
args(labelled)
#> function (x, labels) 
#> NULL
args(labelled_spss)
#> function (x, labels, na_values = NULL, na_range = NULL) 
#> NULL

# Example data
gen = labelled(c(0, 0, 1, 99), 
               labels = c(Female = 0, Male = 1))
attr(gen, "label") = "Patient gender"
attr(gen, "foo") = "bar"
d = tibble(gen)

# Write to SPSS and read resulting file
fname = tempfile()
write_sav(d, fname)
d2 = read_sav(fname)

# The `label` and `labels` attributes survived the conversion,
# but the `foo` attribute (correctly) did not
str(d2)
#> Classes 'tbl_df', 'tbl' and 'data.frame':    4 obs. of  1 variable:
#>  $ gen:Class 'labelled'  atomic [1:4] 0 0 1 99
#>   .. ..- attr(*, "label")= chr "Patient gender"
#>   .. ..- attr(*, "format.spss")= chr "F8.2"
#>   .. ..- attr(*, "labels")= Named num [1:2] 0 1
#>   .. .. ..- attr(*, "names")= chr [1:2] "Female" "Male"

I could probably create a PR if this is of interest.

[hadley]

A PR would be great!

[huftis]

Will do, @hadley. But PR #353 needs to be merged first, as the needed code changes partially overlap.

[hadley]

Ok, just fixed that myself to avoid blocking you.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/