rio export fails for duration variable with label attribute, haven-export works

[hsbizahlenzauber]

MWE

rio export fails for duration variable with label attribute, haven-export works

library("lubridate")
library("rio")
library("haven")

#creating a variable of class period

duration <- seconds_to_period(duration(hours=26000))

#creating data frame

data <- data.frame(duration)

#setting variable label as attribute
#without this attribute, everything works as expected
attr(data$duration, "label") <- "Duration"

#export with haven
#no error message
haven::write_sav(data, path="testHAVEN.sav")

#export with rio
#Error in new_labelled():
#! x must be a numeric or a character vector.
rio::export(data, file="testRIO.sav")

[chainsawriot]

@hsbizahlenzauber Thank you for reporting this. I can reproduce this.

library(lubridate)
#> 
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union

library(rio)

duration <- seconds_to_period(duration(hours = 26000))

#creating data frame

duration_data <- data.frame(duration)
attr(duration_data$duration, "label") <- "Duration"
haven::write_sav(duration_data, path = tempfile(fileext = ".sav"))

rio::export(duration_data, file = tempfile(fileext = ".sav"))
#> Error in `new_labelled()`:
#> ! `x` must be a numeric or a character vector.

^{Created on 2024-06-20 with reprex v2.1.0}

And this is where it fails.

rio/R/standardize_attributes.R

Lines 38 to 58 in 588971f

	restore_labelled <- function(x) {
	# restore labelled variable classes
	x[] <- lapply(x, function(v) {
	if (is.factor(v)) {
	haven::labelled(
	x = as.numeric(v),
	labels = stats::setNames(seq_along(levels(v)), levels(v)),
	label = attr(v, "label", exact = TRUE)
	)
	} else if (!is.null(attr(v, "labels", exact = TRUE)) || !is.null(attr(v, "label", exact = TRUE))) {
	haven::labelled(
	x = v,
	labels = attr(v, "labels", exact = TRUE),
	label = attr(v, "label", exact = TRUE)
	)
	} else {
	v
	}
	})
	x
	}

TBH, I have to study this bunch of code to derive a solution, because I am not super familiar with these variable labeling code. It might take some time.

[chainsawriot]

755cdec #268

[chainsawriot]

@hsbizahlenzauber You know that haven (the package under the hood) does not actually support the Period class of lubridate, right?

library(lubridate)
#> 
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union

library(rio)

duration <- seconds_to_period(duration(hours = 26000))

#creating data frame

duration_data <- data.frame(duration)
duration_data
#>         duration
#> 1 1083d 8H 0M 0S

attr(duration_data$duration, "label") <- "Duration"
haven::write_sav(duration_data, path = x <- tempfile(fileext = ".sav"))

haven::read_sav(x)
#> # A tibble: 1 × 1
#>   duration
#>      <dbl>
#> 1        0

^{Created on 2024-06-20 with reprex v2.1.0}

[hsbizahlenzauber]

@chainsawriot Many thanks for taking care of this and for the tip. I have to admit that I'm not really familiar with the details of different data types. I report when I notice something and I'm very glad that there are people who deal with it.