Using `All()` with a labelled variable fails

[etiennebacher]

Hello, thanks for this amazing package!

I have a problem with tables::tabular() when I use All() in the formula and there's a least one labelled variable in the data:

library(tables)
packageVersion("tables")
#> [1] '0.9.21'

dat <- mtcars[, 1, drop = FALSE]
tabular(All(dat) ~ mean, dat)
#>           
#>      mean 
#>  mpg 20.09

dat$mpg <- haven::labelled(dat$mpg, label = "miles per gallon")
tabular(All(dat) ~ mean, dat)
#> Error in switch(deparse(e[[1]]), `*` = c(factors(e[[2]]), factors(e[[3]])), : EXPR must be a length 1 vector

[dmurdoch]

I can confirm the issue. The problem is that tabular isn't recognizing dat$mpg as a vector of doubles, because haven put class "labelled" on it. I'm not a haven user, so I don't know the recommended way to remove that class, but you will need to do that to get it to work. (I tried zap_label(), but it doesn't work. It leaves the class there. Bug? unclass() works, but leaves the label attribute. Maybe unclass(zap_label())?)

If you read the haven::labelled help page, you'll see that tabular is not alone in having trouble dealing with labelled objects.

[dmurdoch]

Sorry, this really is a bug in tables. It doesn't have much to do with haven: the problem is that deparse(e[[1]]) was two lines long, and the switch() function wants something that's only one line long. This is easy to fix; a bug fix will soon appear on Github.

[etiennebacher]

Great! FYI this will be useful for modelsummary users because there's an option to automatically replace variable names by their labels but it doesn't work with All() due to this issue.

[dmurdoch]

I've now merged the fix into the main. Could you confirm that things are working for you?

[etiennebacher]

Yup, all good, thanks for the quick fix