Can you give me a quick example of this? You can use gather_attrs() to pull them all to the data frame level, after which you could just drop them from there.

I didn't think about using gather_attrs() for this purpose—⁠thank you!

I keep running into situations like following:

## load packages
library("rio")
library("tidyr")

## import and gather
df_auto <- import("http://www.stata-press.com/data/r13/auto.dta") %>% 
  gather(vals, vars, price:foreign)

## output
Warning message:
attributes are not identical across measure variables;
they will be dropped 

I was thinking about a feature like import("stata.dta", remove_attrs = TRUE) but I look forward to using the gather_attrs() path.

Do you know if this happens with the new tidyr functions? I think gather/spread are also deprecated.

I don't think this is limited to deprecated functions, or to tidyr.

Here is another example with left_join from dplyr:

## load packages
library("rio")
library("dplyr")

## import and join 
df_auto <- import("http://www.stata-press.com/data/r13/auto.dta") %>% 
  left_join(., data.frame(trunk = as.numeric(5:23), grade = LETTERS[5:23]), 
            by = "trunk")

## output
Warning message:
Column `trunk` has different attributes on LHS and RHS of join 

In this example, the variable "trunk" does not have a label in the new (RHS) data frame, but if it did, and if that label did not match the label in the LHS data frame, we would still get this warning.

So, on further inspection, this behavior is basically correct. You'll want to gather_attrs() to clear variable label/descriptions before using tidyr::gather().

After that you'll still see warnings on this example because foreign is a factor/categorical variable and the others are all numeric/continuous so you either need to explicitly coerce to numeric (i.e. decide to drop its attributes) or coerce explicitly to factor using factorize() (which will still give you a warning because it's a factor and the others are numeric), or not gather it.