Dplyr grouped column attribute causes "unknown column" error in unite()

[openclosure]

library(dplyr)
library(tidyr)
test <- data.frame(x = c("A", "B", "C"), y = 1, z = c(1, 2, 3)) %>% 
    group_by(x, y, z) %>%
    summarize(count = n())

# Doesn't work, can't find x
test %>% unite(newcol, x, z)
str(test)
# Notice for count, there is an attribute 'vars' with all grouped variable names

# Null out the attributes
attr(test, 'vars') <- NULL
str(test)

# Works
test %>% unite(newcol, x, z)

Can't figure out if this is a bug in dplyr or a bug in tidyr. I think this bug tidyverse/dplyr#859 is related and may not have been fixed in tidyr even though filter now works fine. Sorry if this is already known, I couldn't find any other mentions of this issue.

Edit: It appears that

test %>% tbl_df %>% unite(newcol, x, z) 

fixes the issue. However, the only thing this does is remove 'grouped_df' from the list of classes of the test object. Not sure if this is a bug at this point.

[aammd]

ungroup() also fixes this. It appears that unite() forbids you from modifying a grouping variable. It would be helpful if the error was the same/similar as that given by mutate() when you try to mutate the grouping variable: Error: cannot modify grouping variable