group_by() is converting dataframe to tibble

Question

I have a problem running st_as_sf() after group_by(). My code isn't working when previously I had no issues.

I discovered that group_by() was converting my dataframe to a tibble and st_as_sf() won't work with tibbles. I had peers run the same code and the conversion from dataframe to tibble did not occur for them. Is this something that's happening because of a new update?

R version 4.0.0, RStudio version 1.3.959.

Answer 1

As of dplyr 1.1.0, you can leverage .by. Using .by instead of group_by() has three main advantages:

• grouping is applied only to a single operation so no need to ungroup() afterwards
• original object class() maintained
• in the case of summarise(), group key sort order is preserved. Note that summarise() does not coerce to tibble, it is group_by() that does that

library(dplyr)

df <- data.frame(x = c("b", "b", "a", "a"),
                 y = 1:4)

df %>% group_by(x) %>%
  summarise(sum_y = sum(y)) %>%
  ungroup()

# # A tibble: 2 × 2
#   x     sum_y
#   <chr> <int>
# 1 a         7
# 2 b         3

df %>%
  summarise(sum_y = sum(y), .by = x)

#   x sum_y
# 1 b     3
# 2 a     7

Answer 2

This does seem true, at least with a development version of dplyr.

dd <- data.frame(x=rep(1:4,each=5), y=rnorm(20))
library(dplyr)
str(dd %>% group_by(x) %>% ungroup())
## tibble [20 × 2] (S3: tbl_df/tbl/data.frame)
##  $ x: int [1:20] 1 1 1 1 1 2 2 2 2 2 ...
##  $ y: num [1:20] -0.708 0.976 -1.051 -0.97 -0.225 ...

packageVersion("dplyr") is 0.8.99.9002.

I can't speak for how this behaved in previous versions. You could always add an as.data.frame() at the end of your pipeline ...