Add rbind.rowwise_df() method

Author: DavisVaughan

NAMESPACE

@@ -117,6 +117,7 @@ S3method(print,last_dplyr_warnings)
 S3method(print,src)
 S3method(pull,data.frame)
 S3method(rbind,grouped_df)
+S3method(rbind,rowwise_df)
 S3method(recode,character)
 S3method(recode,factor)
 S3method(recode,numeric)

NEWS.md

@@ -1,5 +1,7 @@
 # dplyr (development version)
 
+* New `rbind()` method for `rowwise_df` to avoid creating corrupt rowwise data frames (r-lib/vctrs#1935).
+
 * In `case_when()`, supplying all size 1 LHS inputs along with a size >1 RHS input is now soft-deprecated. This is an improper usage of `case_when()` that should instead be a series of if statements, like:
 
   ```

R/rowwise.R

@@ -171,3 +171,8 @@ as_tibble.rowwise_df <- function(x, ...) {
 
   rowwise_df(data, group_vars)
 }
+
+#' @export
+rbind.rowwise_df <- function(...) {
+  bind_rows(...)
+}

tests/testthat/test-rowwise.R

@@ -89,6 +89,33 @@ test_that("new_rowwise_df() can add class and attributes (#5918)", {
   expect_equal(attr(df, "a"), "b")
 })
 
+test_that("rbind() works with rowwise data frames by calling bind_rows() (r-lib/vctrs#1935)", {
+  x <- rowwise(tibble(a = 1:2))
+
+  y <- rowwise(tibble(a = 3:4))
+  out <- rbind(x, y)
+  expect_identical(out, rowwise(tibble(a = c(1:2, 3:4))))
+
+  # Important that `.rows` is recreated, not copied over from `x` (r-lib/vctrs#1935)
+  expect_identical(
+    group_data(out),
+    new_tibble(list(.rows = list_of(1L, 2L, 3L, 4L)))
+  )
+
+  # `bind_rows()` returns an object with the class of the first input,
+  # which is roughly how `rbind()` also works
+
+  # With bare tibble
+  y <- tibble(a = 5:6)
+  out <- rbind(x, z)
+  expect_identical(out, rowwise(tibble(a = c(1:2, 5:6))))
+
+  # With grouped_df
+  y <- group_by(tibble(a = 5:6), a)
+  out <- rbind(x, z)
+  expect_identical(out, rowwise(tibble(a = c(1:2, 5:6))))
+})
+
 test_that("validate_rowwise_df() gives useful errors", {
   df1 <- rowwise(tibble(x = 1:4, g = rep(1:2, each = 2)), g)
   groups <- attr(df1, "groups")