tidyr: Tidy Messy Data

test_that("uses input for default column names", {
  df <- tibble(x = 1:2, y = list(1, 1:2))
  out <- df %>% unnest_longer(y)

  expect_named(out, c("x", "y"))
})

test_that("can adjust the column name with `values_to`", {
  df <- tibble(x = 1:2, y = list(1, 1:2))
  out <- df %>% unnest_longer(y, values_to = "y2")

  expect_named(out, c("x", "y2"))
})

test_that("automatically adds id col if named", {
  df <- tibble(x = 1:2, y = list(c(a = 1), c(b = 2)))
  out <- df %>% unnest_longer(y)

  expect_named(out, c("x", "y", "y_id"))
})

test_that("can force integer indexes", {
  df <- tibble(x = 1:2, y = list(1, 2))
  out <- df %>% unnest_longer(y, indices_include = TRUE)
  expect_named(out, c("x", "y", "y_id"))

  out <- df %>% unnest_longer(y, indices_to = "y2")
  expect_named(out, c("x", "y", "y2"))
})

test_that("can handle data frames consistently with vectors", {
  df <- tibble(x = 1:2, y = list(tibble(a = 1:2, b = 2:3)))
  out <- df %>% unnest_longer(y)

  expect_named(out, c("x", "y"))
  expect_equal(nrow(out), 4)
})

test_that("can unnest dates", {
  x <- as.Date(c("2019-08-01", "2019-12-01"))
  df <- tibble(x = as.list(x))
  out <- df %>% unnest_longer(x)
  expect_equal(out$x, x)
})

test_that("unnest_longer - bad inputs generate errors", {
  df <- tibble(x = 1, y = list(mean))

  expect_snapshot((expect_error(
    unnest_longer(df, y)
  )))
})

test_that("list_of columns can be unnested", {
  df <- tibble(x = 1:2, y = list_of(1L, 1:2))
  out <- unnest_longer(df, y)

  expect_named(out, c("x", "y"))
  expect_equal(nrow(out), 3)

  # With id column
  df <- tibble(x = 1:2, y = list_of(c(a = 1L), c(b = 1:2)))
  expect_named(unnest_longer(df, y), c("x", "y", "y_id"))
})

test_that("drops empty rows by default (#1363, #1339)", {
  df <- tibble(
    x = 1:4,
    y = list(NULL, NULL, 1, double())
  )

  out <- unnest_longer(df, y)
  expect_identical(out, tibble(x = 3L, y = 1))
})

test_that("can keep empty rows with `keep_empty` (#1339)", {
  df <- tibble(
    x = 1:4,
    y = list(NULL, NULL, 1, double())
  )

  out <- unnest_longer(df, y, keep_empty = TRUE)
  expect_identical(out, tibble(x = 1:4, y = c(NA, NA, 1, NA)))
})

test_that("keeping empty rows uses `NA` as the index", {
  df <- tibble(
    x = 1:3,
    y = list(NULL, 1:2, integer())
  )

  # Integer case
  out <- unnest_longer(df, y, keep_empty = TRUE, indices_include = TRUE)
  expect_identical(out$y_id, c(NA, 1L, 2L, NA))

  # Character case
  # Trigger names to be generated
  df$y[[2]] <- set_names(df$y[[2]], c("a", "b"))
  out <- unnest_longer(df, y, keep_empty = TRUE)
  expect_identical(out$y_id, c(NA, "a", "b", NA))
})

test_that("named empty vectors force an index column regardless of `keep_empty`", {
  df <- tibble(
    x = 1:2,
    y = list(1:2, set_names(integer(), character()))
  )

  # Empty row is dropped, but names are still forced
  out <- unnest_longer(df, y)
  expect_identical(out$y_id, c("", ""))

  out <- unnest_longer(df, y, keep_empty = TRUE)
  expect_identical(out$y_id, c("", "", NA))
})

test_that("mix of unnamed and named can be unnested (#1029)", {
  df <- tibble(x = 1:4, y = list(1, c(b = 2), NULL, double()))

  out <- unnest_longer(df, y, indices_include = NULL)
  expect_identical(out$y_id, c("", "b"))

  out <- unnest_longer(df, y, indices_include = TRUE)
  expect_identical(out$y_id, c("", "b"))

  out <- unnest_longer(df, y, indices_include = NULL, keep_empty = TRUE)
  expect_identical(out$y_id, c("", "b", NA, NA))

  out <- unnest_longer(df, y, indices_include = TRUE, keep_empty = TRUE)
  expect_identical(out$y_id, c("", "b", NA, NA))
})

test_that("unnesting empty typed column is a no-op and retains column (#1199) (#1196)", {
  df <- tibble(x = integer())
  expect_identical(unnest_longer(df, x), df)

  df <- tibble(x = tibble())
  expect_identical(unnest_longer(df, x), df)

  df <- tibble(x = tibble(a = integer()))
  expect_identical(unnest_longer(df, x), df)

  # In particular, #1196
  df <- tibble(a = tibble(x = 1:2, y = 3:4, z = 5:6))
  expect_identical(unnest_longer(df, a), df)
})

test_that("unnesting empty list retains logical column (#1199)", {
  # Really an unspecified column that `vec_cast_common()` finalizes to logical
  df <- tibble(x = list())
  expect_identical(unnest_longer(df, x), tibble(x = logical()))
})

test_that("unnesting empty list with indices uses integer indices", {
  df <- tibble(x = list())

  out <- unnest_longer(df, x, indices_include = TRUE)
  expect_identical(out$x_id, integer())

  out <- unnest_longer(df, x, indices_include = TRUE, keep_empty = TRUE)
  expect_identical(out$x_id, integer())
})

test_that("unnesting empty list-of retains ptype (#1199)", {
  df <- tibble(x = list_of(.ptype = integer()))
  expect_identical(unnest_longer(df, x), tibble(x = integer()))
})

test_that("unnesting list of data frames utilizes `values_to` (#1195)", {
  df <- tibble(x = list(tibble(a = 1:2), tibble(a = 3:4)))

  expect_identical(
    unnest_longer(df, x, values_to = "foo"),
    tibble(foo = tibble(a = 1:4))
  )
})

test_that("unnesting list of data frames utilizes `indices_include` (#1194)", {
  df <- tibble(x = list(tibble(a = 1:2), tibble(a = 3:4)))

  expect_identical(
    unnest_longer(df, x, indices_include = TRUE),
    tibble(x = tibble(a = 1:4), x_id = c(1L, 2L, 1L, 2L))
  )
})

test_that("can unnest a column with just `list(NULL)` or `list_of(NULL)` (#1193, #1363)", {
  df <- tibble(x = list(NULL))

  expect_identical(
    unnest_longer(df, x),
    tibble(x = logical())
  )
  expect_identical(
    unnest_longer(df, x, keep_empty = TRUE),
    tibble(x = NA)
  )

  df <- tibble(x = list_of(NULL, .ptype = integer()))

  expect_identical(
    unnest_longer(df, x),
    tibble(x = integer())
  )
  expect_identical(
    unnest_longer(df, x, keep_empty = TRUE),
    tibble(x = NA_integer_)
  )
})

test_that("can unnest a column with just `list(integer())`", {
  df <- tibble(x = list(integer()))

  expect_identical(
    unnest_longer(df, x),
    tibble(x = integer())
  )
  expect_identical(
    unnest_longer(df, x, keep_empty = TRUE),
    tibble(x = NA_integer_)
  )
})

test_that("unnesting `list(NULL)` with indices uses integer indices", {
  df <- tibble(x = list(NULL))

  out <- unnest_longer(df, x, indices_include = TRUE)
  expect_identical(out$x_id, integer())

  out <- unnest_longer(df, x, indices_include = TRUE, keep_empty = TRUE)
  expect_identical(out$x_id, NA_integer_)
})

test_that("can unnest one row data frames (#1034)", {
  col <- list(tibble(x = 1, y = 2), tibble(x = 2, y = 3))
  df <- tibble(col = col)

  expect_identical(
    unnest_longer(df, col),
    tibble(col = tibble(x = c(1, 2), y = c(2, 3)))
  )
})

test_that("named vectors are converted to lists with `vec_chop()`", {
  # Equivalent to `df <- tibble(x = list_of(c(a = 1), c(b = 2)))`
  df <- tibble(x = c(a = 1, b = 2))

  out <- unnest_longer(df, x)

  expect_identical(out$x, c(a = 1, b = 2))
  expect_identical(out$x_id, c("a", "b"))
})

test_that("can unnest multiple columns (#740)", {
  df <- tibble(a = list(1:2, 3:4), b = list(1:2, 3:4))
  expect_identical(unnest_longer(df, c(a, b)), unchop(df, c(a, b)))
})

test_that("tidyverse recycling rules are applied when unnesting multiple cols", {
  df <- tibble(a = list(1L, 3:4), b = list(1:2, 4L))

  out <- unnest_longer(df, c(a, b))
  expect_identical(out$a, c(1L, 1L, 3L, 4L))
  expect_identical(out$b, c(1L, 2L, 4L, 4L))
})

test_that("tidyverse recycling rules are applied after `keep_empty`", {
  df <- tibble(a = list(NULL, 3:4), b = list(1:2, 4L))

  expect_snapshot(error = TRUE, {
    unnest_longer(df, c(a, b))
  })

  out <- unnest_longer(df, c(a, b), keep_empty = TRUE, indices_include = TRUE)

  expect_identical(out$a, c(NA, NA, 3L, 4L))
  expect_identical(out$a_id, c(NA, NA, 1L, 2L))

  expect_identical(out$b, c(1L, 2L, 4L, 4L))
  expect_identical(out$b_id, c(1L, 2L, 1L, 1L))
})

test_that("unnesting multiple columns uses independent indices", {
  df <- tibble(a = list(c(x = 1), NULL), b = list(1, 2:3))
  out <- unnest_longer(df, c(a, b), keep_empty = TRUE)

  expect_identical(out$a_id, c("x", NA, NA))
  expect_named(out, c("a", "a_id", "b"))
})

test_that("unnesting multiple columns works with `indices_include = TRUE`", {
  df <- tibble(a = list(c(x = 1), NULL), b = list(1, 2:3))
  out <- unnest_longer(df, c(a, b), keep_empty = TRUE, indices_include = TRUE)

  expect_identical(out$a_id, c("x", NA, NA))
  expect_identical(out$b_id, c(1L, 1L, 2L))
})

test_that("can use glue to name multiple `values_to` cols", {
  df <- tibble(a = list(1, 2:3), b = list(1, 2:3))
  expect_named(
    unnest_longer(df, c(a, b), values_to = "{col}_"),
    c("a_", "b_")
  )
})

test_that("can use glue to name multiple `indices_to` cols", {
  df <- tibble(a = list(1, 2:3), b = list(1, 2:3))
  expect_named(
    unnest_longer(df, c(a, b), indices_to = "{col}_name"),
    c("a", "a_name", "b", "b_name")
  )
})

test_that("default `indices_to` is based on `values_to` (#1201)", {
  df <- tibble(a = list(c(x = 1), 2))
  expect_named(
    unnest_longer(df, a, values_to = "aa"),
    c("aa", "aa_id")
  )
})

test_that("can unnest a vector with a mix of named/unnamed elements (#1200 comment)", {
  df <- tibble(x = c(a = 1L, 2L))
  out <- unnest_longer(df, x)
  expect_identical(out$x, df$x)
  expect_identical(out$x_id, c("a", ""))
})

test_that("can unnest a list with a mix of named/unnamed elements (#1200 comment)", {
  df <- tibble(x = list(a = 1:2, 3:4))
  out <- unnest_longer(df, x)
  expect_identical(out$x, 1:4)
})

test_that("names are preserved when simplification isn't done and a ptype is supplied", {
  df <- tibble(x = list(list(a = 1L), list(b = 1L)))
  ptype <- list(x = integer())

  # Explicit request not to simplify
  out <- unnest_longer(df, x, indices_include = TRUE, ptype = ptype, simplify = FALSE)
  expect_named(out$x, c("a", "b"))
  expect_identical(out$x_id, c("a", "b"))

  df <- tibble(x = list(list(a = 1:2), list(b = 1L)))
  ptype <- list(x = integer())

  # Automatically can't simplify
  out <- unnest_longer(df, x, indices_include = TRUE, ptype = ptype)
  expect_named(out$x, c("a", "b"))
  expect_identical(out$x_id, c("a", "b"))
})

test_that("works with foreign lists recognized by `vec_is_list()` (#1327)", {
  new_foo <- function(...) {
    structure(list(...), class = c("foo", "list"))
  }

  # With empty types
  df <- tibble(x = new_foo(1:2, integer()))
  expect_identical(unnest_longer(df, x), tibble(x = 1:2))
  expect_identical(unnest_longer(df, x, keep_empty = TRUE), tibble(x = c(1:2, NA)))

  # With `NULL`s
  df <- tibble(x = new_foo(1:2, NULL))
  expect_identical(unnest_longer(df, x), tibble(x = 1:2))
  expect_identical(unnest_longer(df, x, keep_empty = TRUE), tibble(x = c(1:2, NA)))
})

test_that("can't currently retain names when simplification isn't done and a ptype is supplied if there is a mix of named/unnamed elements (#1212)", {
  df <- tibble(x = list(list(a = 1L), list(1L)))
  ptype <- list(x = integer())

  out <- unnest_longer(df, x, indices_include = TRUE, ptype = ptype, simplify = FALSE)

  expect_named(out$x, c("a", ""))
  expect_identical(out$x_id, c("a", ""))
})

test_that("can't mix `indices_to` with `indices_include = FALSE`", {
  expect_snapshot((expect_error(
    unnest_longer(mtcars, mpg, indices_to = "x", indices_include = FALSE)
  )))
})

test_that("unnest_longer() validates its inputs", {
  df <- tibble(x = list(list(a = 1L), list(b = 1L)))
  expect_snapshot(error = TRUE, {
    unnest_longer(1)
    unnest_longer(df)
    unnest_longer(df, x, indices_to = "")
    unnest_longer(df, x, indices_include = 1)
    unnest_longer(df, x, values_to = "")
  })
})

test_that("`values_to` and `indices_to` glue can't reach into surrounding env", {
  x <- "foo"

  expect_error(unnest_longer(mtcars, mpg, indices_to = "{x}"))
  expect_error(unnest_longer(mtcars, mpg, values_to = "{x}"))
})

test_that("`values_to` is validated", {
  expect_snapshot({
    (expect_error(unnest_longer(mtcars, mpg, values_to = 1)))
    (expect_error(unnest_longer(mtcars, mpg, values_to = c("x", "y"))))
  })
})

test_that("`indices_to` is validated", {
  expect_snapshot({
    (expect_error(unnest_longer(mtcars, mpg, indices_to = 1)))
    (expect_error(unnest_longer(mtcars, mpg, indices_to = c("x", "y"))))
  })
})

test_that("`indices_include` is validated", {
  expect_snapshot({
    (expect_error(unnest_longer(mtcars, mpg, indices_include = 1)))
    (expect_error(unnest_longer(mtcars, mpg, indices_include = c(TRUE, FALSE))))
  })
})

test_that("`keep_empty` is validated", {
  expect_snapshot({
    (expect_error(unnest_longer(mtcars, mpg, keep_empty = 1)))
    (expect_error(unnest_longer(mtcars, mpg, keep_empty = c(TRUE, FALSE))))
  })
})