tidyr: Tidy Messy Data

test_that("order doesn't matter", {
  df1 <- tibble(x = factor(c("a", "b")), y = 1:2)
  df2 <- tibble(x = factor(c("b", "a")), y = 2:1)
  one <- spread(df1, x, y)
  two <- spread(df2, x, y) %>%
    dplyr::select(a, b) %>%
    dplyr::arrange(a, b)
  expect_identical(one, two)

  df1 <- tibble(z = factor(c("b", "a")), x = factor(c("a", "b")), y = 1:2)
  df2 <- tibble(z = factor(c("a", "b")), x = factor(c("b", "a")), y = 2:1)
  one <- spread(df1, x, y) %>% dplyr::arrange(z)
  two <- spread(df2, x, y)
  expect_identical(one, two)
})

test_that("convert turns strings into integers", {
  df <- tibble(key = "a", value = "1")
  out <- spread(df, key, value, convert = TRUE)
  expect_type(out$a, "integer")
})

test_that("duplicate values for one key is an error", {
  df <- tibble(x = factor(c("a", "b", "b")), y = c(1, 2, 2), z = c(1, 2, 2))
  expect_snapshot(spread(df, x, y), error = TRUE)
})

test_that("factors are spread into columns (#35)", {
  data <- tibble(
    x = factor(c("a", "a", "b", "b")),
    y = factor(c("c", "d", "c", "d")),
    z = factor(c("w", "x", "y", "z"))
  )

  out <- data %>% spread(x, z)
  expect_equal(names(out), c("y", "a", "b"))
  expect_true(all(vapply(out, is.factor, logical(1))))
  expect_identical(levels(out$a), levels(data$z))
  expect_identical(levels(out$b), levels(data$z))
})

test_that("drop = FALSE keeps missing combinations (#25)", {
  df <- tibble(
    x = factor("a", levels = c("a", "b")),
    y = factor("b", levels = c("a", "b")),
    z = 1
  )
  out <- df %>% spread(x, z, drop = FALSE)
  expect_equal(nrow(out), 2)
  expect_equal(ncol(out), 3)
  expect_equal(out$a[2], 1)
})

test_that("drop = FALSE keeps missing combinations of 0-length factors (#56)", {
  df <- tibble(
    x = factor(, levels = c("a", "b")),
    y = factor(, levels = c("a", "b")),
    z = logical()
  )
  out <- df %>% spread(x, z, drop = FALSE)

  expect_equal(nrow(out), 2)
  expect_equal(ncol(out), 3)
  expect_equal(out$a, c(NA, NA))
  expect_equal(out$b, c(NA, NA))
})

test_that("drop = FALSE spread all levels including NA (#254)", {
  l <- c("a", "b", "c", "d")
  df <- tibble(
    x = factor(c("a", "b", "c", NA), levels = l),
    y = factor(c("a", "b", "c", "d")),
    z = factor(c("a", "b", "a", "b"))
  )
  out <- df %>% spread(x, y, drop = FALSE)
  expect_equal(nrow(out), 2)
  expect_equal(ncol(out), 6)
  expect_equal(out$d, factor(c(NA, NA), levels = l))
  expect_equal(out[["<NA>"]], factor(c(NA, "d"), levels = l))
})

test_that("spread preserves class of tibbles", {
  dat <- tibble(
    x = factor(c("a", "a", "b", "b")),
    y = factor(c("c", "d", "c", "d")),
    z = factor(c("w", "x", "y", "z"))
  )
  dat %>%
    spread(x, z) %>%
    expect_s3_class("tbl_df")
})

test_that("dates are spread into columns (#62)", {
  df <- tibble(
    id = c("a", "a", "b", "b"),
    key = c("begin", "end", "begin", "end"),
    date = Sys.Date() + 0:3
  )
  out <- spread(df, key, date)
  expect_identical(names(out), c("id", "begin", "end"))
  expect_s3_class(out$begin, "Date")
  expect_s3_class(out$end, "Date")
})

test_that("spread can produce mixed variable types (#118)", {
  df <- tibble(
    row = rep(1:2, 3),
    column = rep(1:3, each = 2),
    cell_contents = as.character(c(
      rep("Argentina", 2),
      62.485, 64.399,
      1952, 1957
    ))
  )
  out <- spread(df, column, cell_contents, convert = TRUE)
  expect_equal(
    unname(vapply(out, class, "")),
    c("integer", "character", "numeric", "integer")
  )
})

test_that("factors can be used with convert = TRUE to produce mixed types", {
  df <- tibble(
    row = c(1, 2, 1, 2, 1, 2),
    column = c("f", "f", "g", "g", "h", "h"),
    contents = c("aa", "bb", "1", "2", "TRUE", "FALSE")
  )
  out <- df %>% spread(column, contents, convert = TRUE)
  expect_type(out$f, "character")
  expect_type(out$g, "integer")
  expect_type(out$h, "logical")
})

test_that("dates can be used with convert = TRUE", {
  df <- tibble(
    id = c("a", "a", "b", "b"),
    key = c("begin", "end", "begin", "end"),
    date = Sys.Date() + 0:3
  )
  out <- spread(df, key, date, convert = TRUE)
  expect_type(out$begin, "character")
  expect_type(out$end, "character")
})

test_that("vars that are all NA are logical if convert = TRUE (#118)", {
  df <- tibble(
    row = c(1, 2, 1, 2), column = c("f", "f", "g", "g"),
    contents = c("aa", "bb", NA, NA)
  )
  out <- df %>% spread(column, contents, convert = TRUE)
  expect_type(out$g, "logical")
})

test_that("complex values are preserved  (#134)", {
  df <- expand.grid(id = 1:2, key = letters[1:2], stringsAsFactors = TRUE) %>%
    dplyr::mutate(value = 1:4 + 1i)

  out1 <- spread(df, key, value, convert = FALSE)
  out2 <- spread(df, key, value, convert = TRUE)

  expect_equal(out1$a, 1:2 + 1i)
  expect_equal(out2$a, 1:2 + 1i)
  expect_equal(out1$b, 3:4 + 1i)
  expect_equal(out2$b, 3:4 + 1i)
})

test_that("can spread with nested columns", {
  df <- tibble(x = c("a", "a"), y = 1:2, z = list(1:2, 3:5))
  out <- spread(df, x, y)

  expect_equal(out$a, 1:2)
  expect_equal(out$z, df$z)
})

test_that("spreading empty data frame gives empty data frame", {
  df <- tibble(x = character(), y = numeric(), z = character())
  rs <- spread(df, x, y)
  expect_equal(nrow(rs), 0)
  expect_named(rs, "z")

  df <- tibble(x = character(), y = numeric())
  rs <- spread(df, x, y)
  expect_equal(nrow(rs), 0)
  expect_equal(ncol(rs), 0)
})

test_that("spread gives one column when no existing non-spread vars", {
  df <- tibble(
    key = c("a", "b", "c"),
    value = c(1, 2, 3)
  )
  expect_equal(df %>% spread(key, value), tibble(a = 1, b = 2, c = 3))
})


test_that("grouping vars are kept where possible", {
  # Can keep
  df <- tibble(x = 1:2, key = factor(c("a", "b")), value = 1:2)
  out <- df %>%
    dplyr::group_by(x) %>%
    spread(key, value)
  expect_equal(dplyr::groups(out), list(quote(x)))

  # Can't keep
  df <- tibble(key = factor(c("a", "b")), value = 1:2)
  out <- df %>%
    dplyr::group_by(key) %>%
    spread(key, value)
  expect_equal(out, tibble(a = 1L, b = 2L))
})


test_that("col names never contains NA", {
  df <- tibble(x = c(1, NA), y = 1:2)
  df %>%
    spread(x, y) %>%
    expect_named(c("1", "<NA>"))
  df %>%
    spread(x, y, sep = "_") %>%
    expect_named(c("x_1", "x_NA"))
})

test_that("never has row names (#305)", {
  df <- tibble(id = 1:2, x = letters[1:2], y = 1:2)
  expect_false(
    df %>%
      spread(x, y) %>%
      tibble::has_rownames()
  )
})

test_that("overwrites existing columns", {
  df <- tibble(x = 1:2, y = 2:1, key = c("x", "x"), value = 3:4)
  rs <- df %>% spread(key, value)

  expect_named(rs, c("y", "x"))
  expect_equal(rs$x, 3:4)
})

test_that("spread doesn't convert data frames into tibbles", {
  df <- data.frame(x = c("a", "b"), y = 1:2)
  expect_equal(class(spread(df, x, y)), "data.frame")
})

test_that("spread with fill replaces explicit missing values", {
  df <- tibble(key = factor("a"), value = NA)
  out <- spread(df, key, value, fill = 1)
  expect_equal(out, tibble(a = 1))
})

test_that("spread with fill replaces implicit missing values", {
  # Missing keys in some groups
  df <- tibble(
    x = factor(c("G1", "G2")),
    key = factor(c("a", "b")),
    value = c(1, 1)
  )
  out <- spread(df, key, value, fill = 2)
  expect_equal(out, tibble(x = factor(c("G1", "G2")), a = c(1, 2), b = c(2, 1)))

  # Missing factor levels in key with drop = FALSE
  df <- tibble(key = factor("a", levels = c("a", "b")), value = 1)
  out <- spread(df, key, value, fill = 2, drop = FALSE)
  expect_equal(out, tibble(a = 1, b = 2))
})

test_that("ulevels preserves original factor levels", {
  x_na_lev <- factor(c("a", NA), exclude = NULL)
  expect_equal(levels(ulevels(x_na_lev)), c("a", NA))

  x_na_lev_extra <- factor(c("a", NA), levels = c("a", "b", NA), exclude = NULL)
  expect_equal(levels(ulevels(x_na_lev_extra)), c("a", "b", NA))

  x_no_na_lev <- factor(c("a", NA))
  expect_equal(levels(ulevels(x_no_na_lev)), "a")

  x_no_na_lev_extra <- factor(c("a", NA), levels = c("a", "b"))
  expect_equal(levels(ulevels(x_no_na_lev_extra)), c("a", "b"))
})

test_that("ulevels returns unique elements of a list for a list input", {
  test_list <- list(a = 1:6, b = 1:6)
  expect_equal(ulevels(test_list), unique(test_list))
})

test_that("spread works when id column has names (#525)", {
  df <- tibble(
    key = factor(c("a", "b", "c"), levels = letters[1:5]),
    out = 1:3,
    id  = c(a = 1, b = 2, c = 3)
  )
  res <- spread(df, key, out, drop = FALSE)
  expect_equal(names(res), c("id", letters[1:5]))
})