#
# Test plot features
#

library(testthat)
library(volker)
library(dplyr)

# --- Setup ------------------------------------------------------------------

# If vdiffr is not installed, skip all visual tests
if (requireNamespace("vdiffr", quietly = TRUE) && utils::packageVersion('testthat') >= '3.0.3') {
  expect_doppelganger <- vdiffr::expect_doppelganger
} else {
  expect_doppelganger <- function(...) skip("vdiffr is not installed.")
}

# Only run plot tests if explicitly enabled
# because they will fail on different machines due
# to different rendering engines.
# We have plot tests enabled in GitHub Actions for Windows.
# Thus, test them locally before pushing by temporarily
# enabling them. Make sure NOT to push the enabled setting,
# just the test results.
#
# Enable:
# Sys.setenv("R_LOCALTESTS" = "1")
# Disable:
# Sys.unsetenv("R_LOCALTESTS")


# Load the sample data
data <- volker::chatgpt

test_that("Empty plots are empty", {
  test_data <- tibble(var1=c(NA))

  plot_counts(test_data, var1) |>
    expect_error("Check your data: Are they empty?")

})

test_that("`.get_numbers_col()` toggles label column correctly", {
  expect_null(volker:::.get_numbers_col(NULL))
  expect_null(volker:::.get_numbers_col(FALSE))
  expect_null(volker:::.get_numbers_col(character(0)))

  # TRUE means: use prepared label col (default ".values")
  expect_identical(volker:::.get_numbers_col(TRUE), ".values")

  # character values also mean: use prepared label col
  expect_identical(volker:::.get_numbers_col("p"), ".values")
  expect_identical(volker:::.get_numbers_col(c("n","p")), ".values")

  # custom label column name override
  expect_identical(volker:::.get_numbers_col(TRUE, col = ".lab"), ".lab")
})


test_that(".plot_bars() adds text layer only when numbers_col is set", {
  dat <- tibble::tibble(
    item = factor("A"),
    value = factor("TRUE"),
    p = 35,
    n = 3,
    .values = "35%",
    .lab = "X"
  )

  # No numbers -> no geom_text
  p0 <- volker:::.plot_bars(dat, numbers_col = NULL)
  geoms0 <- purrr::map_chr(p0$layers, ~ class(.x$geom)[1])
  expect_false("GeomText" %in% geoms0)

  # numbers_col=".values" -> geom_text exists and maps to ".values"
  p1 <- volker:::.plot_bars(dat, numbers_col = ".values")
  geoms1 <- purrr::map_chr(p1$layers, ~ class(.x$geom)[1])
  expect_true("GeomText" %in% geoms1)

  idx_text <- which(geoms1 == "GeomText")[1]
  # mapping label should reference the chosen column
  expect_true(!is.null(p1$layers[[idx_text]]$mapping$label))
  expect_match(rlang::as_label(p1$layers[[idx_text]]$mapping$label), "\\.values")

  # numbers_col=".lab" -> maps to ".lab"
  p2 <- volker:::.plot_bars(dat, numbers_col = ".lab")
  geoms2 <- purrr::map_chr(p2$layers, ~ class(.x$geom)[1])
  idx_text2 <- which(geoms2 == "GeomText")[1]
  expect_match(rlang::as_label(p2$layers[[idx_text2]]$mapping$label), "\\.lab")
})


test_that(".plot_heatmap() uses numbers_col for text labels", {
  dat <- tibble::tibble(
    item1 = factor(c("A","A","B","B")),
    item2 = factor(c("A","B","A","B")),
    n = c(1,2,3,4),
    npmi = c(0.1, -0.2, 0.3, 0.0),
    .values = c("1","2","3","4"),
    .lab = c("a","b","c","d")
  )

  # Without numbers_col -> no geom_text
  p0 <- volker:::.plot_heatmap(dat, values_col = "npmi", numbers_col = NULL, title = FALSE, labels = FALSE)
  geoms0 <- purrr::map_chr(p0$layers, ~ class(.x$geom)[1])
  expect_false("GeomText" %in% geoms0)
  # With numbers_col -> geom_text and maps to the right column
  p1 <- volker:::.plot_heatmap(dat, values_col = "npmi", numbers_col = ".values", title = FALSE, labels = FALSE)
  geoms1 <- purrr::map_chr(p1$layers, ~ class(.x$geom)[1])
  expect_true("GeomText" %in% geoms1)
  idx_text <- which(geoms1 == "GeomText")[1]
  expect_match(rlang::as_label(p1$layers[[idx_text]]$mapping$label), "\\.values")

  # custom label column
  p2 <- volker:::.plot_heatmap(dat, values_col = "npmi", numbers_col = ".lab", title = FALSE, labels = FALSE)
  geoms2 <- purrr::map_chr(p2$layers, ~ class(.x$geom)[1])
  idx_text2 <- which(geoms2 == "GeomText")[1]
  expect_match(rlang::as_label(p2$layers[[idx_text2]]$mapping$label), "\\.lab")
})


if (Sys.getenv("R_LOCALTESTS") == "1") {

  ggplot2::theme_set(ggplot2::theme_test())
  #ggplot2::theme_set(volker::theme_vlkr())

  test_that("Categorical plots render as expected", {

    expect_doppelganger("plot_counts_one", {
      data |> plot_counts(adopter)
    })

    expect_doppelganger("plot_counts_one with binary data", {
      data |> mutate(adopter = adopter == "I try new offers immediately") |> plot_counts(adopter)
    })

    expect_doppelganger("plot_counts_one with focus category", {
      data |> plot_counts(adopter, category = "I try new offers immediately")
    })

    expect_doppelganger("plot_counts_one_grouped", {
      plot_counts(data, sd_gender, adopter)
    })

    expect_doppelganger("plot_counts_one_grouped with rowsprops", {
      data |> plot_counts(adopter, sd_gender, prop = "rows", numbers = "p")
    })

    expect_doppelganger("plot_counts_one_grouped with cols props", {
      plot_counts(data, adopter, sd_gender, prop = "cols", numbers = "p", ordered = -1)
    })

    expect_doppelganger("plot_counts_one_grouped_reversed", {
      data |> mutate(adopter = factor(adopter, levels = rev(levels(adopter)))) |>
        plot_counts(adopter, sd_gender, prop = "cols")
    })

    expect_doppelganger("plot_counts_one_cor", {
      plot_counts(data, adopter, sd_age, metric = TRUE)
    })

    expect_doppelganger("plot_counts_one_cor with rows props", {
      plot_counts(data, adopter, sd_age, metric = TRUE, prop = "rows")
    })


    expect_doppelganger("plot_counts_items", {
      plot_counts(data, starts_with("cg_adoption_"), numbers = "p")
    })

    expect_doppelganger("plot_counts_items with binary data", {
      data |> mutate(across(starts_with("cg_adoption_advantage"), \(x) x > 3)) |>
        plot_counts(starts_with("cg_adoption_advantage"))
    })

    expect_doppelganger("plot_counts_items_grouped", {
      plot_counts(
        data, starts_with("cg_adoption_"), adopter,
        category = c("agree", "strongly agree")
      )
    })

    # TODO: Fix theming of tile stroke width
    # expect_doppelganger("plot_counts_items_grouped_items", {
    #   plot_counts(
    #     data, starts_with("cg_adoption_"), starts_with("cg_adoption_")
    #   )
    # })

    expect_doppelganger("plot_counts_items_cor", {
      plot_counts(
        data, starts_with("cg_adoption_"), sd_age, metric = TRUE,
        category = c(4, 5)
      )
    })

  })

  test_that("Metric distribution plots render as expected", {

    expect_doppelganger("plot_metrics_one", {
      plot_metrics(data, sd_age)
    })

    expect_doppelganger("plot_metrics_one with box", {
      plot_metrics(data, sd_age, box = TRUE)
    })

    expect_doppelganger("plot_metrics_one with ci", {
      plot_metrics(data, sd_age, ci = TRUE)
    })

    expect_doppelganger("plot_metrics_one_grouped", {
      plot_metrics(data, sd_age, sd_gender)
    })

    expect_doppelganger("plot_metrics_one_cor", {
      plot_metrics(data, sd_age, use_work, metric = TRUE)
    })

    expect_doppelganger("plot_metrics_one_cor with log scale", {
      plot_metrics(data, sd_age, use_work, metric = TRUE, log = TRUE)
    })

    expect_doppelganger("plot_metrics_one_cor without title", {
      plot_metrics(data, sd_age, use_work, metric = TRUE, title = FALSE)
    })

    expect_doppelganger("plot_metrics_items", {
      plot_metrics(data, starts_with("cg_adoption_"))
    })

    expect_doppelganger("plot_metrics_items_grouped", {
      plot_metrics(data, starts_with("cg_adoption_"), sd_gender)
    })

    expect_doppelganger("plot_metrics_items_cor with pearson's r", {
      plot_metrics(data, starts_with("use_"), sd_age, metric = TRUE)
    })

    expect_doppelganger("plot_metrics_items_cor with spearman's rho", {
      plot_metrics(data, starts_with("use_"), sd_age, metric = TRUE, method = "spearman")
    })

    # TODO: Fix theming of tile stroke width
    # expect_doppelganger("plot_metrics_items_cor_items", {
    #   plot_metrics(data, starts_with("cg_adoption_adv"), starts_with("use_"), metric = TRUE)
    # })

    # TODO: Fix theming of tile stroke width
    # expect_doppelganger("plot_metrics_items_cor_items with numbers", {
    #   plot_metrics(
    #     data, starts_with("cg_adoption_adv"),
    #     starts_with("use_"),
    #     metric = TRUE,
    #     numbers = TRUE
    #   )
    # })
  })

  test_that("Index plots render as expected", {

    expect_doppelganger("index_distribution_grouped", {
      data %>%
        add_index(starts_with("cg_adoption_social_")) %>%
        add_index(starts_with("cg_adoption_fearofuse_")) %>%
        add_index(starts_with("cg_adoption_advantage_")) %>%
        labs_apply(tribble(
          ~item_name, ~item_label,
          "idx_cg_adoption_social", "ChatGPT-Erwartung: Social",
          "idx_cg_adoption_fearofuse", "ChatGPT-Erwartung: Fear of Use",
          "idx_cg_adoption_advantage", "ChatGPT-Erwartung: Advantage"
        )) %>%
        plot_metrics(starts_with("idx_cg_adoption"), adopter)
    })

  })

  test_that("New labels render as expected", {

    expect_doppelganger("apply_value_labels_plot", {
      data %>%
        filter(sd_gender != "diverse") %>%
        labs_apply(
          cols = sd_gender,
          values = list(
            "female" = "Weiblich",
            "male" = "Maennlich"
          )
        ) %>%
        plot_metrics(sd_age, sd_gender)
    })
  })


} else {
  skip("R_LOCALTESTS not set; skipping visual tests.")
}