tests/testthat/test_aggregate.R

In ELToulemonde/dataPreparation: Automated Data Preparation

context("test_aggregate.R")
requireNamespace("data.table")
verbose <- TRUE

## aggregate_by_key
# ---------------
test_that("aggregate_by_key: with unique key, should preserve the number of line", {
    # Given
    data("adult")
    adult <- adult[1 : 5000, ] # Reduce it to save time
    adult$key <- seq_len(nrow(adult))
    stored_nrow <- nrow(adult)

    # When
    aggregated_adult <- aggregate_by_key(adult, key = "key", verbose = verbose)

    # Then
    expect_equal(stored_nrow, nrow(aggregated_adult))
})

test_that("aggregate_by_key: with at least 1 aggregated col should generate nbr_lines col", {
    # Given
    data("adult")
    adult <- adult[1 : 5000, ] # Reduce it to save time
    cols <- c("country", "age")

    # When
    aggregated_adult <- aggregate_by_key(adult[, cols], key = "country", verbose = verbose)

    # Then
    expect_true(all(c("country", "nbr_lines") %in% names(aggregated_adult)))
})

test_that("aggregate_by_key: date are not handled, error should be thrown", {
    # Given
    data("messy_adult")
    messy_adult <- messy_adult[1 : 5000, ]
    messy_adult[["date3"]] <- as.Date(messy_adult[["date3"]], format = "%d-%m-%Y")
    cols <- c("country", "date3")
    sapply(messy_adult, class)
    # When and Then
    expect_error(aggregate_by_key(messy_adult[, cols, with = FALSE], key = "country", verbose = verbose),
    "I can only handle: numeric, integer, factor, logical, character columns.")
})

## aggregateAcolumn
# ------------------
test_that("private function: aggregate_a_column: aggregate a character with more
          than thresh unique values should generate one column count of unique.", {
    # Given
    data("adult")
    setDT(adult)
    unique_keys <- unique(adult[, "country", with = FALSE])
    n_country <- nrow(unique_keys)
    adult[["character_col"]] <- sample(LETTERS, nrow(adult), replace = TRUE)
    cols <- c("country", "character_col")

    # When
    aggregated_col <- aggregate_a_column(adult[, cols, with = FALSE], col = "character_col", key = "country",
    unique_keys = unique_keys, thresh = 2)
    # Then
    expect_equal(n_country, nrow(aggregated_col))
    expect_equal(2, ncol(aggregated_col))
    expect_equal(c("country", "n_unique.character_col"), names(aggregated_col))
})

test_that("private function: aggregate_a_column: aggregate a character with les than
          thresh unique values should generate one column per unique values.", {
    # Given
    data("adult")
    setDT(adult)
    unique_keys <- unique(adult[, "country", with = FALSE])
    n_country <- nrow(unique_keys)
    adult[["character_col"]] <- sample(c("A", "B"), nrow(adult), replace = TRUE)
    cols <- c("country", "character_col")

    # When
    aggregated_col <- aggregate_a_column(adult[, cols, with = FALSE], col = "character_col", key = "country",
    unique_keys = unique_keys, thresh = 2)
    # Then
    expect_equal(n_country, nrow(aggregated_col))
    expect_equal(3, ncol(aggregated_col))
    expect_equal(c("country", "character_col.A", "character_col.B"), names(aggregated_col))
})

test_that("private function: aggregate_a_column: aggregate a logical column. Should create
          1 column with count of true value", {
    # Given
    data("adult")
    setDT(adult)
    unique_keys <- unique(adult[, "country", with = FALSE])
    n_country <- nrow(unique_keys)
    adult[["logical_col"]] <- sample(c(TRUE, FALSE), nrow(adult), replace = TRUE)
    cols <- c("country", "logical_col")

    # When
    aggregated_col <- aggregate_a_column(adult[, cols, with = FALSE], col = "logical_col", key = "country",
    unique_keys = unique_keys)
    # Then
    expect_equal(n_country, nrow(aggregated_col))
    expect_equal(2, ncol(aggregated_col))
    expect_equal(c("country", "nbr_true.logical_col"), names(aggregated_col))
})

test_that("private function: aggregate_a_column: aggregate a numeric column with 2 functions.
          Should create 1 column for each function result", {
    # Given
    data("adult")
    setDT(adult)
    unique_keys <- unique(adult[, "country", with = FALSE])
    n_country <- nrow(unique_keys)
    cols <- c("country", "age")
    functions <- c("mean", "sd")

    # When
    aggregated_col <- aggregate_a_column(adult[, cols, with = FALSE], col = "age", key = "country",
    unique_keys = unique_keys, functions = functions)
    # Then
    expect_equal(n_country, nrow(aggregated_col))
    expect_equal(3, ncol(aggregated_col))
    expect_equal(c("country", "mean.age", "sd.age"), names(aggregated_col))
})

test_that("private function: aggregate_a_column: with duplicated col should return same result on both.", {
    # Given
    data("adult")
    setDT(adult)
    unique_keys <- unique(adult[, "country", with = FALSE])
    n_country <- nrow(unique_keys)
    adult[["country2"]] <- adult[["country"]]
    cols <- c("country", "country2")

    # When
    aggregated_col <- aggregate_a_column(adult[, cols, with = FALSE], col = "country2", key = "country",
    unique_keys = unique_keys, functions = functions)
    # Then
    expect_equal(n_country, nrow(aggregated_col))
    expect_equal(aggregated_col[["country"]], aggregated_col[["country2"]])
})

test_that("private function: aggregate_a_column: should thow error on more than 2 cols set", {
    # Given
    data("adult")
    cols <- c("country", "age", "education")

    # When and Then
    expect_error(aggregate_a_column(adult[, cols], col = "country", key = "country", unique_keys = unique_keys),
    ": data_set should have 2 columns.")
})