context("dimSums can collapse arbitrary cube dimensions")

mr_x_mr <- loadCube("cubes/full-cube.json")
mr_x_mr_dims <- dimnames(mr_x_mr)

cat_x_mr <- loadCube("cubes/selected-crosstab-array-last.json")
cat_x_mr_dims <- dimnames(cat_x_mr)
# drop the "No Data" category in fruit
cat_x_mr_dims$fruit <- cat_x_mr_dims$fruit[cat_x_mr_dims$fruit != "No Data"]

cat_x_mr_x_mr <- loadCube("cubes/cat-x-mr-x-mr.json")
cat_x_mr_x_mr_dims <- dimnames(cat_x_mr_x_mr)
# drop the "No Data" category in animal
cat_x_mr_x_mr_dims$animal <- cat_x_mr_x_mr_dims$animal[cat_x_mr_x_mr_dims$animal != "No Data"]

test_that("dimSums(mr_x_mr)", {
    expect_equivalent(
        as.array(dimSums(mr_x_mr, 2)),
        # note this is a 1-d output, condensed here for space
        cubify(
            13068.9587689331, 20954.7013096216,
            6650.64488216886, 9401.03672837049,
            22030.3153523541, 28510.504501164,
            8638.7700564883, 19611.6905531398,
            17341.5840147774, 19639.4826050037,
            14457.4508268457, 9384.25096497542,
            12949.8527858053, 21053.7004299784,
            16873.4231466009, 8805.6717582668,
            9704.93506952685, 11520.0756377588,
            22750.5942352494, 13060.1623918338,
            17343.0337039009, 9312.51198396474,
            15122.0959929982, 23391.9651645574,
            25919.8945071622, 18011.9795313661,
            24395.8900036815,
            dims = mr_x_mr_dims["letters"]
        )
    )

    expect_equivalent(
        as.array(dimSums(mr_x_mr, 1)),
        cubify(
            19935.9325550077,
            8815.06513916879,
            12015.5649554376,
            7967.67530412789,
            25379.6151957024,
            5130.6201930451,
            604.39533293168,
            1757.81059587395,
            3140.06957091528,
            dims = mr_x_mr_dims["offal"]
        )
    )

    # check against a univariate cube with the same data to confirm this is
    # actually the univariate, unconditional margin
    mr_x_self <- loadCube("cubes/natrep-cube.json")
    dimsum_cube <- dimSums(mr_x_mr, 1)
    expect_equal(
        dimsum_cube@dims,
        mr_x_self@dims
    )
    expect_equal(
        dimsum_cube@arrays,
        mr_x_self@arrays
    )
})

test_that("dimSums(mr_x_mr) proportions", {
    first_univariate <- dimSums(mr_x_mr, 2)
    expect_equal(
        as.array(prop.table(first_univariate)),
        # note this is a 1-d output, condensed here for space
        cubify(
            0.343005652705866, 0.526457708906739,
            0.170061597715756, 0.248137953864616,
            0.575295983274584, 0.722413408984012,
            0.215158236529976, 0.497733964220733,
            0.468887081108057, 0.529088607024214,
            0.381627827326823, 0.238897143997144,
            0.348351059638057, 0.561748349106473,
            0.444069716726853, 0.223645636273538,
            0.247956569586389, 0.292280096307629,
            0.572613085059897, 0.327067092416562,
            0.437620721993495, 0.219891051379945,
            0.356885537375432, 0.548258129233669,
            0.641905291100413, 0.457714901465151,
            0.579056506939286,
            dims = mr_x_mr_dims["letters"]
        )
    )

    expect_equal(
        as.array(prop.table(first_univariate, 1)),
        cubify(1, dims = mr_x_mr_dims["letters"])
    )

    second_univariate <- dimSums(mr_x_mr, 1)
    expect_equal(
        as.array(prop.table(second_univariate)),
        cubify(
            0.463523266541455,
            0.204955939573381,
            0.279369620764814,
            0.18525358036403,
            0.590092392548861,
            0.119290222551763,
            0.0140525805968646,
            0.0408702280223159,
            0.073008639082322,
            dims = mr_x_mr_dims["offal"]
        )
    )
    expect_equal(
        as.array(prop.table(second_univariate, 1)),
        cubify(1, dims = mr_x_mr_dims["offal"])
    )

    # check against a univariate cube with the same data to confirm this is
    # actually the univariate, unconditional margin
    mr_x_self <- loadCube("cubes/natrep-cube.json")

    expect_equal(
        prop.table(dimSums(mr_x_mr, 1)),
        prop.table(mr_x_self)
    )
    expect_equal(
        prop.table(dimSums(mr_x_mr, 1), 1),
        prop.table(mr_x_self, 1)
    )
})

test_that("dimSums(cat_x_mr)", {
    expect_equivalent(
        as.array(dimSums(cat_x_mr, 2)),
        cubify(
            22.9672704148528, 45.7789165449064, 86.9728287914322,
            dims = cat_x_mr_dims["zoo"]
        )
    )
    expect_equivalent(
        as.array(prop.table(dimSums(cat_x_mr, 2))),
        cubify(
            0.138355225153969, 0.322687424815186, 0.480529823991341,
            dims = cat_x_mr_dims["zoo"]
        )
    )

    expect_equivalent(
        as.array(prop.table(dimSums(cat_x_mr, 1))),
        cubify(
            0.334028222315447, 0.665971777684553,
            dims = list(fruit = list("rambutan", "satsuma"))
        )
    )
})

test_that("dimSums(cat_x_mr_x_mr)", {
    expect_equivalent(
        as.array(dimSums(cat_x_mr_x_mr, 1)),
        cubify(
            10000, 10000,
            dims = cat_x_mr_x_mr_dims["animal"]
        )
    )

    expect_equivalent(
        as.array(dimSums(cat_x_mr_x_mr, 2)),
        cubify(
            6970, 7029, 6940,
            dims = cat_x_mr_x_mr_dims["opinion_mr"]
        )
    )

    expect_equivalent(
        as.array(dimSums(cat_x_mr_x_mr, 3)),
        cubify(
            3867, 7002,
            dims = cat_x_mr_x_mr_dims["feeling_mr"]
        )
    )
})

test_that("dimSums() prevents use with non-count cubes", {
    mean_cube <- loadCube("cubes/mean-age-food_groups-x-pasta.json")
    expect_error(
        dimSums(mean_cube, 1),
        paste0(
            "You can't use CrunchCubes with measures other than count. ",
            "The cube you provided included measures: mean"
        )
    )
})

test_that("dimSums() errors nicely when it margin is wrong", {
    small_cube <- loadCube("cubes/univariate-categorical.json")
    expect_error(
        dimSums(small_cube, 2),
        "Margin 2 exceeds Cube's number of dimensions (1)",
        fixed = TRUE
    )
})

test_that("only_cube_count() prevents use with non-count cubes", {
    mean_cube <- loadCube("cubes/mean-age-food_groups-x-pasta.json")
    expect_error(
        only_count_cube(mean_cube),
        paste0(
            "You can't use CrunchCubes with measures other than count. ",
            "The cube you provided included measures: mean"
        )
    )
})

with_test_authentication({
    ds <- newDatasetFromFixture("apidocs")

    test_that("dimSums(~x+y, 1) == dimSums(~x)", {
        bivariate_cube <- crtabs(~ allpets + q1, data = ds)
        univariate_allpets <- crtabs(~allpets, data = ds)
        univariate_q1 <- crtabs(~q1, data = ds)
        expect_equal(dimSums(bivariate_cube, 2)@dims, univariate_q1@dims)
        expect_equal(dimSums(bivariate_cube, 2)@arrays, univariate_q1@arrays)
        expect_equal(dimSums(bivariate_cube, 1)@dims, univariate_allpets@dims)
        expect_equal(dimSums(bivariate_cube, 1)@arrays, univariate_allpets@arrays)


        trivariate_cube <- crtabs(~ country + allpets + q1, data = ds)
        expect_equal(
            dimSums(trivariate_cube, c(2, 3))@dims,
            crtabs(~ allpets + q1, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, c(2, 3))@arrays,
            crtabs(~ allpets + q1, data = ds)@arrays
        )

        expect_equal(
            dimSums(trivariate_cube, c(1, 3))@dims,
            crtabs(~ country + q1, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, c(1, 3))@arrays,
            crtabs(~ country + q1, data = ds)@arrays
        )

        expect_equal(
            dimSums(trivariate_cube, c(1, 2))@dims,
            crtabs(~ country + allpets, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, c(1, 2))@arrays,
            crtabs(~ country + allpets, data = ds)@arrays
        )

        expect_equal(
            dimSums(trivariate_cube, 3)@dims,
            crtabs(~q1, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, 3)@arrays,
            crtabs(~q1, data = ds)@arrays
        )

        expect_equal(
            dimSums(trivariate_cube, 1)@dims,
            crtabs(~country, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, 1)@arrays,
            crtabs(~country, data = ds)@arrays
        )

        expect_equal(
            dimSums(trivariate_cube, 2)@dims,
            crtabs(~allpets, data = ds)@dims
        )
        expect_equal(
            dimSums(trivariate_cube, 2)@arrays,
            crtabs(~allpets, data = ds)@arrays
        )
    })

    test_that("only_count_cube() prevents use with non-count cubes", {
        multi_measures <- crtabs(list(mean(ndogs), max(ndogs)) ~ allpets + q1, data = ds)
        expect_error(
            only_count_cube(multi_measures),
            paste0(
                "You can't use CrunchCubes with measures other than count. ",
                "The cube you provided included measures: max and mean"
            )
        )
    })
})