import pytest

pd = pytest.importorskip("pandas")

import numpy as np

import dask.array as da
import dask.dataframe as dd
from dask.dataframe._compat import PANDAS_GT_120
from dask.dataframe.utils import assert_eq

_BASE_UFUNCS = [
    "conj",
    "exp",
    "log",
    "log2",
    "log10",
    "log1p",
    "expm1",
    "sqrt",
    "square",
    "sin",
    "cos",
    "tan",
    "arcsin",
    "arccos",
    "arctan",
    "sinh",
    "cosh",
    "tanh",
    "arcsinh",
    "arccosh",
    "arctanh",
    "deg2rad",
    "rad2deg",
    "isfinite",
    "isinf",
    "isnan",
    "signbit",
    "degrees",
    "radians",
    "rint",
    "fabs",
    "sign",
    "absolute",
    "floor",
    "ceil",
    "trunc",
    "logical_not",
    "cbrt",
    "exp2",
    "negative",
    "reciprocal",
    "spacing",
]


@pytest.mark.parametrize(
    "pandas_input",
    [
        pd.Series(np.random.randint(1, 100, size=20)),
        pd.Series(np.abs(np.random.randn(100))),
        pd.DataFrame(
            {
                "A": np.random.randint(1, 100, size=20),
                "B": np.random.randint(1, 100, size=20),
                "C": np.abs(np.random.randn(20)),
            }
        ),
        pd.Series(
            np.random.randint(1, 100, size=20), index=list("abcdefghijklmnopqrst")
        ),
        pd.Series(np.abs(np.random.randn(20)), index=list("abcdefghijklmnopqrst")),
        pd.DataFrame(
            {
                "A": np.random.randint(1, 100, size=20),
                "B": np.random.randint(1, 100, size=20),
                "C": np.abs(np.random.randn(20)),
            },
            index=list("abcdefghijklmnopqrst"),
        ),
    ],
)
@pytest.mark.parametrize("ufunc", _BASE_UFUNCS)
def test_ufunc(pandas_input, ufunc):
    dafunc = getattr(da, ufunc)
    npfunc = getattr(np, ufunc)

    dask_input = dd.from_pandas(pandas_input, 3)
    pandas_type = pandas_input.__class__
    dask_type = dask_input.__class__

    # applying Dask ufunc doesn't trigger computation
    with pytest.warns(None):
        # Some cause warnings (arcsine)
        assert isinstance(dafunc(dask_input), dask_type)
        assert_eq(dafunc(dask_input), npfunc(pandas_input))

        # applying NumPy ufunc is lazy
        if isinstance(npfunc, np.ufunc):
            assert isinstance(npfunc(dask_input), dask_type)
        else:
            assert isinstance(npfunc(dask_input), pandas_type)
        assert_eq(npfunc(dask_input), npfunc(pandas_input))

        # applying Dask ufunc to normal Series triggers computation
        assert isinstance(dafunc(pandas_input), pandas_type)
        assert_eq(dafunc(dask_input), npfunc(pandas_input))

    # Index
    if pandas_input.index.dtype in [object, str]:
        return
    if ufunc in ("logical_not", "signbit", "isnan", "isinf", "isfinite"):
        return

    with pytest.warns(None):
        assert isinstance(dafunc(dask_input.index), dd.Index)
        assert_eq(
            dafunc(dask_input.index),
            npfunc(pandas_input.index),
            check_divisions=ufunc != "spacing",
        )

        # applying NumPy ufunc is lazy
        if isinstance(npfunc, np.ufunc):
            assert isinstance(npfunc(dask_input.index), dd.Index)
        else:
            assert isinstance(npfunc(dask_input.index), pd.Index)

        assert_eq(
            npfunc(dask_input.index),
            npfunc(dask_input.index),
            check_divisions=ufunc != "spacing",
        )

    # applying Dask ufunc to normal Series triggers computation
    with pytest.warns(None):
        # some (da.log) cause warnings
        assert isinstance(dafunc(pandas_input.index), pd.Index)
        assert_eq(dafunc(pandas_input), npfunc(pandas_input))


@pytest.mark.parametrize(
    "ufunc",
    [
        pytest.param(
            "isreal", marks=pytest.mark.filterwarnings("ignore::FutureWarning")
        ),
        "iscomplex",
        pytest.param("real", marks=pytest.mark.filterwarnings("ignore::FutureWarning")),
        pytest.param("imag", marks=pytest.mark.filterwarnings("ignore::FutureWarning")),
        "angle",
        "fix",
        "i0",
        "sinc",
        "nan_to_num",
    ],
)
def test_ufunc_array_wrap(ufunc):
    """
    some np.ufuncs doesn't call __array_wrap__
    (or __array_ufunc__ starting from numpy v.1.13.0), it should work as below

    - da.ufunc(dd.Series) => dd.Series
    - da.ufunc(pd.Series) => np.ndarray
    - np.ufunc(dd.Series) => np.ndarray
    - np.ufunc(pd.Series) => np.ndarray
    """
    if ufunc == "fix":
        pytest.skip("fix calls floor in a way that we do not yet support")

    dafunc = getattr(da, ufunc)
    npfunc = getattr(np, ufunc)

    s = pd.Series(
        np.random.randint(1, 100, size=20), index=list("abcdefghijklmnopqrst")
    )
    ds = dd.from_pandas(s, 3)

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(ds), dd.Series)
    assert_eq(dafunc(ds), pd.Series(npfunc(s), index=s.index))

    assert isinstance(npfunc(ds), np.ndarray)
    np.testing.assert_equal(npfunc(ds), npfunc(s))

    assert isinstance(dafunc(s), np.ndarray)
    np.testing.assert_array_equal(dafunc(s), npfunc(s))

    df = pd.DataFrame(
        {
            "A": np.random.randint(1, 100, size=20),
            "B": np.random.randint(1, 100, size=20),
            "C": np.abs(np.random.randn(20)),
        },
        index=list("abcdefghijklmnopqrst"),
    )
    ddf = dd.from_pandas(df, 3)

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(ddf), dd.DataFrame)
    # result may be read-only ndarray
    exp = pd.DataFrame(npfunc(df).copy(), columns=df.columns, index=df.index)
    assert_eq(dafunc(ddf), exp)

    assert isinstance(npfunc(ddf), np.ndarray)
    np.testing.assert_array_equal(npfunc(ddf), npfunc(df))

    assert isinstance(dafunc(df), np.ndarray)
    np.testing.assert_array_equal(dafunc(df), npfunc(df))


_UFUNCS_2ARG = [
    "logaddexp",
    "logaddexp2",
    "arctan2",
    "hypot",
    "copysign",
    "nextafter",
    pytest.param("ldexp", marks=[pytest.mark.filterwarnings("ignore::RuntimeWarning")]),
    pytest.param("fmod", marks=[pytest.mark.filterwarnings("ignore::RuntimeWarning")]),
    "logical_and",
    "logical_or",
    "logical_xor",
    "maximum",
    "minimum",
    "fmax",
    "fmin",
    "greater",
    "greater_equal",
    "less",
    "less_equal",
    "not_equal",
    "equal",
    "logical_or",
    "logical_and",
    "logical_xor",
]


@pytest.mark.parametrize("ufunc", _UFUNCS_2ARG)
@pytest.mark.parametrize(
    "make_pandas_input",
    [
        lambda: pd.Series(np.random.randint(1, 100, size=20)),
        lambda: pd.DataFrame(
            np.random.randint(1, 100, size=(20, 2)), columns=["A", "B"]
        ),
    ],
)
def test_ufunc_with_2args(ufunc, make_pandas_input):

    dafunc = getattr(da, ufunc)
    npfunc = getattr(np, ufunc)

    pandas1 = make_pandas_input()
    pandas2 = make_pandas_input()

    dask1 = dd.from_pandas(pandas1, 3)
    dask2 = dd.from_pandas(pandas2, 4)

    pandas_type = pandas1.__class__
    dask_type = dask1.__class__

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(dask1, dask2), dask_type)
    assert_eq(dafunc(dask1, dask2), npfunc(pandas1, pandas2))

    # should be fine with pandas as a second arg, too
    assert isinstance(dafunc(dask1, pandas2), dask_type)
    assert_eq(dafunc(dask1, pandas2), npfunc(pandas1, pandas2))

    # applying NumPy ufunc is lazy
    if isinstance(npfunc, np.ufunc):
        assert isinstance(npfunc(dask1, dask2), dask_type)
        assert isinstance(npfunc(dask1, pandas2), dask_type)
    else:
        assert isinstance(npfunc(dask1, dask2), pandas_type)
        assert isinstance(npfunc(dask1, pandas2), pandas_type)

    assert_eq(npfunc(dask1, dask2), npfunc(pandas1, pandas2))
    assert_eq(npfunc(dask1, pandas2), npfunc(pandas1, pandas2))

    # applying Dask ufunc to normal Series triggers computation
    assert isinstance(dafunc(pandas1, pandas2), pandas_type)
    assert_eq(dafunc(pandas1, pandas2), npfunc(pandas1, pandas2))


@pytest.mark.parametrize(
    "pandas,min,max",
    [
        (pd.Series(np.random.randint(1, 100, size=20)), 5, 50),
        (
            pd.DataFrame(np.random.randint(1, 100, size=(20, 2)), columns=["A", "B"]),
            5.5,
            40.5,
        ),
    ],
)
def test_clip(pandas, min, max):

    dask = dd.from_pandas(pandas, 3)
    pandas_type = pandas.__class__
    dask_type = dask.__class__

    # clip internally calls dd.Series.clip

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(da.clip(dask, min, max), dask_type)
    assert_eq(da.clip(dask, min, max), np.clip(pandas, min, max))

    # applying Numpy ufunc doesn't trigger computation
    assert isinstance(np.clip(dask, min, max), dask_type)
    assert_eq(np.clip(dask, min, max), np.clip(pandas, min, max))

    # applying Dask ufunc to normal pandas objects triggers computation
    assert isinstance(da.clip(pandas, min, max), pandas_type)
    assert_eq(da.clip(pandas, min, max), np.clip(pandas, min, max))


@pytest.mark.parametrize("ufunc", _BASE_UFUNCS)
def test_frame_ufunc_out(ufunc):
    npfunc = getattr(np, ufunc)
    dafunc = getattr(da, ufunc)

    input_matrix = np.random.randint(1, 100, size=(20, 2))

    df = pd.DataFrame(input_matrix, columns=["A", "B"])
    ddf = dd.from_pandas(df, 3)
    df_out = pd.DataFrame(np.random.randint(1, 100, size=(20, 2)), columns=["Y", "Z"])
    ddf_out_np = dd.from_pandas(df_out, 3)
    ddf_out_da = dd.from_pandas(df_out, 3)

    with pytest.warns(None):
        npfunc(ddf, out=ddf_out_np)
        dafunc(ddf, out=ddf_out_da)
        assert_eq(ddf_out_np, ddf_out_da)

    with pytest.warns(None):
        expected = pd.DataFrame(npfunc(input_matrix), columns=["A", "B"])
        assert_eq(ddf_out_np, expected)


def test_frame_2ufunc_out():
    input_matrix = np.random.randint(1, 100, size=(20, 2))

    df = pd.DataFrame(input_matrix, columns=["A", "B"])
    ddf = dd.from_pandas(df, 3)

    # column number mismatch
    df_out = pd.DataFrame(
        np.random.randint(1, 100, size=(20, 3)), columns=["X", "Y", "Z"]
    )
    ddf_out = dd.from_pandas(df_out, 3)

    with pytest.raises(ValueError):
        np.sin(ddf, out=ddf_out)

    # types mismatch
    ddf_out = dd.from_pandas(pd.Series([0]), 1)
    with pytest.raises(TypeError):
        np.sin(ddf, out=ddf_out)

    df_out = pd.DataFrame(np.random.randint(1, 100, size=(20, 2)), columns=["X", "Y"])
    ddf_out = dd.from_pandas(df_out, 3)

    np.sin(ddf, out=ddf_out)
    np.add(ddf_out, 10, out=ddf_out)

    expected = pd.DataFrame(np.sin(input_matrix) + 10, columns=["A", "B"])

    assert_eq(ddf_out, expected)


@pytest.mark.parametrize(
    "arg1",
    [
        pd.Series(np.abs(np.random.randn(100))),
        pd.DataFrame(
            {
                "A": np.random.randint(1, 100, size=20),
                "B": np.random.randint(1, 100, size=20),
                "C": np.abs(np.random.randn(20)),
            }
        ),
    ],
)
@pytest.mark.parametrize("arg2", [2, dd.from_pandas(pd.Series([0]), 1).sum()])
@pytest.mark.parametrize("ufunc", _UFUNCS_2ARG)
def test_mixed_types(ufunc, arg1, arg2):
    npfunc = getattr(np, ufunc)
    dafunc = getattr(da, ufunc)

    dask = dd.from_pandas(arg1, 3)

    pandas_type = arg1.__class__
    dask_type = dask.__class__

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(dask, arg2), dask_type)
    assert_eq(dafunc(dask, arg2), npfunc(dask, arg2))

    # applying NumPy ufunc is lazy
    assert isinstance(npfunc(dask, arg2), dask_type)
    assert_eq(npfunc(dask, arg2), npfunc(arg1, arg2))

    # applying Dask ufunc to normal Series triggers computation
    assert isinstance(dafunc(arg1, arg2), pandas_type)
    assert_eq(dafunc(arg1, arg2), npfunc(arg1, arg2))

    # swapping arguments

    # first parameter of ldexp should be array-like
    if ufunc == "ldexp":
        return

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(arg2, dask), dask_type)
    assert_eq(dafunc(arg2, dask), npfunc(arg2, dask))

    # applying NumPy ufunc is lazy
    assert isinstance(npfunc(arg2, dask), dask_type)
    assert_eq(npfunc(arg2, dask), npfunc(arg2, dask))

    # applying Dask ufunc to normal Series triggers computation
    assert isinstance(dafunc(arg2, arg1), pandas_type)
    assert_eq(dafunc(arg2, arg1), npfunc(arg2, arg1))


@pytest.mark.parametrize("ufunc", _UFUNCS_2ARG)
@pytest.mark.parametrize(
    "pandas,darray",
    [
        (
            pd.Series(np.random.randint(1, 100, size=(100,))),
            da.from_array(np.random.randint(1, 100, size=(100,)), chunks=(50,)),
        ),
        (
            pd.DataFrame(np.random.randint(1, 100, size=(20, 2)), columns=["A", "B"]),
            da.from_array(np.random.randint(1, 100, size=(20, 2)), chunks=(10, 2)),
        ),
    ],
)
def test_2args_with_array(ufunc, pandas, darray):
    dafunc = getattr(da, ufunc)
    npfunc = getattr(np, ufunc)

    dask = dd.from_pandas(pandas, 2)
    dask_type = dask.__class__

    # applying Dask ufunc doesn't trigger computation
    assert isinstance(dafunc(dask, darray), dask_type)
    assert isinstance(dafunc(darray, dask), dask_type)

    np.testing.assert_array_equal(
        dafunc(dask, darray).compute().values, npfunc(pandas.values, darray).compute()
    )

    # applying NumPy ufunc is lazy
    assert isinstance(npfunc(dask, darray), dask_type)
    assert isinstance(npfunc(darray, dask), dask_type)

    np.testing.assert_array_equal(
        npfunc(dask, darray).compute().values, npfunc(pandas.values, darray.compute())
    )
    np.testing.assert_array_equal(
        npfunc(darray, dask).compute().values, npfunc(darray.compute(), pandas.values)
    )


@pytest.mark.parametrize("redfunc", ["sum", "prod", "min", "max", "mean"])
@pytest.mark.parametrize("ufunc", _BASE_UFUNCS)
@pytest.mark.parametrize(
    "pandas",
    [
        pd.Series(np.abs(np.random.randn(100))),
        pd.DataFrame(
            {
                "A": np.random.randint(1, 100, size=20),
                "B": np.random.randint(1, 100, size=20),
                "C": np.abs(np.random.randn(20)),
            }
        ),
    ],
)
def test_ufunc_with_reduction(redfunc, ufunc, pandas):
    dask = dd.from_pandas(pandas, 3)

    np_redfunc = getattr(np, redfunc)
    np_ufunc = getattr(np, ufunc)

    if (
        PANDAS_GT_120
        and (redfunc == "prod")
        and ufunc in ["conj", "square", "negative", "absolute"]
        and isinstance(pandas, pd.DataFrame)
    ):
        # TODO(pandas) follow pandas behaviour?
        # starting with pandas 1.2.0, the ufunc is applied column-wise, and therefore
        # applied on the integer columns separately, overflowing for those columns
        # (instead of being applied on 2D ndarray that was converted to float)
        pytest.xfail("'prod' overflowing with integer columns in pandas 1.2.0")

    with pytest.warns(None):
        assert isinstance(np_redfunc(dask), (dd.DataFrame, dd.Series, dd.core.Scalar))
        assert_eq(np_redfunc(np_ufunc(dask)), np_redfunc(np_ufunc(pandas)))


@pytest.mark.parametrize(
    "pandas",
    [
        pd.Series(np.random.randint(1, 100, size=100)),
        pd.DataFrame(
            {
                "A": np.random.randint(1, 100, size=20),
                "B": np.random.randint(1, 100, size=20),
                "C": np.abs(np.random.randn(20)),
            }
        ),
    ],
)
@pytest.mark.parametrize("scalar", [15, 16.4, np.int64(15), np.float64(16.4)])
def test_ufunc_numpy_scalar_comparison(pandas, scalar):
    # Regression test for issue #3392

    dask_compare = scalar >= dd.from_pandas(pandas, npartitions=3)
    pandas_compare = scalar >= pandas

    assert_eq(dask_compare, pandas_compare)