# Licensed under a 3-clause BSD style license - see LICENSE.rst

import os

import numpy as np
import pytest

from astropy.io.misc.hdf5 import meta_path
from astropy.table import Column, QTable, Table
from astropy.table.table_helpers import simple_table
from astropy.units import allclose as quantity_allclose
from astropy.units.quantity import QuantityInfo
from astropy.utils.compat import NUMPY_LT_1_22
from astropy.utils.compat.optional_deps import HAS_H5PY
from astropy.utils.data import get_pkg_data_filename
from astropy.utils.exceptions import AstropyUserWarning
from astropy.utils.misc import _NOT_OVERWRITING_MSG_MATCH

if HAS_H5PY:
    import h5py

from astropy.io.tests.mixin_columns import compare_attrs, mixin_cols, serialized_names

# HDF5 does not support object dtype (since it stores binary representations).
unsupported_cols = {
    name: col
    for name, col in mixin_cols.items()
    if (isinstance(col, np.ndarray) and col.dtype.kind == "O")
}
mixin_cols = {
    name: col for name, col in mixin_cols.items() if name not in unsupported_cols
}

ALL_DTYPES = [
    np.uint8,
    np.uint16,
    np.uint32,
    np.uint64,
    np.int8,
    np.int16,
    np.int32,
    np.int64,
    np.float32,
    np.float64,
    np.bool_,
    "|S3",
]


def _default_values(dtype):
    if dtype == np.bool_:
        return [0, 1, 1]
    elif dtype == "|S3":
        return [b"abc", b"def", b"ghi"]
    else:
        return [1, 2, 3]


@pytest.fixture
def home_is_tmpdir(monkeypatch, tmp_path):
    """
    Pytest fixture to run a test case with tilde-prefixed paths.

    In the tilde-path case, environment variables are temporarily
    modified so that '~' resolves to the temp directory.
    """
    # For Unix
    monkeypatch.setenv("HOME", str(tmp_path))
    # For Windows
    monkeypatch.setenv("USERPROFILE", str(tmp_path))


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_nopath(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))

    with pytest.warns(
        UserWarning, match="table path was not set via the path= argument"
    ):
        t1.write(test_file)

    t1 = Table.read(test_file, path="__astropy_table__")


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_nopath_nonempty(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))

    t1.write(test_file, path="bubu")

    with pytest.raises(ValueError) as exc:
        t1.write(test_file, append=True)

    assert "table path should always be set via the path=" in exc.value.args[0]


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_notable_nopath(tmp_path):
    test_file = tmp_path / "test.hdf5"
    h5py.File(test_file, "w").close()  # create empty file
    with pytest.raises(ValueError, match="no table found in HDF5 group /"):
        Table.read(test_file, path="/", format="hdf5")


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_nopath(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")
    t2 = Table.read(test_file)

    assert np.all(t1["a"] == t2["a"])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_nopath_multi_tables(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")
    t1.write(test_file, path="the_table_but_different", append=True, overwrite=True)
    with pytest.warns(
        AstropyUserWarning, match=r"path= was not specified but multiple tables"
    ):
        t2 = Table.read(test_file)

    assert np.all(t1["a"] == t2["a"])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_invalid_path(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    with pytest.raises(ValueError) as exc:
        t1.write(test_file, path="test/")
    assert exc.value.args[0] == "table path should end with table name, not /"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_invalid_path(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")
    with pytest.raises(OSError) as exc:
        Table.read(test_file, path="test/")
    assert exc.value.args[0] == "Path test/ does not exist"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_missing_group(tmp_path):
    test_file = tmp_path / "test.hdf5"
    h5py.File(test_file, "w").close()  # create empty file
    with pytest.raises(OSError) as exc:
        Table.read(test_file, path="test/path/table")
    assert exc.value.args[0] == "Path test/path/table does not exist"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_missing_table(tmp_path):
    test_file = tmp_path / "test.hdf5"
    with h5py.File(test_file, "w") as f:
        f.create_group("test").create_group("path")
    with pytest.raises(OSError) as exc:
        Table.read(test_file, path="test/path/table")
    assert exc.value.args[0] == "Path test/path/table does not exist"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_missing_group_fileobj(tmp_path):
    test_file = tmp_path / "test.hdf5"
    with h5py.File(test_file, "w") as f:
        with pytest.raises(OSError) as exc:
            Table.read(f, path="test/path/table")
        assert exc.value.args[0] == "Path test/path/table does not exist"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_simple(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")
    t2 = Table.read(test_file, path="the_table")
    assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing_table(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")
    with pytest.raises(OSError) as exc:
        t1.write(test_file, path="the_table", append=True)
    assert exc.value.args[0] == "Table the_table already exists"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_memory():
    with h5py.File("test", "w", driver="core", backing_store=False) as output_file:
        t1 = Table()
        t1.add_column(Column(name="a", data=[1, 2, 3]))
        t1.write(output_file, path="the_table")
        t2 = Table.read(output_file, path="the_table")
        assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing(tmp_path):
    test_file = tmp_path / "test.hdf5"
    h5py.File(test_file, "w").close()  # create empty file
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    with pytest.raises(OSError, match=_NOT_OVERWRITING_MSG_MATCH):
        t1.write(test_file, path="the_table")


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing_overwrite(tmp_path):
    test_file = tmp_path / "test.hdf5"
    h5py.File(test_file, "w").close()  # create empty file
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table", overwrite=True)
    t2 = Table.read(test_file, path="the_table")
    assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing_append(tmp_path):
    test_file = tmp_path / "test.hdf5"
    h5py.File(test_file, "w").close()  # create empty file
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table_1", append=True)
    t1.write(test_file, path="the_table_2", append=True)
    t2 = Table.read(test_file, path="the_table_1")
    assert np.all(t2["a"] == [1, 2, 3])
    t3 = Table.read(test_file, path="the_table_2")
    assert np.all(t3["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing_append_groups(tmp_path):
    test_file = tmp_path / "test.hdf5"
    with h5py.File(test_file, "w") as f:
        f.create_group("test_1")
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="test_1/the_table_1", append=True)
    t1.write(test_file, path="test_2/the_table_2", append=True)
    t2 = Table.read(test_file, path="test_1/the_table_1")
    assert np.all(t2["a"] == [1, 2, 3])
    t3 = Table.read(test_file, path="test_2/the_table_2")
    assert np.all(t3["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_existing_append_overwrite(tmp_path):
    test_file = tmp_path / "test.hdf5"
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="table1")
    t1.write(test_file, path="table2", append=True)
    t1v2 = Table()
    t1v2.add_column(Column(name="a", data=[4, 5, 6]))
    with pytest.raises(OSError) as exc:
        t1v2.write(test_file, path="table1", append=True)
    assert exc.value.args[0] == "Table table1 already exists"
    t1v2.write(test_file, path="table1", append=True, overwrite=True)
    t2 = Table.read(test_file, path="table1")
    assert np.all(t2["a"] == [4, 5, 6])
    t3 = Table.read(test_file, path="table2")
    assert np.all(t3["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_fileobj(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="the_table")

    import h5py

    with h5py.File(test_file, "r") as input_file:
        t2 = Table.read(input_file, path="the_table")
        assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_filobj_path(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="path/to/data/the_table")

    import h5py

    with h5py.File(test_file, "r") as input_file:
        t2 = Table.read(input_file, path="path/to/data/the_table")
        assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_filobj_group_path(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.write(test_file, path="path/to/data/the_table")

    import h5py

    with h5py.File(test_file, "r") as input_file:
        t2 = Table.read(input_file["path/to"], path="data/the_table")
        assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_wrong_fileobj():
    class FakeFile:
        def read(self):
            pass

    f = FakeFile()

    with pytest.raises(TypeError, match="h5py can only open regular files"):
        Table.read(f, format="hdf5")


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_fileobj(tmp_path):
    test_file = tmp_path / "test.hdf5"

    import h5py

    with h5py.File(test_file, "w") as output_file:
        t1 = Table()
        t1.add_column(Column(name="a", data=[1, 2, 3]))
        t1.write(output_file, path="the_table")

    t2 = Table.read(test_file, path="the_table")
    assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_create_dataset_kwargs(tmp_path):
    test_file = tmp_path / "test.hdf5"
    the_path = "the_table"

    import h5py

    with h5py.File(test_file, "w") as output_file:
        t1 = Table()
        t1.add_column(Column(name="a", data=[1, 2, 3]))
        t1.write(output_file, path=the_path, maxshape=(None,))

    # A roundabout way of checking this, but the table created above should be
    # resizable if the kwarg was passed through successfully
    t2 = Table()
    t2.add_column(Column(name="a", data=[4, 5]))
    with h5py.File(test_file, "a") as output_file:
        output_file[the_path].resize((len(t1) + len(t2),))
        output_file[the_path][len(t1) :] = t2.as_array()

    t3 = Table.read(test_file, path="the_table")
    assert np.all(t3["a"] == [1, 2, 3, 4, 5])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_filobj_group(tmp_path):
    test_file = tmp_path / "test.hdf5"

    import h5py

    with h5py.File(test_file, "w") as output_file:
        t1 = Table()
        t1.add_column(Column(name="a", data=[1, 2, 3]))
        t1.write(output_file, path="path/to/data/the_table")

    t2 = Table.read(test_file, path="path/to/data/the_table")
    assert np.all(t2["a"] == [1, 2, 3])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_write_wrong_type():
    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    with pytest.raises(TypeError) as exc:
        t1.write(1212, path="path/to/data/the_table", format="hdf5")
    assert (
        exc.value.args[0] == "output should be a string or an h5py File or Group object"
    )


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
@pytest.mark.parametrize("dtype", ALL_DTYPES)
def test_preserve_single_dtypes(tmp_path, dtype):
    test_file = tmp_path / "test.hdf5"

    values = _default_values(dtype)

    t1 = Table()
    t1.add_column(Column(name="a", data=np.array(values, dtype=dtype)))
    t1.write(test_file, path="the_table")

    t2 = Table.read(test_file, path="the_table")

    assert np.all(t2["a"] == values)
    assert t2["a"].dtype == dtype


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_preserve_all_dtypes(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()

    for dtype in ALL_DTYPES:
        values = _default_values(dtype)
        t1.add_column(Column(name=str(dtype), data=np.array(values, dtype=dtype)))

    t1.write(test_file, path="the_table")

    t2 = Table.read(test_file, path="the_table")

    for dtype in ALL_DTYPES:
        values = _default_values(dtype)
        assert np.all(t2[str(dtype)] == values)
        assert t2[str(dtype)].dtype == dtype


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_preserve_meta(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))

    t1.meta["a"] = 1
    t1.meta["b"] = "hello"
    t1.meta["c"] = 3.14159
    t1.meta["d"] = True
    t1.meta["e"] = np.array([1, 2, 3])

    t1.write(test_file, path="the_table")

    t2 = Table.read(test_file, path="the_table")

    for key in t1.meta:
        assert np.all(t1.meta[key] == t2.meta[key])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_preserve_serialized(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1["a"] = Column(data=[1, 2, 3], unit="s")
    t1["a"].meta["a0"] = "A0"
    t1["a"].meta["a1"] = {"a1": [0, 1]}
    t1["a"].format = "7.3f"
    t1["a"].description = "A column"
    t1.meta["b"] = 1
    t1.meta["c"] = {"c0": [0, 1]}

    t1.write(test_file, path="the_table", serialize_meta=True, overwrite=True)

    t2 = Table.read(test_file, path="the_table")

    assert t1["a"].unit == t2["a"].unit
    assert t1["a"].format == t2["a"].format
    assert t1["a"].description == t2["a"].description
    assert t1["a"].meta == t2["a"].meta
    assert t1.meta == t2.meta

    # Check that the meta table is fixed-width bytes (see #11299)
    h5 = h5py.File(test_file, "r")
    meta_lines = h5[meta_path("the_table")]
    assert meta_lines.dtype.kind == "S"


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_preserve_serialized_old_meta_format():
    """Test the old meta format

    Only for some files created prior to v4.0, in compatibility mode.
    """
    test_file = get_pkg_data_filename("data/old_meta_example.hdf5")

    t1 = Table()
    t1["a"] = Column(data=[1, 2, 3], unit="s")
    t1["a"].meta["a0"] = "A0"
    t1["a"].meta["a1"] = {"a1": [0, 1]}
    t1["a"].format = "7.3f"
    t1["a"].description = "A column"
    t1.meta["b"] = 1
    t1.meta["c"] = {"c0": [0, 1]}

    t2 = Table.read(test_file, path="the_table")

    assert t1["a"].unit == t2["a"].unit
    assert t1["a"].format == t2["a"].format
    assert t1["a"].description == t2["a"].description
    assert t1["a"].meta == t2["a"].meta
    assert t1.meta == t2.meta


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_preserve_serialized_in_complicated_path(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1["a"] = Column(data=[1, 2, 3], unit="s")
    t1["a"].meta["a0"] = "A0"
    t1["a"].meta["a1"] = {"a1": [0, 1]}
    t1["a"].format = "7.3f"
    t1["a"].description = "A column"
    t1.meta["b"] = 1
    t1.meta["c"] = {"c0": [0, 1]}

    t1.write(
        test_file,
        path="the_table/complicated/path",
        serialize_meta=True,
        overwrite=True,
    )

    t2 = Table.read(test_file, path="the_table/complicated/path")

    assert t1["a"].format == t2["a"].format
    assert t1["a"].unit == t2["a"].unit
    assert t1["a"].description == t2["a"].description
    assert t1["a"].meta == t2["a"].meta
    assert t1.meta == t2.meta


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_metadata_very_large(tmp_path):
    """Test that very large datasets work, now!"""
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1["a"] = Column(data=[1, 2, 3], unit="s")
    t1["a"].meta["a0"] = "A0"
    t1["a"].meta["a1"] = {"a1": [0, 1]}
    t1["a"].format = "7.3f"
    t1["a"].description = "A column"
    t1.meta["b"] = 1
    t1.meta["c"] = {"c0": [0, 1]}
    t1.meta["meta_big"] = "0" * (2**16 + 1)
    t1.meta["meta_biggerstill"] = "0" * (2**18)

    t1.write(test_file, path="the_table", serialize_meta=True, overwrite=True)

    t2 = Table.read(test_file, path="the_table")

    assert t1["a"].unit == t2["a"].unit
    assert t1["a"].format == t2["a"].format
    assert t1["a"].description == t2["a"].description
    assert t1["a"].meta == t2["a"].meta
    assert t1.meta == t2.meta


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_skip_meta(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))

    t1.meta["a"] = 1
    t1.meta["b"] = "hello"
    t1.meta["c"] = 3.14159
    t1.meta["d"] = True
    t1.meta["e"] = np.array([1, 2, 3])
    t1.meta["f"] = str

    wtext = (
        f"Attribute `f` of type {type(t1.meta['f'])} cannot be written to HDF5 files -"
        " skipping"
    )
    with pytest.warns(AstropyUserWarning, match=wtext) as w:
        t1.write(test_file, path="the_table")
    assert len(w) == 1


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_fail_meta_serialize(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t1 = Table()
    t1.add_column(Column(name="a", data=[1, 2, 3]))
    t1.meta["f"] = str

    with pytest.raises(Exception) as err:
        t1.write(test_file, path="the_table", serialize_meta=True)
    assert "cannot represent an object" in str(err.value)
    assert "<class 'str'>" in str(err.value)


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_h5py_objects(tmp_path):
    # Regression test - ensure that Datasets are recognized automatically

    test_file = tmp_path / "test.hdf5"

    import h5py

    with h5py.File(test_file, "w") as output_file:
        t1 = Table()
        t1.add_column(Column(name="a", data=[1, 2, 3]))
        t1.write(output_file, path="the_table")

    f = h5py.File(test_file, mode="r")

    t2 = Table.read(f, path="the_table")
    assert np.all(t2["a"] == [1, 2, 3])

    t3 = Table.read(f["/"], path="the_table")
    assert np.all(t3["a"] == [1, 2, 3])

    t4 = Table.read(f["the_table"])
    assert np.all(t4["a"] == [1, 2, 3])

    f.close()  # don't leave the file open


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_unicode_to_hdf5(tmp_path):
    test_file = tmp_path / "test.hdf5"

    t = Table()
    t["p"] = ["a", "b", "c"]
    t["q"] = [1, 2, 3]
    t["r"] = [b"a", b"b", b"c"]
    t["s"] = ["\u2119", "\u01b4", "\u2602"]
    t.write(test_file, path="the_table", overwrite=True)

    t1 = Table.read(test_file, path="the_table", character_as_bytes=False)
    for col, col1 in zip(t.itercols(), t1.itercols()):
        assert np.all(col == col1)
    assert np.all(t1["p"].info.dtype.kind == "U")
    assert np.all(t1["q"].info.dtype.kind == "i")
    assert np.all(t1["r"].info.dtype.kind == "U")
    assert np.all(t1["s"].info.dtype.kind == "U")

    # Test default (character_as_bytes=True)
    t2 = Table.read(test_file, path="the_table")
    for col, col1 in zip(t.itercols(), t2.itercols()):
        assert np.all(col == col1)
    assert np.all(t2["p"].info.dtype.kind == "S")
    assert np.all(t2["q"].info.dtype.kind == "i")
    assert np.all(t2["r"].info.dtype.kind == "S")
    assert np.all(t2["s"].info.dtype.kind == "S")


def assert_objects_equal(obj1, obj2, attrs, compare_class=True):
    if compare_class:
        assert obj1.__class__ is obj2.__class__

    info_attrs = [
        "info.name",
        "info.format",
        "info.unit",
        "info.description",
        "info.meta",
        "info.dtype",
    ]
    for attr in attrs + info_attrs:
        a1 = obj1
        a2 = obj2
        for subattr in attr.split("."):
            try:
                a1 = getattr(a1, subattr)
                a2 = getattr(a2, subattr)
            except AttributeError:
                a1 = a1[subattr]
                a2 = a2[subattr]

        # Mixin info.meta can None instead of empty OrderedDict(), #6720 would
        # fix this.
        if attr == "info.meta":
            if a1 is None:
                a1 = {}
            if a2 is None:
                a2 = {}

        if isinstance(a1, np.ndarray) and a1.dtype.kind == "f":
            assert quantity_allclose(a1, a2, rtol=1e-15)
        elif isinstance(a1, np.dtype):
            # HDF5 does not perfectly preserve dtype: byte order can change, and
            # unicode gets stored as bytes.  So, we just check safe casting, to
            # ensure we do not, e.g., accidentally change integer to float, etc.
            if NUMPY_LT_1_22 and a1.names:
                # For old numpy, can_cast does not deal well with structured dtype.
                assert a1.names == a2.names
            else:
                assert np.can_cast(a2, a1, casting="safe")
        else:
            assert np.all(a1 == a2)


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_hdf5_mixins_qtable_to_table(tmp_path):
    """Test writing as QTable and reading as Table.  Ensure correct classes
    come out.
    """
    filename = tmp_path / "test_simple.hdf5"

    names = sorted(mixin_cols)

    t = QTable([mixin_cols[name] for name in names], names=names)
    t.write(filename, format="hdf5", path="root", serialize_meta=True)
    t2 = Table.read(filename, format="hdf5", path="root")

    assert t.colnames == t2.colnames

    for name, col in t.columns.items():
        col2 = t2[name]

        attrs = compare_attrs[name]
        compare_class = True

        if isinstance(col.info, QuantityInfo):
            # Downgrade Quantity to Column + unit
            assert type(col2) is Column
            # Class-specific attributes like `value` or `wrap_angle` are lost.
            attrs = ["unit"]
            compare_class = False
            # Compare data values here (assert_objects_equal doesn't know how in this case)
            assert np.all(col.value == col2)

        assert_objects_equal(col, col2, attrs, compare_class)


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
@pytest.mark.parametrize("table_cls", (Table, QTable))
def test_hdf5_mixins_as_one(table_cls, tmp_path):
    """Test write/read all cols at once and validate intermediate column names"""
    filename = tmp_path / "test_simple.hdf5"
    names = sorted(mixin_cols)
    all_serialized_names = []
    for name in names:
        all_serialized_names.extend(serialized_names[name])

    t = table_cls([mixin_cols[name] for name in names], names=names)
    t.meta["C"] = "spam"
    t.meta["comments"] = ["this", "is", "a", "comment"]
    t.meta["history"] = ["first", "second", "third"]

    t.write(filename, format="hdf5", path="root", serialize_meta=True)

    t2 = table_cls.read(filename, format="hdf5", path="root")
    assert t2.meta["C"] == "spam"
    assert t2.meta["comments"] == ["this", "is", "a", "comment"]
    assert t2.meta["history"] == ["first", "second", "third"]

    assert t.colnames == t2.colnames

    # Read directly via hdf5 and confirm column names
    h5 = h5py.File(filename, "r")
    h5_names = list(h5["root"].dtype.names)
    assert h5_names == all_serialized_names
    h5.close()


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
@pytest.mark.parametrize("name_col", list(mixin_cols.items()))
@pytest.mark.parametrize("table_cls", (Table, QTable))
def test_hdf5_mixins_per_column(table_cls, name_col, tmp_path):
    """Test write/read one col at a time and do detailed validation"""
    filename = tmp_path / "test_simple.hdf5"
    name, col = name_col

    c = [1.0, 2.0]
    t = table_cls([c, col, c], names=["c1", name, "c2"])
    t[name].info.description = "my description"
    t[name].info.meta = {"list": list(range(50)), "dict": {"a": "b" * 200}}

    if not t.has_mixin_columns:
        pytest.skip("column is not a mixin (e.g. Quantity subclass in Table)")

    t.write(filename, format="hdf5", path="root", serialize_meta=True)
    t2 = table_cls.read(filename, format="hdf5", path="root")

    assert t.colnames == t2.colnames

    for colname in t.colnames:
        compare = ["data"] if colname in ("c1", "c2") else compare_attrs[colname]
        assert_objects_equal(t[colname], t2[colname], compare)

    # Special case to make sure Column type doesn't leak into Time class data
    if name.startswith("tm"):
        assert t2[name]._time.jd1.__class__ is np.ndarray
        assert t2[name]._time.jd2.__class__ is np.ndarray


@pytest.mark.parametrize("name_col", unsupported_cols.items())
@pytest.mark.xfail(reason="column type unsupported")
def test_fits_unsupported_mixin(self, name_col, tmp_path):
    # Check that we actually fail in writing unsupported columns defined
    # on top.
    filename = tmp_path / "test_simple.fits"
    name, col = name_col
    Table([col], names=[name]).write(
        filename, format="hdf5", path="root", serialize_meta=True
    )


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_round_trip_masked_table_default(tmp_path):
    """Test round-trip of MaskedColumn through HDF5 using default serialization
    that writes a separate mask column.  Note:

    >>> simple_table(masked=True)
    <Table masked=True length=3>
      a      b     c
    int64 float64 str1
    ----- ------- ----
       --     1.0    c
        2     2.0   --
        3      --    e
    """
    filename = tmp_path / "test.h5"

    t = simple_table(masked=True)  # int, float, and str cols with one masked element
    t["c"] = [b"c", b"d", b"e"]
    t["c"].mask[1] = True
    t.write(filename, format="hdf5", path="root", serialize_meta=True)

    t2 = Table.read(filename)
    assert t2.masked is False
    assert t2.colnames == t.colnames
    for name in t2.colnames:
        assert np.all(t2[name].mask == t[name].mask)
        assert np.all(t2[name] == t[name])

        # Data under the mask round-trips also (unmask data to show this).
        t[name].mask = False
        t2[name].mask = False
        assert np.all(t2[name] == t[name])


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_overwrite_serialized_meta():
    # This used to cause an error because the meta data table
    # was not removed from the existing file.

    with h5py.File("test_data.h5", "w", driver="core", backing_store=False) as out:
        t1 = Table()
        t1.add_column(Column(data=[4, 8, 15], unit="cm"))
        t1.write(out, path="data", serialize_meta=True)

        t2 = Table.read(out, path="data")
        assert all(t1 == t2)
        assert t1.info(out=None) == t2.info(out=None)

        t3 = Table()
        t3.add_column(Column(data=[16, 23, 42], unit="g"))
        t3.write(out, path="data", serialize_meta=True, append=True, overwrite=True)

        t2 = Table.read(out, path="data")
        assert all(t3 == t2)
        assert t3.info(out=None) == t2.info(out=None)


@pytest.mark.skipif(not HAS_H5PY, reason="requires h5py")
def test_read_write_tilde_path(home_is_tmpdir):
    test_file = os.path.join("~", "test.hdf5")
    t1 = Table()
    t1["a"] = [1, 2, 3]

    t1.write(test_file, path="the_table")

    t1 = Table.read(test_file, path="the_table")
    t1 = Table.read(test_file, path="the_table", format="hdf5")

    # Ensure the data wasn't written to the literal tilde-prefixed path
    assert not os.path.exists(test_file)