def test_array_concat_scalar(client, op, op_raw):
raw_left = np.array([1, 2, 3])
raw_right = np.array([3, 4])
left = ibis.array(raw_left)
right = ibis.array(raw_right)
expr = op(left, right)
result = client.execute(expr)
expected = op_raw(raw_left, raw_right)
tm.assert_numpy_array_equal(result, expected)
def test_series_to_ibis_literal(core_client):
values = [1, 2, 3, 4]
s = dd.from_pandas(pd.Series(values), npartitions=1)
expr = ibis.array(s)
assert expr.equals(ibis.array(values))
assert core_client.execute(expr) == pytest.approx([1, 2, 3, 4])
def test_array_length_scalar(client):
raw_value = np.array([1, 2, 4])
value = ibis.array(raw_value)
expr = value.length()
result = client.execute(expr)
expected = len(raw_value)
assert result == expected
def test_array_index_scalar(client, index):
raw_value = np.array([-10, 1, 2, 42])
value = ibis.array(raw_value)
expr = value[index]
result = client.execute(expr)
expected = raw_value[index] if index < len(raw_value) else None
assert result == expected
def test_array_slice_scalar(client, start, stop):
raw_value = np.array([-11, 42, 10])
value = ibis.array(raw_value)
expr = value[start:stop]
result = client.execute(expr)
expected = raw_value[start:stop]
tm.assert_numpy_array_equal(result, expected)
def test_array_repeat_scalar(client, n, mul):
raw_array = np.array([1, 2])
array = ibis.array(raw_array)
expr = mul(array, n)
result = client.execute(expr)
if n > 0:
expected = np.tile(raw_array, n)
else:
expected = np.array([], dtype=raw_array.dtype)
tm.assert_numpy_array_equal(result, expected)
def test_array_scalar(con):
expr = ibis.array([1.0, 2.0, 3.0])
assert isinstance(expr, ir.ArrayScalar)
result = con.execute(expr)
expected = np.array([1.0, 2.0, 3.0])
# This does not check whether `result` is an np.array or a list,
# because it varies across backends and backend configurations
assert np.array_equal(result, expected)
def test_array_column(backend, alltypes, df):
expr = ibis.array([alltypes['double_col'], alltypes['double_col']])
assert isinstance(expr, ir.ArrayColumn)
result = expr.execute()
expected = df.apply(
lambda row: [row['double_col'], row['double_col']],
axis=1,
)
backend.assert_series_equal(result, expected, check_names=False)
def test_unnest_unnamed(con):
array_types = con.table("array_types")
df = array_types.execute()
expr = (
array_types.x.cast("!array<int64>")
+ ibis.array([1], type="!array<int64>")
).unnest()
assert not expr.has_name()
result = expr.name("x").execute()
expected = df.x.map(lambda x: x + [1]).explode("x")
tm.assert_series_equal(result, expected.astype("float64"))
