def test_index_copy_datetime(name, dtype, deep=True):
cidx = cudf.DatetimeIndex(["2001", "2002", "2003"])
pidx = cidx.to_pandas()
pidx_copy = pidx.copy(name=name, deep=deep, dtype=dtype)
cidx_copy = cidx.copy(name=name, deep=deep, dtype=dtype)
# By default, cudf.DatetimeIndex uses [ms] as base unit, pandas uses [ns]
if dtype == "int64":
cidx_copy = cidx_copy * 1000000
assert_eq(pidx_copy, cidx_copy)
def _load(self):
print('Loading Environment...')
csv_file = "/media/alan/seagate/Downloads/Binance_BTCUSDT_minute.csv"
columns = ['Date', 'Open', 'High', 'Low', 'Close', 'Volume']
if self.usd_cuda:
df = cudf.read_csv(csv_file)
else:
df = pd.read_csv(csv_file)
df = df.drop(columns=['unix', 'symbol', 'Volume USDT', 'tradecount::'])
df.columns = columns
if self.usd_cuda:
df.index = cudf.DatetimeIndex(df['Date'])
else:
df.index = pd.DatetimeIndex(df['Date'])
df = df.drop(columns=['Date', 'Volume'])
df = df.dropna()
df = df[::-1]
data = df.values
return data
@pytest.mark.parametrize("dtype", NUMERIC_TYPES + ["category"])
def test_index_copy_category(name, dtype, deep=True):
cidx = cudf.core.index.CategoricalIndex([1, 2, 3])
pidx = cidx.to_pandas()
pidx_copy = pidx.copy(name=name, deep=deep, dtype=dtype)
cidx_copy = cidx.copy(name=name, deep=deep, dtype=dtype)
assert_eq(pidx_copy, cidx_copy)
@pytest.mark.parametrize("deep", [True, False])
@pytest.mark.parametrize(
"idx",
[
cudf.DatetimeIndex(["2001", "2002", "2003"]),
cudf.core.index.StringIndex(["a", "b", "c"]),
cudf.Int64Index([1, 2, 3]),
cudf.Float64Index([1.0, 2.0, 3.0]),
cudf.CategoricalIndex([1, 2, 3]),
cudf.CategoricalIndex(["a", "b", "c"]),
],
)
def test_index_copy_deep(idx, deep):
"""Test if deep copy creates a new instance for device data.
The general criterion is to compare `Buffer.ptr` between two data objects.
Specifically for:
- CategoricalIndex, this applies to both `.codes` and `.categories`
- StringIndex, to every element in `._base_children`
- Others, to `.base_data`
No test is defined for RangeIndex.
