def test():
sr = pd.Series([1, 2, 3])
array = np.asarray(sr)
x = np.array([1, 2, 3])
assert np.array_equal(array, x)
y = np.array([4, 5, 6])
z = np.add(sr, y)
assert np.array_equal(z, x + y)
df = pd.DataFrame({"x": x, "y": y})
z = np.add(df["x"], df["y"])
assert np.array_equal(z, x + y)
return
"c1": c1,
"c3": c3_l,
"key1": np.array(key_left, dtype=key_dtype1),
"key2": np.array(key_left, dtype=key_dtype2),
}
)
df2 = pd.DataFrame(
{
"c2": c2,
"c3": c3_r,
"key1": np.array(key_right, dtype=key_dtype1),
"key2": np.array(key_right, dtype=key_dtype2),
}
)
ldf1 = lp.DataFrame(df1)
ldf2 = lp.DataFrame(df2)
join_pandas = df1.merge(df2, on=keys)
join_legate = ldf1.merge(ldf2, on=keys, method="broadcast")
join_legate_hash = ldf1.merge(ldf2, on=keys, method="hash")
assert sort_and_compare(join_pandas, to_pandas(join_legate))
assert sort_and_compare(join_pandas, to_pandas(join_legate_hash))
key_left = list(chain(*[[x] * 3 for x in range(n // 3, 0, -1)]))
for pair in product(key_dtypes, key_dtypes[1:] + key_dtypes[:1]):
key_dtype1, key_dtype2 = pair
print(
"Type: left, Size: %u, Key dtype1: %s, Key dtype2: %s "
% (n, str(key_dtype1), str(key_dtype2))
# Copyright 2021 NVIDIA Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import pandas as pd
from legate import pandas as lp
df = pd.DataFrame({"col1": [1, 2, 3], "col2": [4, 5, 6]})
ldf = lp.DataFrame(df)
df["col3"] = df["col1"]
ldf["col3"] = ldf["col1"]
assert ldf.equals(lp.DataFrame(df))
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import pandas as pd
from legate import pandas as lp
df = pd.DataFrame({
"a": range(10),
"b": range(1, 11),
"c": [str(i) * 3 for i in range(10)],
"d": [str(i % 3) for i in range(10)],
})
df["c"] = df["c"].astype(pd.StringDtype())
df["d"] = df["d"].astype("category")
ldf = lp.DataFrame(df)
assert ldf.tail(2).equals(df.tail(2))
assert ldf.tail(9).equals(df.tail(9))
# See the License for the specific language governing permissions and
# limitations under the License.
#
import pandas as pd
from legate import pandas as lp
from tests.utils import equals, must_fail
def _test(ex, df, *args):
must_fail(ex, df.insert, *args)
df = pd.DataFrame()
ldf = lp.DataFrame()
df.insert(0, "a", 1)
ldf.insert(0, "a", 1)
assert equals(ldf, df)
df = pd.DataFrame(index=[1, 2, 3])
ldf = lp.DataFrame(index=[1, 2, 3])
df.insert(0, "a", 1)
ldf.insert(0, "a", 1)
assert equals(ldf, df)
df = pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
try:
from legate.numpy.random import random
except ModuleNotFoundError:
from numpy.random import random
import pandas as pd
from legate import pandas as lp
x = random(10)
y = random(10)
df = pd.DataFrame({"x": x, "y": y})
ldf1 = lp.DataFrame({"x": x, "y": y})
# FIXME: We don't handle this case correctly now. DataFrame's ctor
# should align all series in the dictionary.
# ldf2 = lp.DataFrame({"x": lp.Series(x), "y": lp.Series(y)})
ldf3 = lp.DataFrame(ldf1)
assert ldf1.equals(lp.DataFrame(df))
# assert ldf2.equals(lp.DataFrame(df))
assert ldf3.equals(lp.DataFrame(df))
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import pandas as pd
from legate import pandas as lp
n = 17
a = [str(i) * 3 for i in range(n)]
for i in range(n):
if i % 4 == 0:
a[i] = None
s = pd.Series(a)
s = s.astype(pd.StringDtype())
ls = lp.DataFrame(s)
out_pd = s.dropna()
out_lp = ls.dropna()
assert out_lp.equals(out_pd)
# limitations under the License.
#
import pandas as pd
from numpy.random import permutation, randn
from legate import pandas as lp
from tests.utils import equals
n = 17
indices = [pd.RangeIndex(1, n + 1), pd.Index(permutation(n))]
for index in indices:
print(f"Index: {index}")
df1 = pd.DataFrame({1: randn(n), 2: randn(n), 5: randn(n)}, index=index)
ldf1 = lp.DataFrame(df1)
df2 = pd.DataFrame({1: randn(n), 2: randn(n), 5: randn(n)}, index=index)
out_pd = df1 + df2
out_lp = ldf1 + df2
assert equals(out_lp, out_pd)
out_pd = df1 + df2.values
out_lp = ldf1 + df2.values
assert equals(out_lp, out_pd)
out_pd = df1.add(df2[1].values, axis=0)
out_lp = ldf1.add(df2[1].values, axis=0)
assert equals(out_lp, out_pd)
out_pd = df1.add(df2[1].to_list(), axis=0)
{"c1": c1, "key1": np.array(key_left, dtype=key_dtype), "c3": c3_l}
)
df1["key"] = df1["key1"]
df1_key_on_index = df1.set_index("key")
df2 = pd.DataFrame(
{
"c2": c2,
"key2": np.array(key_right, dtype=key_dtype),
"c3": c3_r,
}
)
df2["key"] = df2["key2"]
df2_key_on_index = df2.set_index("key")
ldf1 = lp.DataFrame(df1)
ldf1_key_on_index = lp.DataFrame(df1_key_on_index)
ldf2 = lp.DataFrame(df2)
ldf2_key_on_index = lp.DataFrame(df2_key_on_index)
join_pandas2 = df1.merge(
df2_key_on_index, left_on="key1", right_index=True
)
join_pandas4 = df1_key_on_index.merge(
df2, right_on="key2", left_index=True
)
# XXX: Pandas sort the keys in the output when both left_index and
# right_index are True, whereas Legate will not for performance
# reasons. In this test we sorted the keys in the input dataframe
# so that Pandas' join output coincides with Legate's. We can't
# and won't guarantee this semantics equivalence in general.
import pandas as pd
from legate import pandas as lp
from tests.utils import equals, must_fail
def _test(ex, df, *args):
def _loc():
df.loc[args]
must_fail(ex, _loc)
n = 17
for index in [pd.RangeIndex(3, n + 3), pd.Index(list(range(3, n + 3)))]:
df_copy = lp.DataFrame({"a": range(n)}, index=index)
df = lp.DataFrame({"a": range(n)}, index=index)
_test(KeyError, df, n + 3)
_test(KeyError, df, n + 4, "a")
assert len(df.loc[n + 3:n + 4]) == 0
df.loc[n + 3] = 100
assert equals(df_copy, df)
df.loc[n + 3:n + 4] = 200
assert equals(df_copy, df)
pd.RangeIndex(21, 1, -2),
pd.Index(permutation(10)),
]:
print(f"Index: {index}")
df = pd.DataFrame(
{
"a": range(10),
"b": range(1, 11),
"c": [str(i) * 3 for i in range(10)],
"d": [str(i % 3) for i in range(10)],
},
index=index,
)
df["c"] = df["c"].astype(pd.StringDtype())
df["d"] = df["d"].astype("category")
ldf = lp.DataFrame(df)
for idx in range(4):
print(f"Testing ldf.iat[{index[idx + 3]}, {idx}].__getitem__")
out_pd = df.iat[idx + 3, idx]
out_lp = ldf.iat[idx + 3, idx]
assert equals_scalar(out_lp, out_pd)
for idx, val in enumerate([100, 200, "5678"]):
print(f"Testing ldf.iat[{index[idx + 3]}, {idx}].__setitem__")
df.iat[idx + 3, idx] = val
ldf.iat[idx + 3, idx] = val
out_pd = df.iat[idx + 3, idx]
out_lp = ldf.iat[idx + 3, idx]
assert equals_scalar(out_lp, out_pd)
import pandas as pd
from legate import pandas as lp
from tests.utils import equals
indices = [
pd.RangeIndex(3),
pd.RangeIndex(1, 4),
pd.RangeIndex(6, step=2),
pd.RangeIndex(1, 10, step=3),
]
for index in indices:
df = pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]}, index=index)
ldf = lp.DataFrame(df)
# Passing Legate dataframes as arguments
assert equals(lp.DataFrame(ldf, dtype="float64"),
pd.DataFrame(df, dtype="float64"))
assert equals(lp.DataFrame(ldf, columns=["a"]),
pd.DataFrame(df, columns=["a"]))
assert equals(
lp.DataFrame(ldf, columns=["a"], dtype="float64"),
pd.DataFrame(df, columns=["a"], dtype="float64"),
)
# Passing Legate series as arguments
assert equals(lp.DataFrame(ldf["a"]), pd.DataFrame(df["a"]))
assert equals(
lp.DataFrame(ldf["a"], dtype="float32"),
# http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # import pandas as pd from legate import pandas as lp from tests.utils import equals df = pd.DataFrame() ldf = lp.DataFrame() assert equals(ldf, df) df["a"] = 0 ldf["a"] = 0 assert equals(ldf, df) df = pd.DataFrame(columns=["a", "b"]) ldf = lp.DataFrame(columns=["a", "b"]) assert equals(ldf, df) df.loc[:, "a"] = "1" ldf.loc[:, "a"] = "1"
def test(
size_per_proc=1000,
num_procs=1,
num_runs=1,
scale_lhs_only=False,
package="legate",
ty="int64",
key_length=40,
pad_side="right",
):
if package == "legate":
from legate import numpy as np, pandas as pd
from legate.numpy.random import randn
elif package == "cudf":
import cudf as pd
import cupy as np
from cupy.random import randn
elif package == "pandas":
import numpy as np
import pandas as pd
from numpy.random import randn
else:
print("Unknown dataframe package: %s" % package)
assert False
if package == "legate":
from legate.timing import time
def block():
pass
def get_timestamp():
return time()
def compute_elapsed_time(start_ts, stop_ts):
return (stop_ts - start_ts) / 1000.0
else:
import time
def block():
pass
get_timestamp = time.process_time
def compute_elapsed_time(start_ts, stop_ts):
return (stop_ts - start_ts) * 1000.0
size = size_per_proc * num_procs
key = np.arange(size, dtype=np.int64) % size_per_proc
payload = randn(size)
df = pd.DataFrame({"key": key, "payload": payload})
if ty == "int64":
df["key"] = df["key"] * -1
ascending = True
if ty == "string":
df["key"] = (
df["key"]
.astype(str)
.str.pad(width=key_length, side=pad_side, fillchar="0")
)
ascending = False
print("Size: %u, Key dtype: %s" % (size, df["key"].dtype))
block()
for i in range(num_runs):
start_ts = get_timestamp()
result = df.sort_values("key", ignore_index=True, ascending=ascending)
stop_ts = get_timestamp()
print(
"[Run %d] Elapsed time: %lf ms"
% (i + 1, compute_elapsed_time(start_ts, stop_ts))
)
del result
"key1": np.array(key1, dtype=np.int64),
"key2": np.array(key1[::-1], dtype=np.int64),
})
df2 = pd.DataFrame({
"c2": c2,
"key1": np.array(key2, dtype=np.int64),
"key2": np.array(key2[::-1], dtype=np.int64),
})
df3 = pd.DataFrame({"c3": c3, "key1": np.array(key3, dtype=np.int64)})
df4 = pd.DataFrame({
"c4": c3,
"key1": np.array(key3, dtype=np.int64),
"key2": np.array(key3[::-1], dtype=np.int64),
})
ldf1 = lp.DataFrame(df1)
ldf2 = lp.DataFrame(df2)
ldf3 = lp.DataFrame(df3)
ldf4 = lp.DataFrame(df4)
join_pandas = (df1.merge(df2,
on=["key1",
"key2"]).merge(df3,
on="key1").merge(df4,
on="key1"))
join_legate = (ldf1.merge(ldf2,
on=["key1",
"key2"]).merge(ldf3,
on="key1").merge(ldf4,
on="key1"))
def test(
size_per_proc=1000,
num_procs=1,
num_runs=1,
ty="int64",
key_length=10,
scale_lhs_only=False,
package="legate",
):
if package == "legate":
from legate import numpy as np, pandas as pd
from legate.numpy.random import randn
elif package == "cudf":
import cudf as pd
import cupy as np
from cupy.random import randn
elif package == "pandas":
import numpy as np
import pandas as pd
from numpy.random import randn
elif package == "dask" or package == "daskcudf":
import dask.array as da
import dask.dataframe as df
import numpy as np
if package == "daskcudf":
import cudf
else:
print("Unknown dataframe package: %s" % package)
assert False
if package == "legate":
from legate.timing import time
def block(*args):
pass
def get_timestamp():
return time()
def compute_elapsed_time(start_ts, stop_ts):
return (stop_ts - start_ts) / 1000.0
elif package == "dask" or package == "daskcudf":
import time
def block(*args):
for arg in args:
arg.compute()
get_timestamp = time.process_time
def compute_elapsed_time(start_ts, stop_ts):
return (stop_ts - start_ts) * 1000.0
else:
import time
def block(*args):
pass
get_timestamp = time.process_time
def compute_elapsed_time(start_ts, stop_ts):
return (stop_ts - start_ts) * 1000.0
if scale_lhs_only:
size = size_per_proc * num_procs
size_rhs = size // 3
if package == "dask" or package == "daskcudf":
# Dask array does not have randn so use arrange
c1 = da.arange(size, dtype=np.float64, chunks=size_per_proc)
c2 = da.arange(
size_rhs,
dtype=np.float64,
chunks=(size_per_proc + num_procs - 1) // num_procs,
)
else:
c1 = randn(size)
c2 = randn(size_rhs)
key_dtype = np.int64
if package == "dask" or package == "daskcudf":
key_left = (
da.arange(size, dtype=key_dtype, chunks=size_per_proc)
% size_per_proc
)
key_right = da.arange(
size_rhs,
dtype=key_dtype,
chunks=(size_per_proc + num_procs - 1) // num_procs,
)
da.multiply(key_right, 3, out=key_right)
else:
key_left = np.arange(size, dtype=key_dtype) % size_per_proc
key_right = np.arange(size_rhs, dtype=key_dtype)
np.multiply(key_right, 3, out=key_right)
else:
size = size_per_proc * num_procs
size_rhs = size
if package == "dask" or package == "daskcudf":
# Dask array does not have randn so use arrange
c1 = da.arange(size, dtype=np.float64, chunks=size_per_proc)
c2 = da.arange(size, dtype=np.float64, chunks=size_per_proc)
else:
c1 = randn(size)
c2 = randn(size)
key_dtype = np.int64
if package == "dask" or package == "daskcudf":
key_left = da.arange(size, dtype=key_dtype, chunks=size_per_proc)
key_right = da.arange(size, dtype=key_dtype, chunks=size_per_proc)
else:
key_left = np.arange(size, dtype=key_dtype)
key_right = np.arange(size, dtype=key_dtype)
# np.floor_divide(key_right, 3, out=key_right)
# np.multiply(key_right, 3, out=key_right)
if package == "dask" or package == "daskcudf":
df1 = df.multi.concat(
[df.from_dask_array(a) for a in [c1, key_left]], axis=1
)
df1.columns = ["c1", "key"]
df2 = df.multi.concat(
[df.from_dask_array(a) for a in [c2, key_right]], axis=1
)
df2.columns = ["c2", "key"]
if package == "daskcudf":
df1 = df1.map_partitions(cudf.from_pandas)
df2 = df2.map_partitions(cudf.from_pandas)
else:
df1 = pd.DataFrame({"c1": c1, "key": key_left})
df2 = pd.DataFrame({"c2": c2, "key": key_right})
df2["key"] = df2["key"] // 3 * 3
if ty == "string":
df1["key"] = (
df1["key"]
.astype("string")
.str.pad(width=key_length, side="both", fillchar="0")
)
df2["key"] = (
df2["key"]
.astype("string")
.str.pad(width=key_length, side="both", fillchar="0")
)
print(
"Type: inner, Size: %u x %u, Key dtype: %s"
% (size, size_rhs, str(key_dtype))
)
block(df1, df2)
for i in range(num_runs):
start_ts = get_timestamp()
df_result = df1.merge(df2, on="key")
block(df_result)
stop_ts = get_timestamp()
print(
"[Run %d] Elapsed time: %lf ms"
% (i + 1, compute_elapsed_time(start_ts, stop_ts))
)
del df_result
