improve

python/cudf/cudf/testing/_utils.py

@@ -217,7 +217,7 @@ def gen_rand(dtype, size, **kwargs):
         if kwargs.get("positive_only", False):
             return res
         else:
-            return res * 2 - 1
+            return 2 * res - 2
     elif dtype == np.int8 or dtype == np.int16:
         low = kwargs.get("low", -32)
         high = kwargs.get("high", 32)

python/cudf/cudf/tests/test_binops.py

@@ -960,11 +960,11 @@ def gen_df():
 @pytest.mark.parametrize("nulls", _nulls)
 @pytest.mark.parametrize("other", ["df", "scalar"])
 def test_logical_operator_func_dataframe(func, nulls, other):
-    rng = np.random.default_rng(seed=0)
     num_rows = 100
     num_cols = 3
 
     def gen_df():
+        rng = np.random.default_rng(seed=0)
         pdf = pd.DataFrame()
         from string import ascii_lowercase
 
@@ -983,8 +983,12 @@ def gen_df():
 
     pdf1 = gen_df()
     pdf2 = gen_df() if other == "df" else 59.0
-    gdf1 = cudf.DataFrame.from_pandas(pdf1)
-    gdf2 = cudf.DataFrame.from_pandas(pdf2) if other == "df" else 59.0
+    gdf1 = cudf.DataFrame.from_pandas(pdf1, nan_as_null=False)
+    gdf2 = (
+        cudf.DataFrame.from_pandas(pdf2, nan_as_null=False)
+        if other == "df"
+        else 59.0
+    )
 
     got = getattr(gdf1, func)(gdf2)
     expect = getattr(pdf1, func)(pdf2)[list(got._data)]

python/cudf/cudf/tests/test_dataframe.py

@@ -10537,10 +10537,10 @@ def test_dataframe_init_length_error(data, index):
 def test_dataframe_binop_with_mixed_date_types():
     rng = np.random.default_rng(seed=0)
     df = pd.DataFrame(
-        rng.random(2, 2),
+        rng.random(size=(2, 2)),
         columns=pd.Index(["2000-01-03", "2000-01-04"], dtype="datetime64[ns]"),
     )
-    ser = pd.Series(rng.random(3), index=[0, 1, 2])
+    ser = pd.Series(rng.random(size=3), index=[0, 1, 2])
     gdf = cudf.from_pandas(df)
     gser = cudf.from_pandas(ser)
     expected = df - ser

python/cudf/cudf/tests/test_datetime.py

@@ -219,7 +219,7 @@ def test_sort_datetime():
     rng = np.random.default_rng(seed=0)
     df = pd.DataFrame(
         {
-            "data": np.array(
+            "date": np.array(
                 [
                     np.datetime64("2016-11-20"),
                     np.datetime64("2020-11-20"),

python/cudf/cudf/tests/test_indexing.py

@@ -594,7 +594,7 @@ def test_dataframe_series_loc_multiindex(obj):
 def test_series_iloc(nelem):
     # create random cudf.Series
     rng = np.random.default_rng(seed=0)
-    ps = pd.Series(rng.sample(nelem))
+    ps = pd.Series(rng.random(nelem))
 
     # gpu cudf.Series
     gs = cudf.Series(ps)

python/cudf/cudf/tests/test_resampling.py

@@ -50,8 +50,9 @@ def test_series_upsample_simple():
 
 @pytest.mark.parametrize("rule", ["2s", "10s"])
 def test_series_resample_ffill(rule):
-    rng = pd.date_range("1/1/2012", periods=10, freq="5s")
-    ts = pd.Series(rng.integers(0, 500, len(rng)), index=rng)
+    date_idx = pd.date_range("1/1/2012", periods=10, freq="5s")
+    rng = np.random.default_rng(seed=0)
+    ts = pd.Series(rng.integers(0, 500, len(date_idx)), index=date_idx)
     gts = cudf.from_pandas(ts)
     assert_resample_results_equal(
         ts.resample(rule).ffill(), gts.resample(rule).ffill()
@@ -60,8 +61,9 @@ def test_series_resample_ffill(rule):
 
 @pytest.mark.parametrize("rule", ["2s", "10s"])
 def test_series_resample_bfill(rule):
-    rng = pd.date_range("1/1/2012", periods=10, freq="5s")
-    ts = pd.Series(rng.integers(0, 500, len(rng)), index=rng)
+    date_idx = pd.date_range("1/1/2012", periods=10, freq="5s")
+    rng = np.random.default_rng(seed=0)
+    ts = pd.Series(rng.integers(0, 500, len(date_idx)), index=date_idx)
     gts = cudf.from_pandas(ts)
     assert_resample_results_equal(
         ts.resample(rule).bfill(), gts.resample(rule).bfill()
@@ -70,8 +72,9 @@ def test_series_resample_bfill(rule):
 
 @pytest.mark.parametrize("rule", ["2s", "10s"])
 def test_series_resample_asfreq(rule):
-    rng = pd.date_range("1/1/2012", periods=100, freq="5s")
-    ts = pd.Series(rng.integers(0, 500, len(rng)), index=rng)
+    date_range = pd.date_range("1/1/2012", periods=100, freq="5s")
+    rng = np.random.default_rng(seed=0)
+    ts = pd.Series(rng.integers(0, 500, len(date_range)), index=date_range)
     gts = cudf.from_pandas(ts)
     assert_resample_results_equal(
         ts.resample(rule).asfreq(), gts.resample(rule).asfreq()