Merge branch 'branch-23.04' into add_docs_build

docs/cudf/source/api_docs/dataframe.rst

@@ -146,7 +146,6 @@ Computations / descriptive stats
    DataFrame.prod
    DataFrame.product
    DataFrame.quantile
-   DataFrame.quantiles
    DataFrame.rank
    DataFrame.round
    DataFrame.skew

python/cudf/cudf/core/dataframe.py

@@ -5173,13 +5173,21 @@ def from_arrow(cls, table):
 
         if index_col:
             if isinstance(index_col[0], dict):
-                out = out.set_index(
-                    cudf.RangeIndex(
-                        index_col[0]["start"],
-                        index_col[0]["stop"],
-                        name=index_col[0]["name"],
-                    )
+                idx = cudf.RangeIndex(
+                    index_col[0]["start"],
+                    index_col[0]["stop"],
+                    name=index_col[0]["name"],
                 )
+                if len(idx) == len(out):
+                    # `idx` is generated from arrow `pandas_metadata`
+                    # which can get out of date with many of the
+                    # arrow operations. Hence verifying if the
+                    # lengths match, or else don't need to set
+                    # an index at all i.e., Default RangeIndex
+                    # will be set.
+                    # See more about the discussion here:
+                    # https://github.com/apache/arrow/issues/15178
+                    out = out.set_index(idx)
             else:
                 out = out.set_index(index_col[0])
 

python/cudf/cudf/core/indexed_frame.py

@@ -2496,7 +2496,11 @@ def round(self, decimals=0, how="half_even"):
 
         cols = {
             name: col.round(decimals[name], how=how)
-            if (name in decimals and _is_non_decimal_numeric_dtype(col.dtype))
+            if (
+                name in decimals
+                and _is_non_decimal_numeric_dtype(col.dtype)
+                and not is_bool_dtype(col.dtype)
+            )
             else col.copy(deep=True)
             for name, col in self._data.items()
         }

python/cudf/cudf/tests/test_dataframe.py

@@ -3813,17 +3813,22 @@ def test_ndim():
         -3,
         0,
         5,
-        pd.Series([1, 4, 3, -6], index=["w", "x", "y", "z"]),
-        cudf.Series([-4, -2, 12], index=["x", "y", "z"]),
-        {"w": -1, "x": 15, "y": 2},
+        pd.Series(
+            [1, 4, 3, -6],
+            index=["floats", "ints", "floats_with_nan", "floats_same"],
+        ),
+        cudf.Series(
+            [-4, -2, 12], index=["ints", "floats_with_nan", "floats_same"]
+        ),
+        {"floats": -1, "ints": 15, "floats_will_nan": 2},
     ],
 )
 def test_dataframe_round(decimals):
-    pdf = pd.DataFrame(
+    gdf = cudf.DataFrame(
         {
-            "w": np.arange(0.5, 10.5, 1),
-            "x": np.random.normal(-100, 100, 10),
-            "y": np.array(
+            "floats": np.arange(0.5, 10.5, 1),
+            "ints": np.random.normal(-100, 100, 10),
+            "floats_with_na": np.array(
                 [
                     14.123,
                     2.343,
@@ -3832,31 +3837,25 @@ def test_dataframe_round(decimals):
                     -8.302,
                     np.nan,
                     94.313,
-                    -112.236,
+                    None,
                     -8.029,
                     np.nan,
                 ]
             ),
-            "z": np.repeat([-0.6459412758761901], 10),
+            "floats_same": np.repeat([-0.6459412758761901], 10),
+            "bools": np.random.choice([True, None, False], 10),
+            "strings": np.random.choice(["abc", "xyz", None], 10),
+            "struct": np.random.choice([{"abc": 1}, {"xyz": 2}, None], 10),
+            "list": [[1], [2], None, [4], [3]] * 2,
         }
     )
-    gdf = cudf.DataFrame.from_pandas(pdf)
+    pdf = gdf.to_pandas()
 
     if isinstance(decimals, cudf.Series):
         pdecimals = decimals.to_pandas()
     else:
         pdecimals = decimals
 
-    result = gdf.round(decimals)
-    expected = pdf.round(pdecimals)
-    assert_eq(result, expected)
-
-    # with nulls, maintaining existing null mask
-    for c in pdf.columns:
-        arr = pdf[c].to_numpy().astype("float64")  # for pandas nulls
-        arr.ravel()[np.random.choice(10, 5, replace=False)] = np.nan
-        pdf[c] = gdf[c] = arr
-
     result = gdf.round(decimals)
     expected = pdf.round(pdecimals)
 
@@ -10012,3 +10011,17 @@ def test_dataframe_transpose_complex_types(data):
     actual = gdf.T
 
     assert_eq(expected, actual)
+
+
+def test_dataframe_from_arrow_slice():
+    table = pa.Table.from_pandas(
+        pd.DataFrame.from_dict(
+            {"a": ["aa", "bb", "cc"] * 3, "b": [1, 2, 3] * 3}
+        )
+    )
+    table_slice = table.slice(3, 7)
+
+    expected = table_slice.to_pandas()
+    actual = cudf.DataFrame.from_arrow(table_slice)
+
+    assert_eq(expected, actual)