change sample method to work for pandas

narwhals/expression.py

@@ -112,6 +112,9 @@ def __ge__(self, other: Any) -> Expr:
         )
 
     # --- unary ---
+    def __invert__(self) -> Expr:
+        return self.__class__(lambda plx: self._call(plx).__invert__())
+
     def mean(self) -> Expr:
         return self.__class__(lambda plx: self._call(plx).mean())
 

narwhals/pandas_like/dataframe.py

@@ -8,6 +8,7 @@
 
 from narwhals.pandas_like.utils import evaluate_into_exprs
 from narwhals.pandas_like.utils import horizontal_concat
+from narwhals.pandas_like.utils import maybe_reset_indices
 from narwhals.pandas_like.utils import translate_dtype
 from narwhals.pandas_like.utils import validate_dataframe_comparand
 from narwhals.utils import flatten_str
@@ -86,6 +87,7 @@ def select(
         **named_exprs: IntoPandasExpr,
     ) -> Self:
         new_series = evaluate_into_exprs(self, *exprs, **named_exprs)
+        new_series = maybe_reset_indices(new_series)
         df = horizontal_concat(
             [series._series for series in new_series],
             implementation=self._implementation,

narwhals/pandas_like/expr.py

@@ -180,8 +180,16 @@ def n_unique(self) -> Self:
     def unique(self) -> Self:
         return register_expression_call(self, "unique")
 
-    def sample(self, n: int, fraction: float, *, with_replacement: bool) -> Self:
-        return register_expression_call(self, "sample", n, fraction, with_replacement)
+    def sample(
+        self,
+        n: int | None = None,
+        fraction: float | None = None,
+        *,
+        with_replacement: bool = False,
+    ) -> Self:
+        return register_expression_call(
+            self, "sample", n, fraction=fraction, with_replacement=with_replacement
+        )
 
     def alias(self, name: str) -> Self:
         # Define this one manually, so that we can

narwhals/pandas_like/series.py

@@ -291,11 +291,15 @@ def zip_with(self, mask: PandasSeries, other: PandasSeries) -> PandasSeries:
         ser = self._series
         return self._from_series(ser.where(mask, other))
 
-    def sample(self, n: int, fraction: float, *, with_replacement: bool) -> PandasSeries:
+    def sample(
+        self,
+        n: int | None = None,
+        fraction: float | None = None,
+        *,
+        with_replacement: bool = False,
+    ) -> PandasSeries:
         ser = self._series
-        return self._from_series(
-            ser.sample(n=n, frac=fraction, with_replacement=with_replacement)
-        )
+        return self._from_series(ser.sample(n=n, frac=fraction, replace=with_replacement))
 
     def unique(self) -> PandasSeries:
         if self._implementation != "pandas":

narwhals/pandas_like/utils.py

@@ -381,3 +381,15 @@ def reverse_translate_dtype(dtype: DType | type[DType]) -> Any:
         return "bool"
     msg = f"Unknown dtype: {dtype}"
     raise TypeError(msg)
+
+
+def maybe_reset_indices(series: list[PandasSeries]) -> list[PandasSeries]:
+    idx = series[0]._series.index
+    found_non_matching_index = False
+    for s in series[1:]:
+        if s._series.index is not idx and not (s._series.index == idx).all():
+            found_non_matching_index = True
+            break
+    if found_non_matching_index:
+        return [s._from_series(s._series.reset_index(drop=True)) for s in series]
+    return series

narwhals/series.py

@@ -118,9 +118,15 @@ def zip_with(self, mask: Self, other: Self) -> Self:
             self._series.zip_with(self._extract_native(mask), self._extract_native(other))
         )
 
-    def sample(self, n: int, fraction: float, *, with_replacement: bool) -> Self:
+    def sample(
+        self,
+        n: int | None = None,
+        fraction: float | None = None,
+        *,
+        with_replacement: bool = False,
+    ) -> Self:
         return self._from_series(
-            self._series.sample(n, fraction=fraction, with_replacement=with_replacement)
+            self._series.sample(n=n, fraction=fraction, with_replacement=with_replacement)
         )
 
     def to_numpy(self) -> Any:

tests/test_common.py

@@ -15,6 +15,8 @@
 df_pandas = pd.DataFrame({"a": [1, 3, 2], "b": [4, 4, 6], "z": [7.0, 8, 9]})
 df_polars = pl.DataFrame({"a": [1, 3, 2], "b": [4, 4, 6], "z": [7.0, 8, 9]})
 df_lazy = pl.LazyFrame({"a": [1, 3, 2], "b": [4, 4, 6], "z": [7.0, 8, 9]})
+df_pandas_na = pd.DataFrame({"a": [None, 3, 2], "b": [4, 4, 6], "z": [7.0, None, 9]})
+df_lazy_na = pl.LazyFrame({"a": [None, 3, 2], "b": [4, 4, 6], "z": [7.0, None, 9]})
 
 if os.environ.get("CI", None):
     import modin.pandas as mpd
@@ -321,3 +323,21 @@ def test_expr_min_max(df_raw: Any) -> None:
     expected_max = {"a": [3], "b": [6], "z": [9]}
     compare_dicts(result_min, expected_min)
     compare_dicts(result_max, expected_max)
+
+
+@pytest.mark.parametrize("df_raw", [df_polars, df_pandas, df_mpd, df_lazy])
+def test_expr_sample(df_raw: Any) -> None:
+    df = nw.LazyFrame(df_raw)
+    result_shape = nw.to_native(df.select(nw.col("a", "b").sample(n=2)).collect()).shape
+    expected = (2, 2)
+    assert result_shape == expected
+
+
+@pytest.mark.parametrize("df_raw", [df_pandas_na, df_lazy_na])
+def test_expr_na(df_raw: Any) -> None:
+    df = nw.LazyFrame(df_raw)
+    result_nna = nw.to_native(
+        df.filter((~nw.col("a").is_null()) & (~nw.col("z").is_null()))
+    )
+    expected = {"a": [2], "b": [6], "z": [9]}
+    compare_dicts(result_nna, expected)