feat: add Series.drop nulls for pyarrow (narwhals-dev#542)

narwhals/_arrow/expr.py

@@ -189,6 +189,9 @@ def all(self) -> Self:
     def sum(self) -> Self:
         return reuse_series_implementation(self, "sum", returns_scalar=True)
 
+    def drop_nulls(self) -> Self:
+        return reuse_series_implementation(self, "drop_nulls")
+
     def alias(self, name: str) -> Self:
         # Define this one manually, so that we can
         # override `output_names` and not increase depth

narwhals/_arrow/series.py

@@ -203,6 +203,10 @@ def sum(self) -> int:
         pc = get_pyarrow_compute()
         return pc.sum(self._native_series)  # type: ignore[no-any-return]
 
+    def drop_nulls(self) -> ArrowSeries:
+        pc = get_pyarrow_compute()
+        return self._from_native_series(pc.drop_null(self._native_series))
+
     def std(self, ddof: int = 1) -> int:
         pc = get_pyarrow_compute()
         return pc.stddev(self._native_series, ddof=ddof)  # type: ignore[no-any-return]

tests/expr_and_series/drop_nulls_test.py

@@ -0,0 +1,31 @@
+from __future__ import annotations
+
+from typing import Any
+
+import narwhals as nw
+from tests.utils import compare_dicts
+
+
+def test_drop_nulls(constructor: Any) -> None:
+    data = {
+        "A": [1, 2, None, 4],
+        "B": [5, 6, 7, 8],
+        "C": [None, None, None, None],
+        "D": [9, 10, 11, 12],
+    }
+
+    df = nw.from_native(constructor(data), eager_only=True)
+
+    result_a = df.select(nw.col("A").drop_nulls())
+    result_b = df.select(df["B"].drop_nulls())
+    result_c = df.select(nw.col("C").drop_nulls())
+    result_d = df.select(nw.col("D").drop_nulls())
+    expected_a = {"A": [1.0, 2.0, 4.0]}
+    expected_b = {"B": [5, 6, 7, 8]}
+    expected_c = {"C": []}  # type: ignore[var-annotated]
+    expected_d = {"D": [9, 10, 11, 12]}
+
+    compare_dicts(result_a, expected_a)
+    compare_dicts(result_b, expected_b)
+    compare_dicts(result_c, expected_c)
+    compare_dicts(result_d, expected_d)

tests/frame/test_common.py

@@ -159,21 +159,6 @@ def test_expr_na(constructor_with_lazy: Any) -> None:
     compare_dicts(result_nna, expected)
 
 
-def test_drop_nulls(request: Any, constructor_with_lazy: Any) -> None:
-    if "pyarrow_table" in str(constructor_with_lazy):
-        request.applymarker(pytest.mark.xfail)
-
-    df = nw.from_native(constructor_with_lazy(data_na)).lazy()
-
-    result = df.select(nw.col("a").drop_nulls())
-    expected = {"a": [3, 2]}
-    compare_dicts(result, expected)
-
-    result = df.select(df.collect()["a"].drop_nulls())
-    expected = {"a": [3, 2]}
-    compare_dicts(result, expected)
-
-
 @pytest.mark.parametrize(
     ("drop", "left"),
     [

utils/check_backend_completeness.py

@@ -14,7 +14,6 @@
 MISSING = [
     "DataFrame.iter_rows",
     "DataFrame.pipe",
-    "Series.drop_nulls",
     "Series.from_iterable",
     "Series.is_between",
     "Series.quantile",