perf: avoid repeating constant iteration (#2079)

* perf: avoid repeating constant iteration

* fix: revert 2 pandas calls, fix incorrect annotation

I wouldn't have tried this if I knew `tuple` wasn't allowed https://github.com/narwhals-dev/narwhals/actions/runs/13486748618/job/37678868035?pr=2079

Author: dangotbanned

narwhals/_arrow/expr.py

@@ -113,7 +113,7 @@ def func(df: ArrowDataFrame) -> list[ArrowSeries]:
             func,
             depth=0,
             function_name="col",
-            evaluate_output_names=lambda _df: list(column_names),
+            evaluate_output_names=lambda _df: column_names,
             alias_output_names=None,
             backend_version=backend_version,
             version=version,

narwhals/_dask/expr.py

@@ -108,7 +108,7 @@ def func(df: DaskLazyFrame) -> list[dx.Series]:
             func,
             depth=0,
             function_name="col",
-            evaluate_output_names=lambda _df: list(column_names),
+            evaluate_output_names=lambda _df: column_names,
             alias_output_names=None,
             backend_version=backend_version,
             version=version,

narwhals/_duckdb/expr.py

@@ -87,7 +87,7 @@ def func(_: DuckDBLazyFrame) -> list[duckdb.Expression]:
         return cls(
             func,
             function_name="col",
-            evaluate_output_names=lambda _df: list(column_names),
+            evaluate_output_names=lambda _df: column_names,
             alias_output_names=None,
             backend_version=backend_version,
             version=version,

narwhals/_pandas_like/dataframe.py

@@ -277,7 +277,7 @@ def __getitem__(
                 return self._from_native_frame(
                     select_columns_by_name(
                         self._native_frame,
-                        cast("Sequence[str] | _1DArray", item),
+                        cast("list[str] | _1DArray", item),
                         self._backend_version,
                         self._implementation,
                     ),

narwhals/_pandas_like/expr.py

@@ -140,7 +140,7 @@ def func(df: PandasLikeDataFrame) -> list[PandasLikeSeries]:
             func,
             depth=0,
             function_name="col",
-            evaluate_output_names=lambda _df: list(column_names),
+            evaluate_output_names=lambda _df: column_names,
             alias_output_names=None,
             implementation=implementation,
             backend_version=backend_version,

narwhals/_pandas_like/utils.py

@@ -7,7 +7,6 @@
 from typing import TYPE_CHECKING
 from typing import Any
 from typing import Iterable
-from typing import Sequence
 from typing import TypeVar
 from typing import cast
 
@@ -765,7 +764,7 @@ def calculate_timestamp_date(s: pd.Series[int], time_unit: str) -> pd.Series[int
 
 def select_columns_by_name(
     df: T,
-    column_names: Sequence[str] | _1DArray,
+    column_names: list[str] | _1DArray,  # NOTE: Cannot be a tuple!
     backend_version: tuple[int, ...],
     implementation: Implementation,
 ) -> T:

narwhals/_spark_like/expr.py

@@ -128,7 +128,7 @@ def func(df: SparkLikeLazyFrame) -> list[Column]:
         return cls(
             func,
             function_name="col",
-            evaluate_output_names=lambda _df: list(column_names),
+            evaluate_output_names=lambda _df: column_names,
             alias_output_names=None,
             backend_version=backend_version,
             version=version,

narwhals/expr.py

@@ -1535,9 +1535,10 @@ def over(self: Self, *keys: str | Iterable[str]) -> Self:
         if self._metadata.is_changes_length():
             msg = "`.over()` can not be used for expressions which change length."
             raise LengthChangingExprError(msg)
+        flattened = flatten(keys)
         return self.__class__(
             lambda plx: self._to_compliant_expr(plx).over(
-                flatten(keys), kind=self._metadata.kind
+                flattened, kind=self._metadata.kind
             ),
             self._metadata.with_kind(ExprKind.TRANSFORM),
         )

narwhals/selectors.py

@@ -94,8 +94,9 @@ def by_dtype(*dtypes: DType | type[DType] | Iterable[DType | type[DType]]) -> Se
         a: [[2,4]]
         c: [[8.2,4.6]]
     """
+    flattened = flatten(dtypes)
     return Selector(
-        lambda plx: plx.selectors.by_dtype(flatten(dtypes)), ExprMetadata.selector()
+        lambda plx: plx.selectors.by_dtype(flattened), ExprMetadata.selector()
     )