Merge branch 'main' into simp-pandas-group-by

Author: dangotbanned

narwhals/_duckdb/utils.py

@@ -304,15 +304,15 @@ def generate_partition_by_sql(*partition_by: str | Expression) -> str:
 
 
 def generate_order_by_sql(
-    *order_by: str | Expression, ascending: bool, nulls_first: bool
+    *order_by: str | Expression, descending: bool, nulls_last: bool
 ) -> str:
     if not order_by:
         return ""
-    nulls = "nulls first" if nulls_first else "nulls last"
-    if ascending:
-        by_sql = ", ".join([f"{parse_into_expression(x)} asc {nulls}" for x in order_by])
-    else:
+    nulls = "nulls last" if nulls_last else "nulls first"
+    if descending:
         by_sql = ", ".join([f"{parse_into_expression(x)} desc {nulls}" for x in order_by])
+    else:
+        by_sql = ", ".join([f"{parse_into_expression(x)} asc {nulls}" for x in order_by])
     return f"order by {by_sql}"
 
 
@@ -335,9 +335,7 @@ def window_expression(
         msg = f"DuckDB>=1.3.0 is required for this operation. Found: DuckDB {duckdb.__version__}"
         raise NotImplementedError(msg) from exc
     pb = generate_partition_by_sql(*partition_by)
-    ob = generate_order_by_sql(
-        *order_by, ascending=not descending, nulls_first=not nulls_last
-    )
+    ob = generate_order_by_sql(*order_by, descending=descending, nulls_last=nulls_last)
 
     if rows_start and rows_end:
         rows = f"rows between {rows_start} and {rows_end}"

tests/expr_and_series/dt/truncate_test.py

@@ -53,11 +53,6 @@ def test_truncate(
     every: str,
     expected: list[datetime],
 ) -> None:
-    if any(x in str(constructor) for x in ("sqlframe", "pyspark")):
-        # TODO(marco): investigate pyspark, it also localizes to UTC here.
-        request.applymarker(
-            pytest.mark.xfail(reason="https://github.com/eakmanrq/sqlframe/issues/383")
-        )
     if every.endswith("ns") and any(
         x in str(constructor) for x in ("polars", "duckdb", "pyspark", "ibis")
     ):
@@ -109,11 +104,10 @@ def test_truncate_multiples(
     every: str,
     expected: list[datetime],
 ) -> None:
-    if any(x in str(constructor) for x in ("sqlframe", "cudf", "pyspark", "duckdb")):
+    if any(x in str(constructor) for x in ("cudf", "pyspark", "duckdb")):
         # Reasons:
-        # - sqlframe: https://github.com/eakmanrq/sqlframe/issues/383
         # - cudf: https://github.com/rapidsai/cudf/issues/18654
-        # - pyspark: Only multiple 1 is currently supported
+        # - pyspark/sqlframe: Only multiple 1 is currently supported
         request.applymarker(pytest.mark.xfail())
     if every.endswith("ns") and any(
         x in str(constructor) for x in ("polars", "duckdb", "ibis")

tests/utils.py

@@ -77,6 +77,10 @@ def assert_equal_data(result: Any, expected: Mapping[str, Any]) -> None:
         hasattr(result, "_compliant_frame")
         and result._compliant_frame._implementation is Implementation.IBIS
     )
+    is_spark_like = (
+        hasattr(result, "_compliant_frame")
+        and result._compliant_frame._implementation.is_spark_like()
+    )
     if is_duckdb:
         result = from_native(result.to_native().arrow())
     if is_ibis:
@@ -122,6 +126,17 @@ def assert_equal_data(result: Any, expected: Mapping[str, Any]) -> None:
                 are_equivalent_values = pd.isna(rhs)
             elif type(lhs) is date and type(rhs) is datetime:
                 are_equivalent_values = datetime(lhs.year, lhs.month, lhs.day) == rhs
+            elif (
+                is_spark_like
+                and isinstance(lhs, datetime)
+                and isinstance(rhs, datetime)
+                and rhs.tzinfo is None
+                and lhs.tzinfo
+            ):
+                # PySpark converts timezone-naive to timezone-aware by default in many cases.
+                # For now, we just assert that the local result matches the expected one.
+                # https://github.com/narwhals-dev/narwhals/issues/2793
+                are_equivalent_values = lhs.replace(tzinfo=None) == rhs
             else:
                 are_equivalent_values = lhs == rhs
 

tests/v1_test.py

@@ -396,7 +396,7 @@ def test_with_row_index(constructor: Constructor) -> None:
 
     frame = nw_v1.from_native(constructor(data))
 
-    msg = r".*argument after \* must be an iterable, not NoneType$"
+    msg = r"argument after \* must be an iterable, not NoneType|has no len"
     context = (
         pytest.raises(TypeError, match=msg)
         if any(x in str(constructor) for x in ("duckdb", "pyspark"))