snowpark-connect 0.22.1__py3-none-any.whl → 0.24.0__py3-none-any.whl

snowflake/snowpark_connect/relation/map_aggregate.py

@@ -153,11 +153,63 @@ def map_pivot_aggregate(
                     used_columns.add(mapped_col[0].snowpark_name)
 
     if len(columns.grouping_expressions()) == 0:
-        result = (
-            input_df_actual.select(*used_columns)
-            .pivot(pivot_column[1].col, pivot_values if pivot_values else None)
-            .agg(*columns.aggregation_expressions(unalias=True))
-        )
+        # Snowpark doesn't support multiple aggregations in pivot without groupBy
+        # So we need to perform each aggregation separately and then combine results
+        if len(columns.aggregation_expressions(unalias=True)) > 1:
+            agg_expressions = columns.aggregation_expressions(unalias=True)
+            agg_metadata = columns.aggregation_columns
+            num_agg_functions = len(agg_expressions)
+
+            spark_names = []
+            pivot_results = []
+            for i, agg_expr in enumerate(agg_expressions):
+                pivot_result = (
+                    input_df_actual.select(*used_columns)
+                    .pivot(pivot_column[1].col, pivot_values if pivot_values else None)
+                    .agg(agg_expr)
+                )
+                for col_name in pivot_result.columns:
+                    spark_names.append(
+                        f"{pivot_column_name(col_name)}_{agg_metadata[i].spark_name}"
+                    )
+                pivot_results.append(pivot_result)
+
+            result = pivot_results[0]
+            for pivot_result in pivot_results[1:]:
+                result = result.cross_join(pivot_result)
+
+            pivot_columns_per_agg = len(pivot_results[0].columns)
+            reordered_spark_names = []
+            reordered_snowpark_names = []
+            reordered_types = []
+            column_selectors = []
+
+            for pivot_idx in range(pivot_columns_per_agg):
+                for agg_idx in range(num_agg_functions):
+                    current_pos = agg_idx * pivot_columns_per_agg + pivot_idx
+                    if current_pos < len(spark_names):
+                        idx = current_pos + 1  # 1-based indexing for Snowpark
+                        reordered_spark_names.append(spark_names[current_pos])
+                        reordered_snowpark_names.append(f"${idx}")
+                        reordered_types.append(
+                            result.schema.fields[current_pos].datatype
+                        )
+                        column_selectors.append(snowpark_fn.col(f"${idx}"))
+
+            return DataFrameContainer.create_with_column_mapping(
+                dataframe=result.select(*column_selectors),
+                spark_column_names=reordered_spark_names,
+                snowpark_column_names=reordered_snowpark_names,
+                column_qualifiers=[[]] * len(reordered_spark_names),
+                parent_column_name_map=input_container.column_map,
+                snowpark_column_types=reordered_types,
+            )
+        else:
+            result = (
+                input_df_actual.select(*used_columns)
+                .pivot(pivot_column[1].col, pivot_values if pivot_values else None)
+                .agg(*columns.aggregation_expressions(unalias=True))
+            )
     else:
         result = (
             input_df_actual.group_by(*columns.grouping_expressions())

snowflake/snowpark_connect/relation/map_column_ops.py

@@ -266,6 +266,7 @@ def map_project(
 
             aliased_col = mapper.col.alias(snowpark_column)
             select_list.append(aliased_col)
+
             new_snowpark_columns.append(snowpark_column)
             new_spark_columns.append(spark_name)
             column_types.extend(mapper.types)
@@ -422,6 +423,7 @@ def map_sort(
     # TODO: sort.isglobal.
     if not order_specified:
         ascending = None
+
     result = input_df.sort(cols, ascending=ascending)
 
     return DataFrameContainer(
@@ -1075,14 +1077,12 @@ def map_group_map(
     snowpark_grouping_expressions: list[snowpark.Column] = []
     typer = ExpressionTyper(input_df)
     group_name_list: list[str] = []
-    qualifiers = []
     for exp in grouping_expressions:
         new_name, snowpark_column = map_single_column_expression(
             exp, input_container.column_map, typer
         )
         snowpark_grouping_expressions.append(snowpark_column.col)
         group_name_list.append(new_name)
-        qualifiers.append(snowpark_column.get_qualifiers())
     if rel.group_map.func.python_udf is None:
         raise ValueError("group_map relation without python udf is not supported")
 
@@ -1124,13 +1124,14 @@ def map_group_map(
         result = input_df.group_by(*snowpark_grouping_expressions).apply_in_pandas(
             callable_func, output_type
         )
-
-    qualifiers.extend([[]] * (len(result.columns) - len(group_name_list)))
+    # The UDTF `apply_in_pandas` generates a new table whose output schema
+    # can be entirely different from that of the input Snowpark DataFrame.
+    # As a result, the output DataFrame should not use qualifiers based on the input group by columns.
     return DataFrameContainer.create_with_column_mapping(
         dataframe=result,
         spark_column_names=[field.name for field in output_type],
         snowpark_column_names=result.columns,
-        column_qualifiers=qualifiers,
+        column_qualifiers=None,
         parent_column_name_map=input_container.column_map,
     )
 

snowflake/snowpark_connect/relation/map_extension.py

@@ -347,6 +347,13 @@ def map_aggregate(
     raw_groupings: list[tuple[str, TypedColumn]] = []
     raw_aggregations: list[tuple[str, TypedColumn]] = []
 
+    if not is_group_by_all:
+        raw_groupings = [_map_column(exp) for exp in aggregate.grouping_expressions]
+
+    # Set the current grouping columns in context for grouping_id() function
+    grouping_spark_columns = [spark_name for spark_name, _ in raw_groupings]
+    set_current_grouping_columns(grouping_spark_columns)
+
     agg_count = get_sql_aggregate_function_count()
     for exp in aggregate.aggregate_expressions:
         col = _map_column(exp)
@@ -359,13 +366,6 @@ def map_aggregate(
             else:
                 agg_count = new_agg_count
 
-    if not is_group_by_all:
-        raw_groupings = [_map_column(exp) for exp in aggregate.grouping_expressions]
-
-    # Set the current grouping columns in context for grouping_id() function
-    grouping_spark_columns = [spark_name for spark_name, _ in raw_groupings]
-    set_current_grouping_columns(grouping_spark_columns)
-
     # Now create column name lists and assign aliases.
     # In case of GROUP BY ALL, even though groupings are a subset of aggregations,
     # they will have their own aliases so we can drop them later.
@@ -374,23 +374,31 @@ def map_aggregate(
     snowpark_columns: list[str] = []
     snowpark_column_types: list[snowpark_types.DataType] = []
 
-    def _add_column(spark_name: str, snowpark_column: TypedColumn) -> snowpark.Column:
-        alias = make_column_names_snowpark_compatible(
-            [spark_name], plan_id, len(spark_columns)
-        )[0]
+    # Use grouping columns directly without aliases
+    groupings = [col.col for _, col in raw_groupings]
+
+    # Create aliases only for aggregation columns
+    aggregations = []
+    for i, (spark_name, snowpark_column) in enumerate(raw_aggregations):
+        alias = make_column_names_snowpark_compatible([spark_name], plan_id, i)[0]
 
         spark_columns.append(spark_name)
         snowpark_columns.append(alias)
         snowpark_column_types.append(snowpark_column.typ)
 
-        return snowpark_column.col.alias(alias)
-
-    groupings = [_add_column(name, col) for name, col in raw_groupings]
-    aggregations = [_add_column(name, col) for name, col in raw_aggregations]
+        aggregations.append(snowpark_column.col.alias(alias))
 
     match aggregate.group_type:
         case snowflake_proto.Aggregate.GROUP_TYPE_GROUPBY:
-            result = input_df.group_by(groupings)
+            if groupings:
+                # Normal GROUP BY with explicit grouping columns
+                result = input_df.group_by(groupings)
+            else:
+                # No explicit GROUP BY - this is an aggregate over the entire table
+                # Use a dummy constant that will be excluded from the final result
+                result = input_df.with_column(
+                    "__dummy_group__", snowpark_fn.lit(1)
+                ).group_by("__dummy_group__")
         case snowflake_proto.Aggregate.GROUP_TYPE_ROLLUP:
             result = input_df.rollup(groupings)
         case snowflake_proto.Aggregate.GROUP_TYPE_CUBE:
@@ -410,28 +418,54 @@ def map_aggregate(
                 f"Unsupported GROUP BY type: {other}"
             )
 
-    result = result.agg(*aggregations)
+    result = result.agg(*aggregations, exclude_grouping_columns=True)
+
+    # If we added a dummy grouping column, make sure it's excluded
+    if not groupings and "__dummy_group__" in result.columns:
+        result = result.drop("__dummy_group__")
+
+    # Apply HAVING condition if present
+    if aggregate.HasField("having_condition"):
+        from snowflake.snowpark_connect.expression.hybrid_column_map import (
+            create_hybrid_column_map_for_having,
+        )
+
+        # Create aggregated DataFrame column map
+        aggregated_column_map = DataFrameContainer.create_with_column_mapping(
+            dataframe=result,
+            spark_column_names=spark_columns,
+            snowpark_column_names=snowpark_columns,
+            snowpark_column_types=snowpark_column_types,
+        ).column_map
+
+        # Create hybrid column map that can resolve both input and aggregate contexts
+        hybrid_map = create_hybrid_column_map_for_having(
+            input_df=input_df,
+            input_column_map=input_container.column_map,
+            aggregated_df=result,
+            aggregated_column_map=aggregated_column_map,
+            aggregate_expressions=list(aggregate.aggregate_expressions),
+            grouping_expressions=list(aggregate.grouping_expressions),
+            spark_columns=spark_columns,
+            raw_aggregations=raw_aggregations,
+        )
+
+        # Map the HAVING condition using hybrid resolution
+        _, having_column = hybrid_map.resolve_expression(aggregate.having_condition)
+
+        # Apply the HAVING filter
+        result = result.filter(having_column.col)
 
     if aggregate.group_type == snowflake_proto.Aggregate.GROUP_TYPE_GROUPING_SETS:
         # Immediately drop extra columns. Unlike other GROUP BY operations,
         # grouping sets don't allow ORDER BY with columns that aren't in the aggregate list.
-        result = result.select(result.columns[-len(spark_columns) :])
+        result = result.select(result.columns[-len(aggregations) :])
 
-    # Build a parent column map that includes groupings.
-    result_container = DataFrameContainer.create_with_column_mapping(
+    # Return only aggregation columns in the column map
+    return DataFrameContainer.create_with_column_mapping(
         dataframe=result,
         spark_column_names=spark_columns,
         snowpark_column_names=snowpark_columns,
         snowpark_column_types=snowpark_column_types,
-    )
-
-    # Drop the groupings.
-    grouping_count = len(groupings)
-
-    return DataFrameContainer.create_with_column_mapping(
-        result.drop(snowpark_columns[:grouping_count]),
-        spark_columns[grouping_count:],
-        snowpark_columns[grouping_count:],
-        snowpark_column_types[grouping_count:],
-        parent_column_name_map=result_container.column_map,
+        parent_column_name_map=input_df._column_map,
     )

snowflake/snowpark_connect/relation/map_local_relation.py

@@ -4,6 +4,7 @@
 
 import json
 import re
+from json import JSONDecodeError
 
 import numpy as np
 import pyarrow as pa
@@ -19,6 +20,7 @@ from snowflake.snowpark_connect.column_name_handler import (
 )
 from snowflake.snowpark_connect.dataframe_container import DataFrameContainer
 from snowflake.snowpark_connect.type_mapping import (
+    get_python_sql_utils_class,
     map_json_schema_to_snowpark,
     map_pyarrow_to_snowpark_types,
     map_simple_types,
@@ -34,7 +36,12 @@ def parse_local_relation_schema_string(rel: relation_proto.Relation):
     # schema_str can be a dict, or just a type string, e.g. INTEGER.
     schema_str = rel.local_relation.schema
     assert schema_str
-    schema_dict = json.loads(schema_str)
+    try:
+        schema_dict = json.loads(schema_str)
+    except JSONDecodeError:
+        # Legacy scala clients sends unparsed struct type strings like "struct<id:bigint,a:int,b:double>"
+        spark_datatype = get_python_sql_utils_class().parseDataType(schema_str)
+        schema_dict = json.loads(spark_datatype.json())
 
     column_metadata = {}
     if isinstance(schema_dict, dict):

snowflake/snowpark_connect/relation/map_row_ops.py

@@ -2,6 +2,7 @@
 # Copyright (c) 2012-2025 Snowflake Computing Inc. All rights reserved.
 #
 
+
 import pyspark.sql.connect.proto.expressions_pb2 as expressions_proto
 import pyspark.sql.connect.proto.relations_pb2 as relation_proto
 from pyspark.errors.exceptions.base import AnalysisException, IllegalArgumentException
@@ -551,6 +552,7 @@ def map_filter(
     _, condition = map_single_column_expression(
         rel.filter.condition, input_container.column_map, typer
     )
+
     result = input_df.filter(condition.col)
 
     return DataFrameContainer(

snowflake/snowpark_connect/relation/map_sql.py

@@ -77,6 +77,9 @@ from ..expression.map_sql_expression import (
 from ..utils.identifiers import spark_to_sf_single_id
 
 _ctes = ContextVar[dict[str, relation_proto.Relation]]("_ctes", default={})
+_having_condition = ContextVar[expressions_proto.Expression | None](
+    "_having_condition", default=None
+)
 
 
 def _is_sql_select_statement_helper(sql_string: str) -> bool:
@@ -1146,6 +1149,7 @@ def map_logical_plan_relation(
                             grouping_expressions=grouping_expressions,
                             aggregate_expressions=aggregate_expressions,
                             grouping_sets=grouping_sets,
+                            having_condition=_having_condition.get(),
                         )
                     )
                 )
@@ -1389,12 +1393,25 @@ def map_logical_plan_relation(
                 )
             )
         case "UnresolvedHaving":
-            proto = relation_proto.Relation(
-                filter=relation_proto.Filter(
-                    input=map_logical_plan_relation(rel.child()),
-                    condition=map_logical_plan_expression(rel.havingCondition()),
+            # Store the having condition in context and process the child aggregate
+            child_relation = rel.child()
+            if str(child_relation.getClass().getSimpleName()) != "Aggregate":
+                raise SnowparkConnectNotImplementedError(
+                    "UnresolvedHaving can only be applied to Aggregate relations"
                 )
-            )
+
+            # Store having condition in a context variable for the Aggregate case to pick up
+            having_condition = map_logical_plan_expression(rel.havingCondition())
+
+            # Store in thread-local context (similar to how _ctes works)
+            token = _having_condition.set(having_condition)
+
+            try:
+                # Recursively call map_logical_plan_relation on the child Aggregate
+                # The Aggregate case will pick up the having condition from context
+                proto = map_logical_plan_relation(child_relation, plan_id)
+            finally:
+                _having_condition.reset(token)
         case "UnresolvedHint":
             proto = relation_proto.Relation(
                 hint=relation_proto.Hint(

snowflake/snowpark_connect/relation/read/map_read.py

@@ -95,7 +95,8 @@ def map_read(
             if len(rel.read.data_source.paths) > 0:
                 # Normalize paths to ensure consistent behavior
                 clean_source_paths = [
-                    str(Path(path)) for path in rel.read.data_source.paths
+                    path.rstrip("/") if is_cloud_path(path) else str(Path(path))
+                    for path in rel.read.data_source.paths
                 ]
 
                 result = _read_file(

snowflake/snowpark_connect/relation/read/map_read_parquet.py

@@ -54,10 +54,17 @@ def map_read_parquet(
     if len(paths) == 1:
         df = _read_parquet_with_partitions(session, reader, paths[0])
     else:
+        is_merge_schema = options.config.get("mergeschema")
         df = _read_parquet_with_partitions(session, reader, paths[0])
+        schema_cols = df.columns
         for p in paths[1:]:
             reader._user_schema = None
-            df = df.union_all(_read_parquet_with_partitions(session, reader, p))
+            df = df.union_all_by_name(
+                _read_parquet_with_partitions(session, reader, p),
+                allow_missing_columns=True,
+            )
+        if not is_merge_schema:
+            df = df.select(*schema_cols)
 
     renamed_df, snowpark_column_names = rename_columns_as_snowflake_standard(
         df, rel.common.plan_id

snowflake/snowpark_connect/relation/read/reader_config.py

@@ -398,3 +398,12 @@ class ParquetReaderConfig(ReaderWriterConfig):
             ),
             options,
         )
+
+    def convert_to_snowpark_args(self) -> dict[str, Any]:
+        snowpark_args = super().convert_to_snowpark_args()
+
+        # Should be determined by spark.sql.parquet.binaryAsString, but currently Snowpark Connect only supports
+        # the default value (false). TODO: Add support for spark.sql.parquet.binaryAsString equal to "true".
+        snowpark_args["BINARY_AS_TEXT"] = False
+
+        return snowpark_args