Merge branch 'main' into tpcds

Author: nickgerrets

duckdb

@@ -21,9 +21,11 @@
 
 #include "duckdb/parser/expression/comparison_expression.hpp"
 
-#include "substrait/plan.pb.h"
-#include "google/protobuf/util/json_util.h"
 #include "duckdb/main/client_data.hpp"
+#include "google/protobuf/util/json_util.h"
+#include "substrait/plan.pb.h"
+
+#include "duckdb/main/relation/table_relation.hpp"
 
 namespace duckdb {
 const std::unordered_map<std::string, std::string> SubstraitToDuckDB::function_names_remap = {
@@ -660,21 +662,22 @@ int32_t SkipColumnNames(const LogicalType &type) {
 	return columns_to_skip;
 }
 
-Relation *GetProjectionRelation(Relation &relation, string &error) {
+Relation *GetProjectionOrTableRelation(Relation &relation, string &error) {
 	error += RelationTypeToString(relation.type);
 	switch (relation.type) {
+	case RelationType::TABLE_RELATION:
 	case RelationType::PROJECTION_RELATION:
 		error += " -> ";
 		return &relation;
 	case RelationType::LIMIT_RELATION:
 		error += " -> ";
-		return GetProjectionRelation(*relation.Cast<LimitRelation>().child, error);
+		return GetProjectionOrTableRelation(*relation.Cast<LimitRelation>().child, error);
 	case RelationType::ORDER_RELATION:
 		error += " -> ";
-		return GetProjectionRelation(*relation.Cast<OrderRelation>().child, error);
+		return GetProjectionOrTableRelation(*relation.Cast<OrderRelation>().child, error);
 	case RelationType::SET_OPERATION_RELATION:
 		error += " -> ";
-		return GetProjectionRelation(*relation.Cast<SetOpRelation>().right, error);
+		return GetProjectionOrTableRelation(*relation.Cast<SetOpRelation>().right, error);
 	default:
 		throw NotImplementedException(
 		    "Relation %s is not yet implemented as a possible root chain type of from_substrait function", error);
@@ -683,15 +686,20 @@ Relation *GetProjectionRelation(Relation &relation, string &error) {
 
 shared_ptr<Relation> SubstraitToDuckDB::TransformRootOp(const substrait::RelRoot &sop) {
 	vector<string> aliases;
-	auto column_names = sop.names();
+	const auto &column_names = sop.names();
 	vector<unique_ptr<ParsedExpression>> expressions;
 	int id = 1;
 	auto child = TransformOp(sop.input());
 	string error;
-	auto first_projection = GetProjectionRelation(*child, error);
-	auto &columns = first_projection->Cast<ProjectionRelation>().columns;
+	auto first_projection_or_table = GetProjectionOrTableRelation(*child, error);
+	vector<ColumnDefinition> *column_definitions;
+	if (first_projection_or_table->type == RelationType::PROJECTION_RELATION) {
+		column_definitions = &first_projection_or_table->Cast<ProjectionRelation>().columns;
+	} else {
+		column_definitions = &first_projection_or_table->Cast<TableRelation>().description->columns;
+	}
 	int32_t i = 0;
-	for (auto &column : columns) {
+	for (auto &column : *column_definitions) {
 		aliases.push_back(column_names[i++]);
 		auto column_type = column.GetType();
 		i += SkipColumnNames(column.GetType());

duckdb-r

@@ -163,6 +163,8 @@ static void ToSubFunction(ClientContext &context, TableFunctionInput &data_p, Da
 		return;
 	}
 	auto new_conn = Connection(*context.db);
+	// If error(varchar) gets implemented in substrait this can be removed
+	new_conn.Query("SET scalar_subquery_error_on_multiple_rows=false;");
 
 	unique_ptr<LogicalOperator> query_plan;
 	string serialized;
@@ -195,6 +197,8 @@ static void ToJsonFunction(ClientContext &context, TableFunctionInput &data_p, D
 		return;
 	}
 	auto new_conn = Connection(*context.db);
+	// If error(varchar) gets implemented in substrait this can be removed
+	new_conn.Query("SET scalar_subquery_error_on_multiple_rows=false;");
 
 	unique_ptr<LogicalOperator> query_plan;
 	string serialized;

src/from_substrait.cpp

@@ -0,0 +1,78 @@
+# name: test/sql/test_direct_scan.test
+# description: Test that a direct table scan works
+# group: [sql]
+
+require substrait
+
+statement ok
+PRAGMA enable_verification
+
+statement ok
+create table users (user_id varchar, name varchar, paid_for_service bool);
+
+statement ok
+insert into users values ('1', 'Pedro', false);
+
+statement ok
+CALL get_substrait('FROM users')
+
+query III
+CALL from_substrait_json('{
+  "relations": [
+    {
+      "root": {
+        "input": {
+          "read": {
+            "common": {
+              "direct": {}
+            },
+            "baseSchema": {
+              "names": [
+                "user_id",
+                "name",
+                "paid_for_service"
+              ],
+              "struct": {
+                "types": [
+                  {
+                    "string": {
+                      "nullability": "NULLABILITY_NULLABLE"
+                    }
+                  },
+                  {
+                    "string": {
+                      "nullability": "NULLABILITY_NULLABLE"
+                    }
+                  },
+                  {
+                    "bool": {
+                      "nullability": "NULLABILITY_NULLABLE"
+                    }
+                  }
+                ],
+                "nullability": "NULLABILITY_REQUIRED"
+              }
+            },
+            "namedTable": {
+              "names": [
+                "users"
+              ]
+            }
+          }
+        },
+        "names": [
+          "user_id",
+          "name",
+          "paid_for_service"
+        ]
+      }
+    }
+  ],
+  "version": {
+    "minorNumber": 52,
+    "producer": "spark-substrait-gateway"
+  }
+}
+')
+----
+1	Pedro	false