Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 18 additions & 0 deletions .github/patches/extensions/sqlsmith/fix.patch
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,15 @@ index 72e8f5b..8eafec0 100644
}

return delete_statement;
@@ -302,7 +303,7 @@ void StatementGenerator::GenerateCTEs(QueryNode &node) {
}
while (RandomPercentage(20)) {
auto cte = make_uniq<CommonTableExpressionInfo>();
- cte->query = unique_ptr_cast<SQLStatement, SelectStatement>(GenerateSelect());
+ cte->query_node = std::move(unique_ptr_cast<SQLStatement, SelectStatement>(GenerateSelect())->node);
for (idx_t i = 0; i < 1 + RandomValue(10); i++) {
cte->aliases.push_back(GenerateIdentifier());
}
@@ -957,9 +958,12 @@ unique_ptr<ParsedExpression> StatementGenerator::GenerateWindowFunction(optional
case ExpressionType::WINDOW_NTILE:
case ExpressionType::WINDOW_FIRST_VALUE:
Expand Down Expand Up @@ -80,6 +89,15 @@ index 4602928..3e56aca 100644
#include "duckdb/parser/statement/select_statement.hpp"
#endif

@@ -206,7 +209,7 @@ void StatementSimplifier::Simplify(CommonTableExpressionMap &cte) {
SimplifyMap(cte.map);
for (auto &cte_child : cte.map) {
// simplify individual ctes
- Simplify(cte_child.second->query->node);
+ Simplify(cte_child.second->query_node);
}
}

@@ -357,8 +360,6 @@ void StatementSimplifier::SimplifyExpression(duckdb::unique_ptr<ParsedExpression
SimplifyChildExpression(expr, window.filter_expr);
SimplifyChildExpression(expr, window.start_expr);
Expand Down
4 changes: 1 addition & 3 deletions extension/autocomplete/transformer/transform_create_view.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -55,9 +55,7 @@ void PEGTransformerFactory::WrapRecursiveView(unique_ptr<CreateViewInfo> &info,
auto cte_info = make_uniq<CommonTableExpressionInfo>();
cte_info->aliases = info->aliases;

auto cte_select = make_uniq<SelectStatement>();
cte_select->node = std::move(inner_node);
cte_info->query = std::move(cte_select);
cte_info->query_node = std::move(inner_node);

outer_select->cte_map.map.insert(info->view_name, std::move(cte_info));

Expand Down
42 changes: 37 additions & 5 deletions extension/autocomplete/transformer/transform_select.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,12 @@
#include "duckdb/parser/tableref/at_clause.hpp"
#include "duckdb/parser/query_node/set_operation_node.hpp"
#include "duckdb/parser/tableref/pivotref.hpp"
#include "duckdb/parser/statement/insert_statement.hpp"
#include "duckdb/parser/statement/update_statement.hpp"
#include "duckdb/parser/statement/delete_statement.hpp"
#include "duckdb/parser/query_node/insert_query_node.hpp"
#include "duckdb/parser/query_node/update_query_node.hpp"
#include "duckdb/parser/query_node/delete_query_node.hpp"

namespace duckdb {

Expand Down Expand Up @@ -1597,7 +1603,15 @@ CommonTableExpressionMap PEGTransformerFactory::TransformWithClause(PEGTransform
transformer.Transform<pair<string, unique_ptr<CommonTableExpressionInfo>>>(with_statement_list[entry_idx]);

if (is_recursive) {
auto &query_node = with_entry.second->query->node;
auto &query_node = with_entry.second->query_node;
if (!query_node) {
throw ParserException("Recursive CTEs with DML statements are not supported");
}
if (query_node->type == QueryNodeType::INSERT_QUERY_NODE ||
query_node->type == QueryNodeType::UPDATE_QUERY_NODE ||
query_node->type == QueryNodeType::DELETE_QUERY_NODE) {
throw ParserException("Recursive CTEs with DML statements are not supported");
}
if (!query_node->modifiers.empty()) {
for (auto &modifier : query_node->modifiers) {
if (modifier->type == ResultModifierType::LIMIT_MODIFIER ||
Expand Down Expand Up @@ -1645,7 +1659,7 @@ PEGTransformerFactory::TransformWithStatement(PEGTransformer &transformer, optio
auto table_ref = transformer.Transform<unique_ptr<TableRef>>(list_pr.Child<ListParseResult>(5));
D_ASSERT(table_ref->type == TableReferenceType::SUBQUERY);
auto subquery_ref = unique_ptr_cast<TableRef, SubqueryRef>(std::move(table_ref));
result->query = std::move(subquery_ref->subquery);
result->query_node = std::move(subquery_ref->subquery->node);
return make_pair(cte_name, std::move(result));
}

Expand All @@ -1657,10 +1671,28 @@ unique_ptr<TableRef> PEGTransformerFactory::TransformCTEBody(PEGTransformer &tra
// CTEBodyContent <- SelectStatementInternal / Statement
auto &content_list = inner->Cast<ListParseResult>();
auto &body_choice = content_list.Child<ChoiceParseResult>(0);
if (body_choice.result->name != "SelectStatementInternal") {
throw ParserException("A CTE needs a SELECT");
if (body_choice.result->name == "SelectStatementInternal") {
auto select_statement = transformer.Transform<unique_ptr<SelectStatement>>(body_choice.result);
return make_uniq<SubqueryRef>(std::move(select_statement));
}
// DML body (INSERT / UPDATE / DELETE) - transform as a Statement and extract its QueryNode
auto sql_stmt = transformer.Transform<unique_ptr<SQLStatement>>(body_choice.result);
unique_ptr<QueryNode> query_node;
switch (sql_stmt->type) {
case StatementType::INSERT_STATEMENT:
query_node = unique_ptr_cast<InsertQueryNode, QueryNode>(std::move(sql_stmt->Cast<InsertStatement>().node));
break;
case StatementType::UPDATE_STATEMENT:
query_node = unique_ptr_cast<UpdateQueryNode, QueryNode>(std::move(sql_stmt->Cast<UpdateStatement>().node));
break;
case StatementType::DELETE_STATEMENT:
query_node = unique_ptr_cast<DeleteQueryNode, QueryNode>(std::move(sql_stmt->Cast<DeleteStatement>().node));
break;
default:
throw ParserException("A CTE body must be a SELECT, INSERT, UPDATE, or DELETE statement");
}
auto select_statement = transformer.Transform<unique_ptr<SelectStatement>>(body_choice.result);
auto select_statement = make_uniq<SelectStatement>();
select_statement->node = std::move(query_node);
return make_uniq<SubqueryRef>(std::move(select_statement));
}

Expand Down
17 changes: 12 additions & 5 deletions src/include/duckdb/parser/common_table_expression_info.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -13,25 +13,32 @@

namespace duckdb {

class QueryNode;
class SelectStatement;
class Serializer;
class Deserializer;

struct CommonTableExpressionInfo {
CommonTableExpressionInfo() = default;
~CommonTableExpressionInfo();

//! Used by deserialization: prefers query_node; if null, uses query->node.
CommonTableExpressionInfo(unique_ptr<SelectStatement> query, unique_ptr<QueryNode> query_node);

vector<string> aliases;
vector<unique_ptr<ParsedExpression>> key_targets;
vector<unique_ptr<ParsedExpression>> payload_aggregates;

unique_ptr<SelectStatement> query;
//! The root QueryNode for this CTE (SELECT, INSERT, UPDATE, or DELETE)
unique_ptr<QueryNode> query_node;
CTEMaterialize materialized = CTEMaterialize::CTE_MATERIALIZE_DEFAULT;

public:
CTEMaterialize GetMaterializedForSerialization(Serializer &serializer) const;
unique_ptr<SelectStatement> GetQueryForSerialization(Serializer &serializer) const;

void Serialize(Serializer &serializer) const;
static unique_ptr<CommonTableExpressionInfo> Deserialize(Deserializer &deserializer);
unique_ptr<CommonTableExpressionInfo> Copy();

private:
CTEMaterialize GetMaterializedForSerialization(Serializer &serializer) const;
};

} // namespace duckdb
4 changes: 2 additions & 2 deletions src/include/duckdb/parser/transformer.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -302,8 +302,8 @@ class Transformer {
string TransformAlias(duckdb_libpgquery::PGAlias *root, vector<string> &column_name_alias);
vector<string> TransformStringList(duckdb_libpgquery::PGList *list);
void TransformCTE(duckdb_libpgquery::PGWithClause &de_with_clause, CommonTableExpressionMap &cte_map);
unique_ptr<SelectStatement> TransformRecursiveCTE(duckdb_libpgquery::PGCommonTableExpr &node,
CommonTableExpressionInfo &info);
unique_ptr<QueryNode> TransformRecursiveCTE(duckdb_libpgquery::PGCommonTableExpr &node,
CommonTableExpressionInfo &info);

unique_ptr<ParsedExpression> TransformUnaryOperator(const string &op, unique_ptr<ParsedExpression> child);
unique_ptr<ParsedExpression> TransformBinaryOperator(string op, unique_ptr<ParsedExpression> left,
Expand Down
14 changes: 12 additions & 2 deletions src/include/duckdb/storage/serialization/nodes.json
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,10 @@
{
"id": 101,
"name": "query",
"type": "SelectStatement*"
"type": "SelectStatement*",
"version": "v2.0.0",
"status": "read_only",
"serialize_property": "GetQueryForSerialization(serializer)"
},
{
"id": 102,
Expand All @@ -30,8 +33,15 @@
"id": 104,
"name": "payload_aggregates",
"type": "vector<ParsedExpression*>"
},
{
"id": 105,
"name": "query_node",
"type": "QueryNode*",
"version": "v2.0.0"
}
]
],
"constructor": ["query", "query_node"]
},
{
"class": "CommonTableExpressionMap",
Expand Down
2 changes: 1 addition & 1 deletion src/main/appender.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -483,7 +483,7 @@ unique_ptr<SQLStatement> BaseAppender::ParseStatement(unique_ptr<TableRef> table

// Create the CTE info.
auto cte_info = make_uniq<CommonTableExpressionInfo>();
cte_info->query = std::move(cte_select);
cte_info->query_node = std::move(cte_select->node);
cte_info->materialized = CTEMaterialize::CTE_MATERIALIZE_NEVER;

// Add the appender data as a CTE to the CTE map of the statement.
Expand Down
2 changes: 1 addition & 1 deletion src/main/relation/query_relation.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,7 @@ BoundStatement QueryRelation::Bind(Binder &binder) {
select->node = std::move(select_node);

auto cte_info = make_uniq<CommonTableExpressionInfo>();
cte_info->query = std::move(select);
cte_info->query_node = std::move(select->node);

auto subquery = make_uniq<SubqueryRef>(std::move(select_stmt), "query_relation");
auto top_level_select = make_uniq<SelectStatement>();
Expand Down
21 changes: 21 additions & 0 deletions src/optimizer/cte_inlining.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -87,6 +87,23 @@ bool CTEInlining::EndsInAggregateOrDistinct(const LogicalOperator &op) {
return false;
}

static bool HasSideEffects(const LogicalOperator &op) {
switch (op.type) {
case LogicalOperatorType::LOGICAL_INSERT:
case LogicalOperatorType::LOGICAL_UPDATE:
case LogicalOperatorType::LOGICAL_DELETE:
return true;
default:
break;
}
for (auto &child : op.children) {
if (HasSideEffects(*child)) {
return true;
}
}
return false;
}

void CTEInlining::TryInlining(unique_ptr<LogicalOperator> &op) {
if (op->type == LogicalOperatorType::LOGICAL_PREPARE) {
// we are in a prepare statement, if we have to copy an operator during inlining,
Expand All @@ -104,6 +121,10 @@ void CTEInlining::TryInlining(unique_ptr<LogicalOperator> &op) {
auto &cte = op->Cast<LogicalMaterializedCTE>();
auto ref_count = CountCTEReferences(*op, cte.table_index);
if (ref_count == 0) {
if (HasSideEffects(*cte.children[0])) {
// DML CTEs must always execute for side effects even when unreferenced
return;
}
// this CTE is not referenced, we can remove it
op = std::move(op->children[1]);
return;
Expand Down
4 changes: 3 additions & 1 deletion src/parser/parsed_expression_iterator.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -372,7 +372,9 @@ void ParsedExpressionIterator::EnumerateQueryNodeChildren(
}

for (auto &kv : node.cte_map.map) {
EnumerateQueryNodeChildren(*kv.second->query->node, expr_callback, ref_callback);
if (kv.second->query_node) {
EnumerateQueryNodeChildren(*kv.second->query_node, expr_callback, ref_callback);
}
}
}

Expand Down
20 changes: 11 additions & 9 deletions src/parser/query_node.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -5,10 +5,6 @@
#include "duckdb/parser/query_node/recursive_cte_node.hpp"
#include "duckdb/parser/query_node/cte_node.hpp"
#include "duckdb/common/limits.hpp"
#include "duckdb/common/serializer/serializer.hpp"
#include "duckdb/common/serializer/deserializer.hpp"
#include "duckdb/parser/statement/select_statement.hpp"

namespace duckdb {

CommonTableExpressionMap::CommonTableExpressionMap() {
Expand All @@ -27,7 +23,9 @@ CommonTableExpressionMap CommonTableExpressionMap::Copy() const {
for (auto &al : kv.second->payload_aggregates) {
kv_info->payload_aggregates.push_back(al->Copy());
}
kv_info->query = unique_ptr_cast<SQLStatement, SelectStatement>(kv.second->query->Copy());
if (kv.second->query_node) {
kv_info->query_node = kv.second->query_node->Copy();
}
kv_info->materialized = kv.second->materialized;
res.map[kv.first] = std::move(kv_info);
}
Expand All @@ -42,7 +40,7 @@ string CommonTableExpressionMap::ToString() const {
// check if there are any recursive CTEs
bool has_recursive = false;
for (auto &kv : map) {
if (kv.second->query->node->type == QueryNodeType::RECURSIVE_CTE_NODE) {
if (kv.second->query_node && kv.second->query_node->type == QueryNodeType::RECURSIVE_CTE_NODE) {
has_recursive = true;
break;
}
Expand Down Expand Up @@ -89,7 +87,8 @@ string CommonTableExpressionMap::ToString() const {
} else {
result += " AS (";
}
result += cte.query->ToString();
D_ASSERT(cte.query_node);
result += cte.query_node->ToString();
result += ")";
first_cte = false;
}
Expand Down Expand Up @@ -169,7 +168,8 @@ bool QueryNode::Equals(const QueryNode *other) const {
if (!ParsedExpression::ListEquals(entry.second->payload_aggregates, other_entry->second->payload_aggregates)) {
return false;
}
if (!entry.second->query->Equals(*other->cte_map.map.at(entry.first)->query)) {
if (!entry.second->query_node ||
!entry.second->query_node->Equals(other->cte_map.map.at(entry.first)->query_node.get())) {
return false;
}
}
Expand All @@ -191,7 +191,9 @@ void QueryNode::CopyProperties(QueryNode &other) const {
for (auto &agg : kv.second->payload_aggregates) {
kv_info->payload_aggregates.push_back(agg->Copy());
}
kv_info->query = unique_ptr_cast<SQLStatement, SelectStatement>(kv.second->query->Copy());
if (kv.second->query_node) {
kv_info->query_node = kv.second->query_node->Copy();
}
kv_info->materialized = kv.second->materialized;
other.cte_map.map[kv.first] = std::move(kv_info);
}
Expand Down
Loading
Loading