ClickHouse/dbms/Interpreters/ExecuteScalarSubqueriesVisitor.cpp

182 lines
6.3 KiB
C++
Raw Normal View History

#include <Parsers/ASTFunction.h>
#include <Parsers/ASTLiteral.h>
#include <Parsers/ASTSubquery.h>
2019-01-16 17:26:14 +00:00
#include <Parsers/ASTSelectQuery.h>
#include <Parsers/ASTTablesInSelectQuery.h>
#include <Parsers/ASTExpressionList.h>
#include <Interpreters/Context.h>
#include <Interpreters/misc.h>
#include <Interpreters/InterpreterSelectWithUnionQuery.h>
#include <Interpreters/ExecuteScalarSubqueriesVisitor.h>
2019-06-14 17:15:30 +00:00
#include <Interpreters/addTypeConversionToAST.h>
#include <DataStreams/IBlockInputStream.h>
2019-10-19 20:36:35 +00:00
#include <DataStreams/materializeBlock.h>
#include <DataTypes/DataTypeAggregateFunction.h>
2019-10-19 20:36:35 +00:00
#include <DataTypes/DataTypeTuple.h>
2019-10-19 20:36:35 +00:00
#include <Columns/ColumnTuple.h>
2019-06-14 17:15:30 +00:00
#include <IO/WriteHelpers.h>
namespace DB
{
namespace ErrorCodes
{
extern const int INCORRECT_RESULT_OF_SCALAR_SUBQUERY;
extern const int TOO_MANY_ROWS;
}
2019-01-09 16:16:59 +00:00
bool ExecuteScalarSubqueriesMatcher::needChildVisit(ASTPtr & node, const ASTPtr & child)
{
/// Processed
2019-03-11 13:22:51 +00:00
if (node->as<ASTSubquery>() || node->as<ASTFunction>())
return false;
/// Don't descend into subqueries in FROM section
2019-03-11 13:22:51 +00:00
if (node->as<ASTTableExpression>())
return false;
2019-03-11 13:22:51 +00:00
if (node->as<ASTSelectQuery>())
2019-01-09 16:16:59 +00:00
{
/// Do not go to FROM, JOIN, UNION.
2019-03-11 13:22:51 +00:00
if (child->as<ASTTableExpression>() || child->as<ASTSelectQuery>())
2019-01-09 16:16:59 +00:00
return false;
}
return true;
}
2019-02-22 13:33:56 +00:00
void ExecuteScalarSubqueriesMatcher::visit(ASTPtr & ast, Data & data)
{
2019-03-11 13:22:51 +00:00
if (const auto * t = ast->as<ASTSubquery>())
visit(*t, ast, data);
2019-03-11 13:22:51 +00:00
if (const auto * t = ast->as<ASTFunction>())
2019-02-22 13:33:56 +00:00
visit(*t, ast, data);
}
2019-10-19 20:36:35 +00:00
/// Converting to literal values might take a fair amount of overhead when the value is large, (e.g.
/// Array, BitMap, etc.), This conversion is required for constant folding, index lookup, branch
/// elimination. However, these optimizations should never be related to large values, thus we
/// blacklist them here.
static bool worthConvertingToLiteral(const Block & scalar)
{
auto scalar_type_name = scalar.safeGetByPosition(0).type->getFamilyName();
std::set<String> useless_literal_types = {"Array", "Tuple", "AggregateFunction", "Function", "Set", "LowCardinality"};
return !useless_literal_types.count(scalar_type_name);
}
void ExecuteScalarSubqueriesMatcher::visit(const ASTSubquery & subquery, ASTPtr & ast, Data & data)
{
2019-10-19 20:36:35 +00:00
auto hash = subquery.getTreeHash();
auto scalar_query_hash_str = toString(hash.first) + "_" + toString(hash.second);
Block scalar;
if (data.context.hasQueryContext() && data.context.getQueryContext().hasScalar(scalar_query_hash_str))
scalar = data.context.getQueryContext().getScalar(scalar_query_hash_str);
else if (data.scalars.count(scalar_query_hash_str))
scalar = data.scalars[scalar_query_hash_str];
else
{
2019-10-19 20:36:35 +00:00
Context subquery_context = data.context;
Settings subquery_settings = data.context.getSettings();
subquery_settings.max_result_rows = 1;
2020-03-08 23:48:08 +00:00
subquery_settings.extremes = false;
2019-10-19 20:36:35 +00:00
subquery_context.setSettings(subquery_settings);
ASTPtr subquery_select = subquery.children.at(0);
BlockIO res = InterpreterSelectWithUnionQuery(
subquery_select, subquery_context, SelectQueryOptions(QueryProcessingStage::Complete, data.subquery_depth + 1)).execute();
Block block;
try
{
2019-10-19 20:36:35 +00:00
block = res.in->read();
if (!block)
{
/// Interpret subquery with empty result as Null literal
auto ast_new = std::make_unique<ASTLiteral>(Null());
ast_new->setAlias(ast->tryGetAlias());
ast = std::move(ast_new);
return;
}
if (block.rows() != 1 || res.in->read())
throw Exception("Scalar subquery returned more than one row", ErrorCodes::INCORRECT_RESULT_OF_SCALAR_SUBQUERY);
}
catch (const Exception & e)
{
if (e.code() == ErrorCodes::TOO_MANY_ROWS)
throw Exception("Scalar subquery returned more than one row", ErrorCodes::INCORRECT_RESULT_OF_SCALAR_SUBQUERY);
else
throw;
}
2019-10-19 20:36:35 +00:00
block = materializeBlock(block);
size_t columns = block.columns();
if (columns == 1)
scalar = block;
else
2019-10-19 20:36:35 +00:00
{
ColumnWithTypeAndName ctn;
ctn.type = std::make_shared<DataTypeTuple>(block.getDataTypes());
ctn.column = ColumnTuple::create(block.getColumns());
scalar.insert(ctn);
}
}
2019-10-19 20:36:35 +00:00
const Settings & settings = data.context.getSettingsRef();
// Always convert to literals when there is no query context.
if (!settings.enable_scalar_subquery_optimization || worthConvertingToLiteral(scalar) || !data.context.hasQueryContext())
{
2019-10-19 20:36:35 +00:00
auto lit = std::make_unique<ASTLiteral>((*scalar.safeGetByPosition(0).column)[0]);
lit->alias = subquery.alias;
lit->prefer_alias_to_column_name = subquery.prefer_alias_to_column_name;
2019-10-19 20:36:35 +00:00
ast = addTypeConversionToAST(std::move(lit), scalar.safeGetByPosition(0).type->getName());
}
else
{
2019-10-19 20:36:35 +00:00
auto func = makeASTFunction("__getScalar", std::make_shared<ASTLiteral>(scalar_query_hash_str));
func->alias = subquery.alias;
func->prefer_alias_to_column_name = subquery.prefer_alias_to_column_name;
ast = std::move(func);
}
2019-10-19 20:36:35 +00:00
data.scalars[scalar_query_hash_str] = std::move(scalar);
}
2019-02-22 13:33:56 +00:00
void ExecuteScalarSubqueriesMatcher::visit(const ASTFunction & func, ASTPtr & ast, Data & data)
{
/// Don't descend into subqueries in arguments of IN operator.
/// But if an argument is not subquery, than deeper may be scalar subqueries and we need to descend in them.
std::vector<ASTPtr *> out;
if (functionIsInOrGlobalInOperator(func.name))
{
for (auto & child : ast->children)
{
if (child != func.arguments)
out.push_back(&child);
else
for (size_t i = 0, size = func.arguments->children.size(); i < size; ++i)
2019-03-11 13:22:51 +00:00
if (i != 1 || !func.arguments->children[i]->as<ASTSubquery>())
out.push_back(&func.arguments->children[i]);
}
}
else
for (auto & child : ast->children)
out.push_back(&child);
2019-02-22 13:33:56 +00:00
for (ASTPtr * add_node : out)
2019-02-22 15:45:47 +00:00
Visitor(data).visit(*add_node);
}
}