ClickHouse/dbms/Interpreters/ExpressionAnalyzer.cpp

1166 lines
43 KiB
C++
Raw Normal View History

#include <Poco/Util/Application.h>
2016-11-20 12:43:20 +00:00
#include <Poco/String.h>
#include <Core/Block.h>
#include <Parsers/ASTFunction.h>
#include <Parsers/ASTIdentifier.h>
#include <Parsers/ASTLiteral.h>
#include <Parsers/ASTQualifiedAsterisk.h>
#include <Parsers/ASTExpressionList.h>
#include <Parsers/ASTSelectQuery.h>
2018-02-26 09:05:06 +00:00
#include <Parsers/ASTSelectWithUnionQuery.h>
#include <Parsers/ASTSubquery.h>
#include <Parsers/ASTOrderByElement.h>
#include <Parsers/formatAST.h>
#include <Parsers/DumpASTNode.h>
2017-12-25 00:38:45 +00:00
#include <DataTypes/DataTypeNullable.h>
#include <DataTypes/NestedUtils.h>
#include <DataTypes/DataTypesNumber.h>
#include <DataTypes/DataTypeLowCardinality.h>
2018-08-28 14:40:07 +00:00
#include <Columns/IColumn.h>
#include <Interpreters/ExpressionAnalyzer.h>
#include <Interpreters/ExpressionActions.h>
#include <Interpreters/LogicalExpressionsOptimizer.h>
#include <Interpreters/ExternalDictionariesLoader.h>
#include <Interpreters/Set.h>
2019-09-03 16:56:32 +00:00
#include <Interpreters/AnalyzedJoin.h>
2020-02-11 18:27:52 +00:00
#include <Interpreters/JoinSwitcher.h>
2020-02-17 17:21:03 +00:00
#include <Interpreters/Join.h>
#include <Interpreters/MergeJoin.h>
#include <AggregateFunctions/AggregateFunctionFactory.h>
#include <AggregateFunctions/parseAggregateFunctionParameters.h>
2015-04-16 06:12:35 +00:00
#include <Storages/StorageDistributed.h>
#include <Storages/StorageJoin.h>
#include <DataStreams/copyData.h>
#include <DataStreams/IBlockInputStream.h>
#include <Dictionaries/IDictionary.h>
#include <Common/typeid_cast.h>
#include <Common/StringUtils/StringUtils.h>
2017-06-06 17:18:32 +00:00
#include <ext/range.h>
#include <DataTypes/DataTypeFactory.h>
#include <Functions/FunctionsMiscellaneous.h>
#include <Parsers/ExpressionListParsers.h>
#include <Parsers/parseQuery.h>
2018-07-27 16:21:43 +00:00
#include <Parsers/queryToString.h>
#include <Interpreters/interpretSubquery.h>
#include <Interpreters/DatabaseAndTableWithAlias.h>
#include <Interpreters/misc.h>
#include <Interpreters/ActionsVisitor.h>
#include <Interpreters/GlobalSubqueriesVisitor.h>
2019-08-12 19:27:09 +00:00
#include <Interpreters/GetAggregatesVisitor.h>
namespace DB
{
2018-10-10 16:23:27 +00:00
using LogAST = DebugASTLog<false>; /// set to true to enable logs
namespace ErrorCodes
{
2020-02-25 18:02:41 +00:00
extern const int UNKNOWN_TYPE_OF_AST_NODE;
extern const int UNKNOWN_IDENTIFIER;
extern const int ILLEGAL_PREWHERE;
extern const int LOGICAL_ERROR;
}
namespace
{
/// Check if there is an ignore function. It's used for disabling constant folding in query
/// predicates because some performance tests use ignore function as a non-optimize guard.
bool allowEarlyConstantFolding(const ExpressionActions & actions, const Settings & settings)
{
if (!settings.enable_early_constant_folding)
return false;
for (auto & action : actions.getActions())
{
if (action.type == action.APPLY_FUNCTION && action.function_base)
{
auto name = action.function_base->getName();
if (name == "ignore")
return false;
}
}
return true;
}
}
bool sanitizeBlock(Block & block)
{
for (auto & col : block)
{
if (!col.column)
{
if (isNotCreatable(col.type->getTypeId()))
return false;
col.column = col.type->createColumn();
}
else if (isColumnConst(*col.column) && !col.column->empty())
col.column = col.column->cloneEmpty();
}
return true;
}
Squashed commit of the following: commit e712f469a55ff34ad34b482b15cc4153b7ad7233 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:59:13 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 2a002823084e3a79bffcc17d479620a68eb0644b Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:58:30 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 9e06f407c8ee781ed8ddf98bdfcc31846bf2a0fe Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:55:14 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 9581620f1e839f456fa7894aa1f996d5162ac6cd Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:54:22 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 2a8564c68cb6cc3649fafaf401256d43c9a2e777 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:47:34 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit cf60632d78ec656be3304ef4565e859bb6ce80ba Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:40:09 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit ee3d1dc6e0c4ca60e3ac1e0c30d4b3ed1e66eca0 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:22:49 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 65592ef7116a90104fcd524b53ef8b7cf22640f2 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:18:17 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 37972c257320d3b7e7b294e0fdeffff218647bfd Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:17:06 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit dd909d149974ce5bed2456de1261aa5a368fd3ff Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:16:28 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 3cf43266ca7e30adf01212b1a739ba5fe43639fd Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:15:42 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 6731a3df96d1609286e2536b6432916af7743f0f Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:13:35 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 1b5727e0d56415b7add4cb76110105358663602c Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:11:18 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit bbcf726a55685b8e72f5b40ba0bf1904bd1c0407 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:09:04 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit c03b477d5e2e65014e8906ecfa2efb67ee295af1 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:06:30 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 2986e2fb0466bc18d73693dcdded28fccc0dc66b Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:05:44 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit 5d6cdef13d2e02bd5c4954983334e9162ab2635b Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:04:53 2017 +0300 Less dependencies [#CLICKHOUSE-2] commit f2b819b25ce8b2ccdcb201eefb03e1e6f5aab590 Author: Alexey Milovidov <milovidov@yandex-team.ru> Date: Sat Jan 14 11:01:47 2017 +0300 Less dependencies [#CLICKHOUSE-2]
2017-01-14 09:00:19 +00:00
ExpressionAnalyzer::ExpressionAnalyzer(
2018-08-27 17:58:43 +00:00
const ASTPtr & query_,
const SyntaxAnalyzerResultPtr & syntax_analyzer_result_,
const Context & context_,
size_t subquery_depth_,
2019-08-13 13:48:09 +00:00
bool do_global)
: query(query_), context(context_), settings(context.getSettings())
2019-08-13 13:48:09 +00:00
, subquery_depth(subquery_depth_)
2018-11-09 17:23:48 +00:00
, syntax(syntax_analyzer_result_)
{
2017-04-02 17:37:49 +00:00
/// external_tables, subqueries_for_sets for global subqueries.
/// Replaces global subqueries with the generated names of temporary tables that will be sent to remote servers.
2019-08-13 13:48:09 +00:00
initGlobalSubqueriesAndExternalTables(do_global);
/// has_aggregation, aggregation_keys, aggregate_descriptions, aggregated_columns.
2017-04-02 17:37:49 +00:00
/// This analysis should be performed after processing global subqueries, because otherwise,
/// if the aggregate function contains a global subquery, then `analyzeAggregation` method will save
/// in `aggregate_descriptions` the information about the parameters of this aggregate function, among which
/// global subquery. Then, when you call `initGlobalSubqueriesAndExternalTables` method, this
/// the global subquery will be replaced with a temporary table, resulting in aggregate_descriptions
/// will contain out-of-date information, which will lead to an error when the query is executed.
analyzeAggregation();
}
2018-10-17 11:06:46 +00:00
bool ExpressionAnalyzer::isRemoteStorage() const
{
return storage() && storage()->isRemote();
}
void ExpressionAnalyzer::analyzeAggregation()
{
2017-04-02 17:37:49 +00:00
/** Find aggregation keys (aggregation_keys), information about aggregate functions (aggregate_descriptions),
* as well as a set of columns obtained after the aggregation, if any,
* or after all the actions that are usually performed before aggregation (aggregated_columns).
*
2017-04-02 17:37:49 +00:00
* Everything below (compiling temporary ExpressionActions) - only for the purpose of query analysis (type output).
*/
2019-03-11 13:22:51 +00:00
auto * select_query = query->as<ASTSelectQuery>();
ExpressionActionsPtr temp_actions = std::make_shared<ExpressionActions>(sourceColumns(), context);
if (select_query)
{
bool is_array_join_left;
ASTPtr array_join_expression_list = select_query->arrayJoinExpressionList(is_array_join_left);
if (array_join_expression_list)
{
2020-01-27 21:57:44 +00:00
getRootActionsNoMakeSet(array_join_expression_list, true, temp_actions, false);
addMultipleArrayJoinAction(temp_actions, is_array_join_left);
array_join_columns.clear();
for (auto & column : temp_actions->getSampleBlock().getNamesAndTypesList())
if (syntax->array_join_result_to_source.count(column.name))
array_join_columns.emplace_back(column);
}
const ASTTablesInSelectQueryElement * join = select_query->join();
if (join)
{
2020-01-27 21:57:44 +00:00
getRootActionsNoMakeSet(analyzedJoin().leftKeysList(), true, temp_actions, false);
addJoinAction(temp_actions);
}
}
has_aggregation = makeAggregateDescriptions(temp_actions);
if (select_query && (select_query->groupBy() || select_query->having()))
has_aggregation = true;
if (has_aggregation)
{
2019-08-13 13:48:09 +00:00
getSelectQuery(); /// assertSelect()
/// Find out aggregation keys.
if (select_query->groupBy())
{
NameSet unique_keys;
ASTs & group_asts = select_query->groupBy()->children;
for (ssize_t i = 0; i < ssize_t(group_asts.size()); ++i)
{
ssize_t size = group_asts.size();
2020-01-27 21:57:44 +00:00
getRootActionsNoMakeSet(group_asts[i], true, temp_actions, false);
const auto & column_name = group_asts[i]->getColumnName();
const auto & block = temp_actions->getSampleBlock();
if (!block.has(column_name))
throw Exception("Unknown identifier (in GROUP BY): " + column_name, ErrorCodes::UNKNOWN_IDENTIFIER);
const auto & col = block.getByName(column_name);
/// Constant expressions have non-null column pointer at this stage.
if (col.column && isColumnConst(*col.column))
{
/// But don't remove last key column if no aggregate functions, otherwise aggregation will not work.
if (!aggregate_descriptions.empty() || size > 1)
{
if (i + 1 < static_cast<ssize_t>(size))
group_asts[i] = std::move(group_asts.back());
group_asts.pop_back();
--i;
continue;
}
}
NameAndTypePair key{column_name, col.type};
/// Aggregation keys are uniqued.
if (!unique_keys.count(key.name))
{
unique_keys.insert(key.name);
aggregation_keys.push_back(key);
/// Key is no longer needed, therefore we can save a little by moving it.
aggregated_columns.push_back(std::move(key));
}
}
if (group_asts.empty())
{
select_query->setExpression(ASTSelectQuery::Expression::GROUP_BY, {});
2020-03-08 23:48:08 +00:00
has_aggregation = select_query->having() || !aggregate_descriptions.empty();
}
}
2020-03-08 23:48:08 +00:00
for (const auto & desc : aggregate_descriptions)
aggregated_columns.emplace_back(desc.column_name, desc.function->getReturnType());
}
else
{
aggregated_columns = temp_actions->getSampleBlock().getNamesAndTypesList();
}
}
2019-08-13 13:48:09 +00:00
void ExpressionAnalyzer::initGlobalSubqueriesAndExternalTables(bool do_global)
{
2018-10-17 10:59:05 +00:00
if (do_global)
{
GlobalSubqueriesVisitor::Data subqueries_data(context, subquery_depth, isRemoteStorage(),
2018-10-17 10:59:05 +00:00
external_tables, subqueries_for_sets, has_global_subqueries);
GlobalSubqueriesVisitor(subqueries_data).visit(query);
2018-10-17 10:59:05 +00:00
}
}
NamesAndTypesList ExpressionAnalyzer::sourceWithJoinedColumns() const
{
auto result_columns = sourceColumns();
result_columns.insert(result_columns.end(), array_join_columns.begin(), array_join_columns.end());
result_columns.insert(result_columns.end(),
analyzedJoin().columnsAddedByJoin().begin(), analyzedJoin().columnsAddedByJoin().end());
return result_columns;
}
void SelectQueryExpressionAnalyzer::tryMakeSetForIndexFromSubquery(const ASTPtr & subquery_or_table_name)
{
auto set_key = PreparedSetKey::forSubquery(*subquery_or_table_name);
if (prepared_sets.count(set_key))
return; /// Already prepared.
if (auto set_ptr_from_storage_set = isPlainStorageSetInSubquery(subquery_or_table_name))
{
prepared_sets.insert({set_key, set_ptr_from_storage_set});
return;
}
auto interpreter_subquery = interpretSubquery(subquery_or_table_name, context, {}, query_options);
BlockIO res = interpreter_subquery->execute();
SetPtr set = std::make_shared<Set>(settings.size_limits_for_set, true);
2018-04-19 21:34:04 +00:00
set->setHeader(res.in->getHeader());
res.in->readPrefix();
while (Block block = res.in->read())
{
/// If the limits have been exceeded, give up and let the default subquery processing actions take place.
2018-07-02 18:57:14 +00:00
if (!set->insertFromBlock(block))
return;
}
2019-11-01 10:58:29 +00:00
set->finishInsert();
res.in->readSuffix();
prepared_sets[set_key] = std::move(set);
}
SetPtr SelectQueryExpressionAnalyzer::isPlainStorageSetInSubquery(const ASTPtr & subquery_or_table_name)
2019-11-13 12:43:55 +00:00
{
const auto * table = subquery_or_table_name->as<ASTIdentifier>();
if (!table)
return nullptr;
2020-03-13 10:30:55 +00:00
auto table_id = context.resolveStorageID(subquery_or_table_name);
2020-03-06 20:38:19 +00:00
const auto storage = DatabaseCatalog::instance().getTable(table_id);
if (storage->getName() != "Set")
return nullptr;
const auto storage_set = std::dynamic_pointer_cast<StorageSet>(storage);
return storage_set->getSet();
2019-11-13 12:43:55 +00:00
}
2020-01-11 09:50:41 +00:00
/// Performance optimisation for IN() if storage supports it.
void SelectQueryExpressionAnalyzer::makeSetsForIndex(const ASTPtr & node)
{
2019-08-12 19:27:09 +00:00
if (!node || !storage() || !storage()->supportsIndexForIn())
return;
for (auto & child : node->children)
{
/// Don't descend into subqueries.
2019-03-11 13:22:51 +00:00
if (child->as<ASTSubquery>())
continue;
/// Don't descend into lambda functions
2019-03-11 13:22:51 +00:00
const auto * func = child->as<ASTFunction>();
if (func && func->name == "lambda")
continue;
2019-08-12 19:27:09 +00:00
makeSetsForIndex(child);
}
2019-03-11 13:22:51 +00:00
const auto * func = node->as<ASTFunction>();
if (func && functionIsInOperator(func->name))
{
const IAST & args = *func->arguments;
2019-08-12 19:27:09 +00:00
const ASTPtr & left_in_operand = args.children.at(0);
2019-08-12 19:27:09 +00:00
if (storage()->mayBenefitFromIndexForIn(left_in_operand, context))
{
2018-03-16 06:39:32 +00:00
const ASTPtr & arg = args.children.at(1);
2019-03-11 13:22:51 +00:00
if (arg->as<ASTSubquery>() || arg->as<ASTIdentifier>())
{
if (settings.use_index_for_in_with_subqueries)
tryMakeSetForIndexFromSubquery(arg);
}
else
{
ExpressionActionsPtr temp_actions = std::make_shared<ExpressionActions>(sourceWithJoinedColumns(), context);
2019-08-12 19:27:09 +00:00
getRootActions(left_in_operand, true, temp_actions);
Block sample_block_with_calculated_columns = temp_actions->getSampleBlock();
2019-08-12 19:27:09 +00:00
if (sample_block_with_calculated_columns.has(left_in_operand->getColumnName()))
makeExplicitSet(func, sample_block_with_calculated_columns, true, context,
settings.size_limits_for_set, prepared_sets);
}
}
}
}
void ExpressionAnalyzer::getRootActions(const ASTPtr & ast, bool no_subqueries, ExpressionActionsPtr & actions, bool only_consts)
{
LogAST log;
ActionsVisitor::Data visitor_data(context, settings.size_limits_for_set, subquery_depth,
sourceColumns(), actions, prepared_sets, subqueries_for_sets,
2020-01-27 21:57:44 +00:00
no_subqueries, false, only_consts, !isRemoteStorage());
ActionsVisitor(visitor_data, log.stream()).visit(ast);
visitor_data.updateActions(actions);
}
void ExpressionAnalyzer::getRootActionsNoMakeSet(const ASTPtr & ast, bool no_subqueries, ExpressionActionsPtr & actions, bool only_consts)
{
LogAST log;
ActionsVisitor::Data visitor_data(context, settings.size_limits_for_set, subquery_depth,
sourceColumns(), actions, prepared_sets, subqueries_for_sets,
no_subqueries, true, only_consts, !isRemoteStorage());
ActionsVisitor(visitor_data, log.stream()).visit(ast);
visitor_data.updateActions(actions);
}
bool ExpressionAnalyzer::makeAggregateDescriptions(ExpressionActionsPtr & actions)
{
for (const ASTFunction * node : aggregates())
{
AggregateDescription aggregate;
aggregate.column_name = node->getColumnName();
const ASTs & arguments = node->arguments->children;
aggregate.argument_names.resize(arguments.size());
DataTypes types(arguments.size());
for (size_t i = 0; i < arguments.size(); ++i)
{
getRootActions(arguments[i], true, actions);
const std::string & name = arguments[i]->getColumnName();
types[i] = actions->getSampleBlock().getByName(name).type;
aggregate.argument_names[i] = name;
}
aggregate.parameters = (node->parameters) ? getAggregateFunctionParametersArray(node->parameters) : Array();
aggregate.function = AggregateFunctionFactory::instance().get(node->name, types, aggregate.parameters);
aggregate_descriptions.push_back(aggregate);
}
2016-03-05 02:30:20 +00:00
return !aggregates().empty();
2016-03-05 02:30:20 +00:00
}
2019-08-13 13:48:09 +00:00
const ASTSelectQuery * ExpressionAnalyzer::getSelectQuery() const
{
2019-03-11 13:22:51 +00:00
const auto * select_query = query->as<ASTSelectQuery>();
if (!select_query)
throw Exception("Not a select query", ErrorCodes::LOGICAL_ERROR);
2019-08-13 13:48:09 +00:00
return select_query;
}
2013-06-14 16:38:54 +00:00
const ASTSelectQuery * SelectQueryExpressionAnalyzer::getAggregatingQuery() const
{
if (!has_aggregation)
throw Exception("No aggregation", ErrorCodes::LOGICAL_ERROR);
2019-08-13 13:48:09 +00:00
return getSelectQuery();
}
2013-06-14 16:38:54 +00:00
void ExpressionAnalyzer::initChain(ExpressionActionsChain & chain, const NamesAndTypesList & columns) const
{
if (chain.steps.empty())
{
chain.steps.emplace_back(std::make_shared<ExpressionActions>(columns, context));
}
}
2016-07-23 02:25:09 +00:00
/// "Big" ARRAY JOIN.
void ExpressionAnalyzer::addMultipleArrayJoinAction(ExpressionActionsPtr & actions, bool array_join_is_left) const
{
NameSet result_columns;
for (const auto & result_source : syntax->array_join_result_to_source)
{
/// Assign new names to columns, if needed.
if (result_source.first != result_source.second)
actions->add(ExpressionAction::copyColumn(result_source.second, result_source.first));
2017-04-02 17:37:49 +00:00
/// Make ARRAY JOIN (replace arrays with their insides) for the columns in these new names.
result_columns.insert(result_source.first);
}
actions->add(ExpressionAction::arrayJoin(result_columns, array_join_is_left, context));
}
bool SelectQueryExpressionAnalyzer::appendArrayJoin(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
bool is_array_join_left;
ASTPtr array_join_expression_list = select_query->arrayJoinExpressionList(is_array_join_left);
if (!array_join_expression_list)
return false;
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(array_join_expression_list, only_types, step.actions);
addMultipleArrayJoinAction(step.actions, is_array_join_left);
return true;
}
void ExpressionAnalyzer::addJoinAction(ExpressionActionsPtr & actions, JoinPtr join) const
{
actions->add(ExpressionAction::ordinaryJoin(syntax->analyzed_join, join));
}
bool SelectQueryExpressionAnalyzer::appendJoin(ExpressionActionsChain & chain, bool only_types)
{
const ASTTablesInSelectQueryElement * ast_join = getSelectQuery()->join();
if (!ast_join)
return false;
JoinPtr table_join = makeTableJoin(*ast_join);
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(analyzedJoin().leftKeysList(), only_types, step.actions);
addJoinAction(step.actions, table_join);
return true;
}
2019-12-19 15:50:28 +00:00
static JoinPtr tryGetStorageJoin(const ASTTablesInSelectQueryElement & join_element, std::shared_ptr<AnalyzedJoin> analyzed_join,
const Context & context)
{
const auto & table_to_join = join_element.table_expression->as<ASTTableExpression &>();
2017-04-02 17:37:49 +00:00
/// TODO This syntax does not support specifying a database name.
if (table_to_join.database_and_table_name)
{
2020-03-13 10:30:55 +00:00
auto table_id = context.resolveStorageID(table_to_join.database_and_table_name);
2020-03-04 20:29:52 +00:00
StoragePtr table = DatabaseCatalog::instance().tryGetTable(table_id);
if (table)
{
auto * storage_join = dynamic_cast<StorageJoin *>(table.get());
if (storage_join)
2019-12-19 15:50:28 +00:00
return storage_join->getJoin(analyzed_join);
}
}
return {};
}
2019-09-04 16:37:05 +00:00
static ExpressionActionsPtr createJoinedBlockActions(const Context & context, const AnalyzedJoin & analyzed_join)
{
ASTPtr expression_list = analyzed_join.rightKeysList();
auto syntax_result = SyntaxAnalyzer(context).analyze(expression_list, analyzed_join.columnsFromJoinedTable());
2019-09-04 16:37:05 +00:00
return ExpressionAnalyzer(expression_list, syntax_result, context).getActions(true, false);
}
static std::shared_ptr<IJoin> makeJoin(std::shared_ptr<AnalyzedJoin> analyzed_join, const Block & sample_block)
{
bool allow_merge_join = analyzed_join->allowMergeJoin();
if (analyzed_join->forceHashJoin() || (analyzed_join->preferMergeJoin() && !allow_merge_join))
return std::make_shared<Join>(analyzed_join, sample_block);
else if (analyzed_join->forceMergeJoin() || (analyzed_join->preferMergeJoin() && allow_merge_join))
return std::make_shared<MergeJoin>(analyzed_join, sample_block);
return std::make_shared<JoinSwitcher>(analyzed_join, sample_block);
}
JoinPtr SelectQueryExpressionAnalyzer::makeTableJoin(const ASTTablesInSelectQueryElement & join_element)
{
/// Two JOINs are not supported with the same subquery, but different USINGs.
auto join_hash = join_element.getTreeHash();
String join_subquery_id = toString(join_hash.first) + "_" + toString(join_hash.second);
2019-09-09 19:43:37 +00:00
SubqueryForSet & subquery_for_join = subqueries_for_sets[join_subquery_id];
/// Special case - if table name is specified on the right of JOIN, then the table has the type Join (the previously prepared mapping).
2019-09-09 19:43:37 +00:00
if (!subquery_for_join.join)
2019-12-19 15:50:28 +00:00
subquery_for_join.join = tryGetStorageJoin(join_element, syntax->analyzed_join, context);
2019-09-09 19:43:37 +00:00
if (!subquery_for_join.join)
{
/// Actions which need to be calculated on joined block.
2019-09-04 16:37:05 +00:00
ExpressionActionsPtr joined_block_actions = createJoinedBlockActions(context, analyzedJoin());
2019-09-09 19:43:37 +00:00
if (!subquery_for_join.source)
{
NamesWithAliases required_columns_with_aliases =
analyzedJoin().getRequiredColumns(joined_block_actions->getSampleBlock(), joined_block_actions->getRequiredColumns());
makeSubqueryForJoin(join_element, std::move(required_columns_with_aliases), subquery_for_join);
}
/// TODO You do not need to set this up when JOIN is only needed on remote servers.
2019-09-09 19:43:37 +00:00
subquery_for_join.setJoinActions(joined_block_actions); /// changes subquery_for_join.sample_block inside
subquery_for_join.join = makeJoin(syntax->analyzed_join, subquery_for_join.sample_block);
}
return subquery_for_join.join;
}
void SelectQueryExpressionAnalyzer::makeSubqueryForJoin(const ASTTablesInSelectQueryElement & join_element,
2019-09-09 19:43:37 +00:00
NamesWithAliases && required_columns_with_aliases,
SubqueryForSet & subquery_for_set) const
{
/** For GLOBAL JOINs (in the case, for example, of the push method for executing GLOBAL subqueries), the following occurs
* - in the addExternalStorage function, the JOIN (SELECT ...) subquery is replaced with JOIN _data1,
* in the subquery_for_set object this subquery is exposed as source and the temporary table _data1 as the `table`.
* - this function shows the expression JOIN _data1.
*/
Names original_columns;
for (auto & pr : required_columns_with_aliases)
original_columns.push_back(pr.first);
auto interpreter = interpretSubquery(join_element.table_expression, context, original_columns, query_options);
subquery_for_set.makeSource(interpreter, std::move(required_columns_with_aliases));
}
bool SelectQueryExpressionAnalyzer::appendPrewhere(
ExpressionActionsChain & chain, bool only_types, const Names & additional_required_columns)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
if (!select_query->prewhere())
return false;
initChain(chain, sourceColumns());
2018-04-12 09:45:24 +00:00
auto & step = chain.getLastStep();
getRootActions(select_query->prewhere(), only_types, step.actions);
String prewhere_column_name = select_query->prewhere()->getColumnName();
2018-04-12 09:45:24 +00:00
step.required_output.push_back(prewhere_column_name);
2018-06-29 11:42:44 +00:00
step.can_remove_required_output.push_back(true);
2018-04-12 09:45:24 +00:00
{
/// Remove unused source_columns from prewhere actions.
auto tmp_actions = std::make_shared<ExpressionActions>(sourceColumns(), context);
getRootActions(select_query->prewhere(), only_types, tmp_actions);
2018-04-12 09:45:24 +00:00
tmp_actions->finalize({prewhere_column_name});
auto required_columns = tmp_actions->getRequiredColumns();
NameSet required_source_columns(required_columns.begin(), required_columns.end());
2018-10-04 08:58:19 +00:00
/// Add required columns to required output in order not to remove them after prewhere execution.
/// TODO: add sampling and final execution to common chain.
for (const auto & column : additional_required_columns)
{
if (required_source_columns.count(column))
{
step.required_output.push_back(column);
step.can_remove_required_output.push_back(true);
}
}
2018-04-12 09:45:24 +00:00
auto names = step.actions->getSampleBlock().getNames();
NameSet name_set(names.begin(), names.end());
for (const auto & column : sourceColumns())
2018-04-12 09:45:24 +00:00
if (required_source_columns.count(column.name) == 0)
name_set.erase(column.name);
Names required_output(name_set.begin(), name_set.end());
step.actions->finalize(required_output);
}
{
/// Add empty action with input = {prewhere actions output} + {unused source columns}
/// Reasons:
/// 1. Remove remove source columns which are used only in prewhere actions during prewhere actions execution.
/// Example: select A prewhere B > 0. B can be removed at prewhere step.
/// 2. Store side columns which were calculated during prewhere actions execution if they are used.
/// Example: select F(A) prewhere F(A) > 0. F(A) can be saved from prewhere step.
2018-06-25 13:08:35 +00:00
/// 3. Check if we can remove filter column at prewhere step. If we can, action will store single REMOVE_COLUMN.
2018-04-12 09:45:24 +00:00
ColumnsWithTypeAndName columns = step.actions->getSampleBlock().getColumnsWithTypeAndName();
auto required_columns = step.actions->getRequiredColumns();
NameSet prewhere_input_names(required_columns.begin(), required_columns.end());
NameSet unused_source_columns;
for (const auto & column : sourceColumns())
2018-04-12 09:45:24 +00:00
{
if (prewhere_input_names.count(column.name) == 0)
{
columns.emplace_back(column.type, column.name);
unused_source_columns.emplace(column.name);
}
}
chain.steps.emplace_back(std::make_shared<ExpressionActions>(std::move(columns), context));
2018-04-12 09:45:24 +00:00
chain.steps.back().additional_input = std::move(unused_source_columns);
}
return true;
}
void SelectQueryExpressionAnalyzer::appendPreliminaryFilter(ExpressionActionsChain & chain, ExpressionActionsPtr actions, String column_name)
{
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
2019-10-05 19:34:57 +00:00
// FIXME: assert(filter_info);
step.actions = std::move(actions);
step.required_output.push_back(std::move(column_name));
step.can_remove_required_output = {true};
chain.addStep();
}
bool SelectQueryExpressionAnalyzer::appendWhere(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
if (!select_query->where())
return false;
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
step.required_output.push_back(select_query->where()->getColumnName());
2018-06-29 11:42:44 +00:00
step.can_remove_required_output = {true};
2018-04-23 19:05:46 +00:00
getRootActions(select_query->where(), only_types, step.actions);
return true;
}
bool SelectQueryExpressionAnalyzer::appendGroupBy(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getAggregatingQuery();
if (!select_query->groupBy())
return false;
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
ASTs asts = select_query->groupBy()->children;
2020-03-08 23:48:08 +00:00
for (const auto & ast : asts)
{
2020-03-08 23:48:08 +00:00
step.required_output.emplace_back(ast->getColumnName());
getRootActions(ast, only_types, step.actions);
}
return true;
}
void SelectQueryExpressionAnalyzer::appendAggregateFunctionsArguments(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getAggregatingQuery();
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
2020-03-08 23:48:08 +00:00
for (const auto & desc : aggregate_descriptions)
for (const auto & name : desc.argument_names)
step.required_output.emplace_back(name);
/// Collect aggregates removing duplicates by node.getColumnName()
/// It's not clear why we recollect aggregates (for query parts) while we're able to use previously collected ones (for entire query)
/// @note The original recollection logic didn't remove duplicates.
GetAggregatesVisitor::Data data;
GetAggregatesVisitor(data).visit(select_query->select());
if (select_query->having())
GetAggregatesVisitor(data).visit(select_query->having());
if (select_query->orderBy())
GetAggregatesVisitor(data).visit(select_query->orderBy());
/// TODO: data.aggregates -> aggregates()
for (const ASTFunction * node : data.aggregates)
for (auto & argument : node->arguments->children)
getRootActions(argument, only_types, step.actions);
}
bool SelectQueryExpressionAnalyzer::appendHaving(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getAggregatingQuery();
if (!select_query->having())
return false;
initChain(chain, aggregated_columns);
ExpressionActionsChain::Step & step = chain.steps.back();
step.required_output.push_back(select_query->having()->getColumnName());
getRootActions(select_query->having(), only_types, step.actions);
return true;
}
void SelectQueryExpressionAnalyzer::appendSelect(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
initChain(chain, aggregated_columns);
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(select_query->select(), only_types, step.actions);
for (const auto & child : select_query->select()->children)
2018-03-01 01:25:06 +00:00
step.required_output.push_back(child->getColumnName());
}
bool SelectQueryExpressionAnalyzer::appendOrderBy(ExpressionActionsChain & chain, bool only_types, bool optimize_read_in_order,
ManyExpressionActions & order_by_elements_actions)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
if (!select_query->orderBy())
return false;
initChain(chain, aggregated_columns);
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(select_query->orderBy(), only_types, step.actions);
for (auto & child : select_query->orderBy()->children)
{
const auto * ast = child->as<ASTOrderByElement>();
2020-03-08 23:48:08 +00:00
if (!ast || ast->children.empty())
throw Exception("Bad order expression AST", ErrorCodes::UNKNOWN_TYPE_OF_AST_NODE);
ASTPtr order_expression = ast->children.at(0);
step.required_output.push_back(order_expression->getColumnName());
}
if (optimize_read_in_order)
{
auto all_columns = sourceWithJoinedColumns();
for (auto & child : select_query->orderBy()->children)
2019-12-10 23:18:24 +00:00
{
order_by_elements_actions.emplace_back(std::make_shared<ExpressionActions>(all_columns, context));
2019-12-10 23:18:24 +00:00
getRootActions(child, only_types, order_by_elements_actions.back());
}
}
return true;
}
bool SelectQueryExpressionAnalyzer::appendLimitBy(ExpressionActionsChain & chain, bool only_types)
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
if (!select_query->limitBy())
return false;
initChain(chain, aggregated_columns);
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(select_query->limitBy(), only_types, step.actions);
NameSet aggregated_names;
for (const auto & column : aggregated_columns)
{
2019-05-25 10:51:48 +00:00
step.required_output.push_back(column.name);
aggregated_names.insert(column.name);
}
for (const auto & child : select_query->limitBy()->children)
2019-05-25 10:51:48 +00:00
{
auto child_name = child->getColumnName();
if (!aggregated_names.count(child_name))
step.required_output.push_back(std::move(child_name));
}
return true;
}
void SelectQueryExpressionAnalyzer::appendProjectResult(ExpressionActionsChain & chain) const
{
2019-08-13 13:48:09 +00:00
const auto * select_query = getSelectQuery();
initChain(chain, aggregated_columns);
ExpressionActionsChain::Step & step = chain.steps.back();
NamesWithAliases result_columns;
ASTs asts = select_query->select()->children;
2020-03-08 23:48:08 +00:00
for (const auto & ast : asts)
{
2020-03-08 23:48:08 +00:00
String result_name = ast->getAliasOrColumnName();
if (required_result_columns.empty() || required_result_columns.count(result_name))
2018-02-26 09:05:06 +00:00
{
2020-03-08 23:48:08 +00:00
result_columns.emplace_back(ast->getColumnName(), result_name);
2018-02-26 09:05:06 +00:00
step.required_output.push_back(result_columns.back().second);
}
}
step.actions->add(ExpressionAction::project(result_columns));
}
void ExpressionAnalyzer::appendExpression(ExpressionActionsChain & chain, const ASTPtr & expr, bool only_types)
{
initChain(chain, sourceColumns());
ExpressionActionsChain::Step & step = chain.steps.back();
getRootActions(expr, only_types, step.actions);
step.required_output.push_back(expr->getColumnName());
}
ExpressionActionsPtr ExpressionAnalyzer::getActions(bool add_aliases, bool project_result)
{
ExpressionActionsPtr actions = std::make_shared<ExpressionActions>(sourceColumns(), context);
NamesWithAliases result_columns;
Names result_names;
ASTs asts;
2019-03-11 13:22:51 +00:00
if (const auto * node = query->as<ASTExpressionList>())
asts = node->children;
else
2018-08-27 17:58:43 +00:00
asts = ASTs(1, query);
2020-03-08 23:48:08 +00:00
for (const auto & ast : asts)
{
2020-03-08 23:48:08 +00:00
std::string name = ast->getColumnName();
std::string alias;
if (add_aliases)
2020-03-08 23:48:08 +00:00
alias = ast->getAliasOrColumnName();
else
alias = name;
result_columns.emplace_back(name, alias);
result_names.push_back(alias);
2020-03-08 23:48:08 +00:00
getRootActions(ast, false, actions);
}
if (add_aliases)
{
if (project_result)
actions->add(ExpressionAction::project(result_columns));
else
actions->add(ExpressionAction::addAliases(result_columns));
}
2018-09-05 13:04:28 +00:00
if (!(add_aliases && project_result))
{
2017-04-02 17:37:49 +00:00
/// We will not delete the original columns.
for (const auto & column_name_type : sourceColumns())
result_names.push_back(column_name_type.name);
}
actions->finalize(result_names);
return actions;
}
ExpressionActionsPtr ExpressionAnalyzer::getConstActions()
{
ExpressionActionsPtr actions = std::make_shared<ExpressionActions>(NamesAndTypesList(), context);
getRootActions(query, true, actions, true);
return actions;
}
ExpressionActionsPtr SelectQueryExpressionAnalyzer::simpleSelectActions()
{
ExpressionActionsChain new_chain(context);
appendSelect(new_chain, false);
return new_chain.getLastActions();
}
ExpressionAnalysisResult::ExpressionAnalysisResult(
SelectQueryExpressionAnalyzer & query_analyzer,
bool first_stage_,
bool second_stage_,
bool only_types,
const FilterInfoPtr & filter_info_,
const Block & source_header)
: first_stage(first_stage_)
, second_stage(second_stage_)
, need_aggregate(query_analyzer.hasAggregation())
{
/// first_stage: Do I need to perform the first part of the pipeline - running on remote servers during distributed processing.
/// second_stage: Do I need to execute the second part of the pipeline - running on the initiating server during distributed processing.
/** First we compose a chain of actions and remember the necessary steps from it.
* Regardless of from_stage and to_stage, we will compose a complete sequence of actions to perform optimization and
* throw out unnecessary columns based on the entire query. In unnecessary parts of the query, we will not execute subqueries.
*/
const ASTSelectQuery & query = *query_analyzer.getSelectQuery();
const Context & context = query_analyzer.context;
const Settings & settings = context.getSettingsRef();
const ConstStoragePtr & storage = query_analyzer.storage();
bool finalized = false;
size_t where_step_num = 0;
auto finalize_chain = [&](ExpressionActionsChain & chain)
{
if (!finalized)
{
chain.finalize();
finalize(chain, context, where_step_num);
chain.clear();
}
finalized = true;
};
{
ExpressionActionsChain chain(context);
Names additional_required_columns_after_prewhere;
if (storage && (query.sampleSize() || settings.parallel_replicas_count > 1))
{
Names columns_for_sampling = storage->getColumnsRequiredForSampling();
additional_required_columns_after_prewhere.insert(additional_required_columns_after_prewhere.end(),
columns_for_sampling.begin(), columns_for_sampling.end());
}
if (storage && query.final())
{
Names columns_for_final = storage->getColumnsRequiredForFinal();
additional_required_columns_after_prewhere.insert(additional_required_columns_after_prewhere.end(),
columns_for_final.begin(), columns_for_final.end());
}
if (storage && filter_info_)
{
filter_info = filter_info_;
query_analyzer.appendPreliminaryFilter(chain, filter_info->actions, filter_info->column_name);
}
if (query_analyzer.appendPrewhere(chain, !first_stage, additional_required_columns_after_prewhere))
{
prewhere_info = std::make_shared<PrewhereInfo>(
chain.steps.front().actions, query.prewhere()->getColumnName());
if (allowEarlyConstantFolding(*prewhere_info->prewhere_actions, settings))
{
Block before_prewhere_sample = source_header;
if (sanitizeBlock(before_prewhere_sample))
{
prewhere_info->prewhere_actions->execute(before_prewhere_sample);
auto & column_elem = before_prewhere_sample.getByName(query.prewhere()->getColumnName());
/// If the filter column is a constant, record it.
if (column_elem.column)
prewhere_constant_filter_description = ConstantFilterDescription(*column_elem.column);
}
}
chain.addStep();
}
query_analyzer.appendArrayJoin(chain, only_types || !first_stage);
if (query_analyzer.appendJoin(chain, only_types || !first_stage))
{
before_join = chain.getLastActions();
if (!hasJoin())
throw Exception("No expected JOIN", ErrorCodes::LOGICAL_ERROR);
chain.addStep();
}
if (query_analyzer.appendWhere(chain, only_types || !first_stage))
{
where_step_num = chain.steps.size() - 1;
before_where = chain.getLastActions();
if (allowEarlyConstantFolding(*before_where, settings))
{
Block before_where_sample;
if (chain.steps.size() > 1)
before_where_sample = chain.steps[chain.steps.size() - 2].actions->getSampleBlock();
else
before_where_sample = source_header;
if (sanitizeBlock(before_where_sample))
{
before_where->execute(before_where_sample);
auto & column_elem = before_where_sample.getByName(query.where()->getColumnName());
/// If the filter column is a constant, record it.
if (column_elem.column)
where_constant_filter_description = ConstantFilterDescription(*column_elem.column);
}
}
chain.addStep();
}
if (need_aggregate)
{
query_analyzer.appendGroupBy(chain, only_types || !first_stage);
query_analyzer.appendAggregateFunctionsArguments(chain, only_types || !first_stage);
before_aggregation = chain.getLastActions();
finalize_chain(chain);
if (query_analyzer.appendHaving(chain, only_types || !second_stage))
{
before_having = chain.getLastActions();
chain.addStep();
}
}
bool has_stream_with_non_joned_rows = (before_join && before_join->getTableJoinAlgo()->hasStreamWithNonJoinedRows());
optimize_read_in_order =
settings.optimize_read_in_order
&& storage && query.orderBy()
&& !query_analyzer.hasAggregation()
&& !query.final()
&& !has_stream_with_non_joned_rows;
/// If there is aggregation, we execute expressions in SELECT and ORDER BY on the initiating server, otherwise on the source servers.
query_analyzer.appendSelect(chain, only_types || (need_aggregate ? !second_stage : !first_stage));
selected_columns = chain.getLastStep().required_output;
has_order_by = query_analyzer.appendOrderBy(chain, only_types || (need_aggregate ? !second_stage : !first_stage),
optimize_read_in_order, order_by_elements_actions);
before_order_and_select = chain.getLastActions();
chain.addStep();
if (query_analyzer.appendLimitBy(chain, only_types || !second_stage))
{
before_limit_by = chain.getLastActions();
chain.addStep();
}
query_analyzer.appendProjectResult(chain);
final_projection = chain.getLastActions();
finalize_chain(chain);
}
/// Before executing WHERE and HAVING, remove the extra columns from the block (mostly the aggregation keys).
removeExtraColumns();
checkActions();
}
void ExpressionAnalysisResult::finalize(const ExpressionActionsChain & chain, const Context & context_, size_t where_step_num)
{
if (hasPrewhere())
{
const ExpressionActionsChain::Step & step = chain.steps.at(0);
prewhere_info->remove_prewhere_column = step.can_remove_required_output.at(0);
Names columns_to_remove;
for (size_t i = 1; i < step.required_output.size(); ++i)
{
if (step.can_remove_required_output[i])
columns_to_remove.push_back(step.required_output[i]);
}
if (!columns_to_remove.empty())
{
auto columns = prewhere_info->prewhere_actions->getSampleBlock().getNamesAndTypesList();
ExpressionActionsPtr actions = std::make_shared<ExpressionActions>(columns, context_);
for (const auto & column : columns_to_remove)
actions->add(ExpressionAction::removeColumn(column));
prewhere_info->remove_columns_actions = std::move(actions);
}
columns_to_remove_after_prewhere = std::move(columns_to_remove);
}
else if (hasFilter())
{
/// Can't have prewhere and filter set simultaneously
filter_info->do_remove_column = chain.steps.at(0).can_remove_required_output.at(0);
}
if (hasWhere())
remove_where_filter = chain.steps.at(where_step_num).can_remove_required_output.at(0);
}
void ExpressionAnalysisResult::removeExtraColumns()
{
if (hasFilter())
filter_info->actions->prependProjectInput();
if (hasWhere())
before_where->prependProjectInput();
if (hasHaving())
before_having->prependProjectInput();
}
void ExpressionAnalysisResult::checkActions()
{
/// Check that PREWHERE doesn't contain unusual actions. Unusual actions are that can change number of rows.
if (hasPrewhere())
{
auto check_actions = [](const ExpressionActionsPtr & actions)
{
if (actions)
for (const auto & action : actions->getActions())
if (action.type == ExpressionAction::Type::JOIN || action.type == ExpressionAction::Type::ARRAY_JOIN)
throw Exception("PREWHERE cannot contain ARRAY JOIN or JOIN action", ErrorCodes::ILLEGAL_PREWHERE);
};
check_actions(prewhere_info->prewhere_actions);
check_actions(prewhere_info->alias_actions);
check_actions(prewhere_info->remove_columns_actions);
}
}
}