bool filterBlockWithQuery(ASTPtr query, Block & block, const Context & context)
{
query = query->clone();
const ASTSelectQuery & select = typeid_cast<ASTSelectQuery & >(*query);
if (!select.where_expression && !select.prewhere_expression)
return false;
NameSet columns;
for (const auto & it : block.getColumnsList())
columns.insert(it.name);
/// Составим выражение, вычисляющее выражения в WHERE и PREWHERE, зависящие только от имеющихся столбцов.
std::vector<ASTPtr> functions;
if (select.where_expression)
extractFunctions(select.where_expression, columns, functions);
if (select.prewhere_expression)
extractFunctions(select.prewhere_expression, columns, functions);
ASTPtr expression_ast = buildWhereExpression(functions);
if (!expression_ast)
return false;
/// Распарсим и вычислим выражение.
ExpressionAnalyzer analyzer(expression_ast, context, {}, block.getColumnsList());
ExpressionActionsPtr actions = analyzer.getActions(false);
actions->execute(block);
/// Отфильтруем блок.
String filter_column_name = expression_ast->getColumnName();
ColumnPtr filter_column = block.getByName(filter_column_name).column;
if (auto converted = filter_column->convertToFullColumnIfConst())
filter_column = converted;
const IColumn::Filter & filter = dynamic_cast<ColumnUInt8 &>(*filter_column).getData();
if (std::accumulate(filter.begin(), filter.end(), 0ul) == filter.size())
return false;
for (size_t i = 0; i < block.columns(); ++i)
{
ColumnPtr & column = block.safeGetByPosition(i).column;
column = column->filter(filter, -1);
}
return true;
}
bool filterBlockWithQuery(const ASTPtr & query, Block & block, const Context & context)
{
const ASTSelectQuery & select = typeid_cast<const ASTSelectQuery & >(*query);
if (!select.where_expression && !select.prewhere_expression)
return false;
NameSet columns;
for (const auto & it : block.getColumnsList())
columns.insert(it.name);
/// We will create an expression that evaluates the expressions in WHERE and PREWHERE, depending only on the existing columns.
std::vector<ASTPtr> functions;
if (select.where_expression)
extractFunctions(select.where_expression, columns, functions);
if (select.prewhere_expression)
extractFunctions(select.prewhere_expression, columns, functions);
ASTPtr expression_ast = buildWhereExpression(functions);
if (!expression_ast)
return false;
/// Let's parse and calculate the expression.
ExpressionAnalyzer analyzer(expression_ast, context, {}, block.getColumnsList());
ExpressionActionsPtr actions = analyzer.getActions(false);
actions->execute(block);
/// Filter the block.
String filter_column_name = expression_ast->getColumnName();
ColumnPtr filter_column = block.getByName(filter_column_name).column;
if (auto converted = filter_column->convertToFullColumnIfConst())
filter_column = converted;
const IColumn::Filter & filter = dynamic_cast<ColumnUInt8 &>(*filter_column).getData();
if (countBytesInFilter(filter) == 0)
return false;
for (size_t i = 0; i < block.columns(); ++i)
{
ColumnPtr & column = block.safeGetByPosition(i).column;
column = column->filter(filter, -1);
}
return true;
}
int main(int argc, char ** argv)
{
using namespace DB;
try
{
std::string input = "SELECT x, s1, s2, "
"/*"
"2 + x * 2, x * 2, x % 3 == 1, "
"s1 == 'abc', s1 == s2, s1 != 'abc', s1 != s2, "
"s1 < 'abc', s1 < s2, s1 > 'abc', s1 > s2, "
"s1 <= 'abc', s1 <= s2, s1 >= 'abc', s1 >= s2, "
"*/"
"s1 < s2 AND x % 3 < x % 5";
ParserSelectQuery parser;
ASTPtr ast = parseQuery(parser, input.data(), input.data() + input.size(), "", 0);
formatAST(*ast, std::cerr);
std::cerr << std::endl;
Context context = Context::createGlobal();
NamesAndTypesList columns
{
{"x", std::make_shared<DataTypeInt16>()},
{"s1", std::make_shared<DataTypeString>()},
{"s2", std::make_shared<DataTypeString>()}
};
auto syntax_result = SyntaxAnalyzer(context, {}).analyze(ast, columns);
ExpressionAnalyzer analyzer(ast, syntax_result, context);
ExpressionActionsChain chain(context);
analyzer.appendSelect(chain, false);
analyzer.appendProjectResult(chain);
chain.finalize();
ExpressionActionsPtr expression = chain.getLastActions();
size_t n = argc == 2 ? atoi(argv[1]) : 10;
Block block;
{
ColumnWithTypeAndName column;
column.name = "x";
column.type = std::make_shared<DataTypeInt16>();
auto col = ColumnInt16::create();
auto & vec_x = col->getData();
vec_x.resize(n);
for (size_t i = 0; i < n; ++i)
vec_x[i] = i % 9;
column.column = std::move(col);
block.insert(column);
}
const char * strings[] = {"abc", "def", "abcd", "defg", "ac"};
{
ColumnWithTypeAndName column;
column.name = "s1";
column.type = std::make_shared<DataTypeString>();
auto col = ColumnString::create();
for (size_t i = 0; i < n; ++i)
col->insert(std::string(strings[i % 5]));
column.column = std::move(col);
block.insert(column);
}
{
ColumnWithTypeAndName column;
column.name = "s2";
column.type = std::make_shared<DataTypeString>();
auto col = ColumnString::create();
for (size_t i = 0; i < n; ++i)
col->insert(std::string(strings[i % 3]));
column.column = std::move(col);
block.insert(column);
}
{
Stopwatch stopwatch;
stopwatch.start();
expression->execute(block);
stopwatch.stop();
std::cout << std::fixed << std::setprecision(2)
<< "Elapsed " << stopwatch.elapsedSeconds() << " sec."
<< ", " << n / stopwatch.elapsedSeconds() << " rows/sec."
<< std::endl;
}
auto is = std::make_shared<OneBlockInputStream>(block);
LimitBlockInputStream lis(is, 20, std::max(0, static_cast<int>(n) - 20));
WriteBufferFromOStream out_buf(std::cout);
BlockOutputStreamPtr out = FormatFactory::instance().getOutput("TabSeparated", out_buf, block, context);
copyData(lis, *out);
}
catch (const Exception & e)
{
std::cerr << e.displayText() << std::endl;
}
return 0;
}