#include "usql.h" #include "exception.h" #include "csvreader.h" #include "ml_date.h" #include #include namespace usql { std::unique_ptr USql::execute(const std::string &command) { std::unique_ptr node = m_parser.parse(command); return execute(*node); } std::unique_ptr
USql::execute(Node &node) { // TODO optimize execution nodes here switch (node.node_type) { case NodeType::create_table: return execute_create_table(static_cast(node)); case NodeType::create_table_as_select: return execute_create_table_as_table(static_cast(node)); case NodeType::insert_into: return execute_insert_into_table(static_cast(node)); case NodeType::select_from: return execute_select(static_cast(node)); case NodeType::delete_from: return execute_delete(static_cast(node)); case NodeType::update_table: return execute_update(static_cast(node)); case NodeType::load_table: return execute_load(static_cast(node)); case NodeType::save_table: return execute_save(static_cast(node)); default: return create_stmt_result_table(-1, "unknown statement"); } } std::unique_ptr
USql::execute_create_table(CreateTableNode &node) { // TODO check table does not exists Table table{node.table_name, node.cols_defs}; m_tables.push_back(table); return create_stmt_result_table(0, "table created"); } std::unique_ptr
USql::execute_create_table_as_table(CreateTableAsSelectNode &node) { // TODO check table does not exists auto select = execute_select((SelectFromTableNode &) *node.select_table); // create table Table new_table{node.table_name, select->m_col_defs}; m_tables.push_back(new_table); // copy rows // must be here, if rows are put into new_table, they are lost during m_tables.push_table Table *table = find_table(node.table_name); for( Row& orig_row : select->m_rows) { table->addCopyOfRow(orig_row); } select.release(); // is it correct? hoping not to release select table here and then when releasing CreateTableAsSelectNode return create_stmt_result_table(0, "table created"); } std::unique_ptr
USql::execute_insert_into_table(InsertIntoTableNode &node) { // TODO check column names.size = values.size // find table Table *table_def = find_table(node.table_name); // prepare empty new_row Row new_row = table_def->createEmptyRow(); // copy values for (size_t i = 0; i < node.cols_names.size(); i++) { ColDefNode col_def = table_def->get_column_def(node.cols_names[i].name); // TODO validate value auto value = evalValueNode(table_def, new_row, node.cols_values[i].get()); if (col_def.type == ColumnType::integer_type) { new_row.setColumnValue(col_def.order, value->getIntValue()); } else if (col_def.type == ColumnType::float_type) { new_row.setColumnValue(col_def.order, value->getDoubleValue()); } else { new_row.setColumnValue(col_def.order, value->getStringValue()); } } // append new_row table_def->addRow(new_row); return create_stmt_result_table(0, "insert succeeded"); } std::unique_ptr
USql::execute_select(SelectFromTableNode &node) { // find source table Table *table = find_table(node.table_name); // create result table std::vector result_tbl_col_defs{}; std::vector source_table_col_index{}; for (int i = 0; i < node.cols_names->size(); i++) { auto [ src_tbl_col_index, col_def ] = getColumnDefinition(table, &node.cols_names->operator[](i), i); source_table_col_index.push_back(src_tbl_col_index); result_tbl_col_defs.push_back(col_def); } auto result = std::make_unique
("result", result_tbl_col_defs); // execute access plan for (auto row = begin(table->m_rows); row != end(table->m_rows); ++row) { // eval where for row if (evalWhere(node.where.get(), table, *row)) { // prepare empty row Row new_row = result->createEmptyRow(); // copy column values for (auto idx = 0; idx < result->columns_count(); idx++) { auto row_col_index = source_table_col_index[idx]; if (row_col_index == -1) { // todo its function TODO col_names zmenit na colValues auto evaluated_value = evalValueNode(table, *row, node.cols_names->operator[](idx).value.get()); ValueNode *col_value = evaluated_value.get(); if (!col_value->isNull()) { // TODO sjednotit nasledujici if (result_tbl_col_defs[idx].type == ColumnType::integer_type) new_row.setColumnValue(idx, col_value->getIntValue()); if (result_tbl_col_defs[idx].type == ColumnType::float_type) new_row.setColumnValue(idx, col_value->getDoubleValue()); if (result_tbl_col_defs[idx].type == ColumnType::varchar_type) new_row.setColumnValue(idx, col_value->getStringValue()); } // TODO set to null } else { ColValue *col_value = row->ithColumn(row_col_index); if (!col_value->isNull()) { if (result_tbl_col_defs[idx].type == ColumnType::integer_type) new_row.setColumnValue(idx, col_value->getIntValue()); else if (result_tbl_col_defs[idx].type == ColumnType::float_type) new_row.setColumnValue(idx, col_value->getDoubleValue()); else if (result_tbl_col_defs[idx].type == ColumnType::varchar_type) new_row.setColumnValue(idx, col_value->getStringValue()); } // TODO set to null } } // add row to result result->m_rows.push_back(new_row); } } return std::move(result); } std::tuple USql::getColumnDefinition(Table *table, SelectColNode *select_col_node, int col_order ) { std::string new_col_name = select_col_node->name; if (select_col_node->value->node_type == NodeType::column_name) { ColDefNode cdef = table->get_column_def(new_col_name); return std::make_tuple(cdef.order, cdef); } else if (select_col_node->value->node_type == NodeType::function) { auto node = static_cast(select_col_node->value.get()); if (node->function == "to_string") { ColDefNode cdef = ColDefNode{new_col_name, ColumnType::varchar_type, col_order, 64, true}; return std::make_tuple(-1, cdef); } else if (node->function == "to_date") { ColDefNode cdef = ColDefNode{new_col_name, ColumnType::integer_type, col_order, 1, true}; return std::make_tuple(-1, cdef); } throw Exception("Unsupported function"); } throw Exception("Unsupported node type"); } std::unique_ptr
USql::execute_delete(DeleteFromTableNode &node) { // find source table Table *table = find_table(node.table_name); // execute access plan auto it = table->m_rows.begin(); for (; it != table->m_rows.end();) { if (evalWhere(node.where.get(), table, *it)) { // TODO this can be really expensive operation it = table->m_rows.erase(it); } else { ++it; } } return create_stmt_result_table(0, "delete succeeded"); } std::unique_ptr
USql::execute_update(UpdateTableNode &node) { // find source table Table *table = find_table(node.table_name); // execute access plan for (auto row = begin(table->m_rows); row != end(table->m_rows); ++row) { // eval where for row if (evalWhere(node.where.get(), table, *row)) { int i = 0; for (const auto& col : node.cols_names) { // TODO cache it like in select ColDefNode cdef = table->get_column_def(col.name); std::unique_ptr new_val = evalArithmeticOperator(cdef.type, static_cast(*node.values[i]), table, *row); if (cdef.type == ColumnType::integer_type) { row->setColumnValue(cdef.order, new_val->getIntValue()); } else if (cdef.type == ColumnType::float_type) { row->setColumnValue(cdef.order, new_val->getDoubleValue()); } else if (cdef.type == ColumnType::varchar_type) { row->setColumnValue(cdef.order, new_val->getStringValue()); } else { throw Exception("Implement me!"); } i++; } } } return create_stmt_result_table(0, "delete succeeded"); } std::unique_ptr
USql::execute_load(LoadIntoTableNode &node) { // find source table Table *table_def = find_table(node.table_name); // read data std::ifstream ifs(node.filename); std::string content((std::istreambuf_iterator(ifs)), (std::istreambuf_iterator())); CsvReader csvparser{}; auto csv = csvparser.parseCSV(content); std::vector &colDefs = table_def->m_col_defs; for (auto it = csv.begin() + 1; it != csv.end(); ++it) { std::vector csv_line = *it; // prepare empty new_row Row new_row = table_def->createEmptyRow(); // copy values for (size_t i = 0; i < table_def->columns_count(); i++) { ColDefNode col_def = table_def->get_column_def(colDefs[i].name); // TODO validate value if (col_def.type == ColumnType::integer_type) { new_row.setColumnValue(col_def.order, std::stol(csv_line[i])); } else if (col_def.type == ColumnType::float_type) { new_row.setColumnValue(col_def.order, std::stof(csv_line[i])); } else { new_row.setColumnValue(col_def.order, csv_line[i]); } } // append new_row table_def->addRow(new_row); } return create_stmt_result_table(0, "load succeeded"); } std::unique_ptr
USql::execute_save(SaveTableNode &node) { // find source table Table *table_def = find_table(node.table_name); // header std::string out_string; for(int i = 0; i < table_def->m_col_defs.size(); i++) { if (i > 0) out_string += ","; out_string += table_def->m_col_defs[i].name; } // rows for (auto it = table_def->m_rows.begin() + 1; it != table_def->m_rows.end(); ++it) { std::string csv_line; for(int i = 0; i < table_def->m_col_defs.size(); i++) { if (i > 0) csv_line += ","; auto col = it->ithColumn(i); if (!col->isNull()) { csv_line += col->getStringValue(); // TODO handle enclosing commas etc } } out_string += "\n"; out_string += csv_line; } // save data std::ofstream file(node.filename); file << out_string; file.close(); return create_stmt_result_table(0, "save succeeded"); } bool USql::evalWhere(Node *where, Table *table, Row &row) const { switch (where->node_type) { // no where clause case NodeType::true_node: return true; case NodeType::relational_operator: // just one condition return evalRelationalOperator(*((RelationalOperatorNode *) where), table, row); case NodeType::logical_operator: return evalLogicalOperator(*((LogicalOperatorNode *) where), table, row); default: throw Exception("Wrong node type"); } return false; } bool USql::evalRelationalOperator(const RelationalOperatorNode &filter, Table *table, Row &row) const { std::unique_ptr left_value = evalValueNode(table, row, filter.left.get()); std::unique_ptr right_value = evalValueNode(table, row, filter.right.get()); double comparator; if (left_value->node_type == NodeType::int_value && right_value->node_type == NodeType::int_value) { comparator = left_value->getIntValue() - right_value->getIntValue(); } else if ((left_value->node_type == NodeType::int_value && right_value->node_type == NodeType::float_value) || (left_value->node_type == NodeType::float_value && right_value->node_type == NodeType::int_value) || (left_value->node_type == NodeType::float_value && right_value->node_type == NodeType::float_value)) { comparator = left_value->getDoubleValue() - right_value->getDoubleValue(); } else if (left_value->node_type == NodeType::string_value || right_value->node_type == NodeType::string_value) { comparator = left_value->getStringValue().compare(right_value->getStringValue()); } else { // TODO throw exception } switch (filter.op) { case RelationalOperatorType::equal: return comparator == 0.0; case RelationalOperatorType::not_equal: return comparator != 0.0; case RelationalOperatorType::greater: return comparator > 0.0; case RelationalOperatorType::greater_equal: return comparator >= 0.0; case RelationalOperatorType::lesser: return comparator < 0.0; case RelationalOperatorType::lesser_equal: return comparator <= 0.0; } throw Exception("invalid relational operator"); } std::unique_ptr USql::evalValueNode(Table *table, Row &row, Node *node) { if (node->node_type == NodeType::database_value || node->node_type == NodeType::column_name) { // TODO sjednotit return evalDatabaseValueNode(table, row, node); } else if (node->node_type == NodeType::int_value || node->node_type == NodeType::float_value || node->node_type == NodeType::string_value) { return evalLiteralValueNode(table, row, node); } else if (node->node_type == NodeType::function) { return evalFunctionValueNode(table, row, node); } throw Exception("unsupported node type"); } std::unique_ptr USql::evalDatabaseValueNode(Table *table, Row &row, Node *node) { auto *dvl = static_cast(node); ColDefNode col_def = table->get_column_def( dvl->col_name); // TODO optimize it to just get this def once auto db_value = row.ithColumn(col_def.order); if (col_def.type == ColumnType::integer_type) { return std::__1::make_unique(db_value->getIntValue()); } if (col_def.type == ColumnType::float_type) { return std::__1::make_unique(db_value->getDoubleValue()); } if (col_def.type == ColumnType::varchar_type) { return std::__1::make_unique(db_value->getStringValue()); } throw Exception("unknown database value type"); } std::unique_ptr USql::evalLiteralValueNode(Table *table, Row &row, Node *node) { if (node->node_type == NodeType::int_value) { auto *ivl = static_cast(node); return std::make_unique(ivl->value); } else if (node->node_type == NodeType::float_value) { auto *ivl = static_cast(node); return std::make_unique(ivl->value); } else if (node->node_type == NodeType::string_value) { auto *ivl = static_cast(node); return std::make_unique(ivl->value); } throw Exception("invalid type"); } std::unique_ptr USql::evalFunctionValueNode(Table *table, Row &row, Node *node) { auto *fnc = static_cast(node); std::vector> evaluatedPars; for(auto & param : fnc->params) { evaluatedPars.push_back(evalValueNode(table, row, param.get())); } // TODO use some enum if (fnc->function == "lower") { std::string str = evaluatedPars[0]->getStringValue(); std::transform(str.begin(), str.end(), str.begin(), [](unsigned char c) -> unsigned char { return std::tolower(c); }); return std::make_unique(str); } if (fnc->function == "upper") { std::string str = evaluatedPars[0]->getStringValue(); std::transform(str.begin(), str.end(), str.begin(), [](unsigned char c) -> unsigned char { return std::toupper(c); }); return std::make_unique(str); } if (fnc->function == "to_date") { std::string date = evaluatedPars[0]->getStringValue(); std::string format = evaluatedPars[1]->getStringValue(); long epoch_time = string_to_date(date, format); return std::make_unique(epoch_time); } if (fnc->function == "to_string") { long date = evaluatedPars[0]->getIntValue(); std::string format = evaluatedPars[1]->getStringValue(); std::string formated_date = date_to_string(date, format); return std::make_unique(formated_date); } throw Exception("invalid function"); } bool USql::evalLogicalOperator(LogicalOperatorNode &node, Table *pTable, Row &row) const { bool left = evalRelationalOperator(static_cast(*node.left), pTable, row); if ((node.op == LogicalOperatorType::and_operator && !left) || (node.op == LogicalOperatorType::or_operator && left)) return left; bool right = evalRelationalOperator(static_cast(*node.right), pTable, row); return right; } std::unique_ptr USql::evalArithmeticOperator(ColumnType outType, ArithmeticalOperatorNode &node, Table *table, Row &row) const { if (node.op == ArithmeticalOperatorType::copy_value) { return evalValueNode(table, row, node.left.get()); } std::unique_ptr left = evalValueNode(table, row, node.left.get()); std::unique_ptr right = evalValueNode(table, row, node.right.get()); if (outType == ColumnType::float_type) { double l = ((ValueNode *) left.get())->getDoubleValue(); double r = ((ValueNode *) right.get())->getDoubleValue(); switch (node.op) { case ArithmeticalOperatorType::plus_operator: return std::make_unique(l + r); case ArithmeticalOperatorType::minus_operator: return std::make_unique(l - r); case ArithmeticalOperatorType::multiply_operator: return std::make_unique(l * r); case ArithmeticalOperatorType::divide_operator: return std::make_unique(l / r); default: throw Exception("implement me!!"); } } else if (outType == ColumnType::integer_type) { long l = ((ValueNode *) left.get())->getIntValue(); long r = ((ValueNode *) right.get())->getIntValue(); switch (node.op) { case ArithmeticalOperatorType::plus_operator: return std::make_unique(l + r); case ArithmeticalOperatorType::minus_operator: return std::make_unique(l - r); case ArithmeticalOperatorType::multiply_operator: return std::make_unique(l * r); case ArithmeticalOperatorType::divide_operator: return std::make_unique(l / r); default: throw Exception("implement me!!"); } } else if (outType == ColumnType::varchar_type) { std::string l = ((ValueNode *) left.get())->getStringValue(); std::string r = ((ValueNode *) right.get())->getStringValue(); switch (node.op) { case ArithmeticalOperatorType::plus_operator: return std::make_unique(l + r); default: throw Exception("implement me!!"); } } throw Exception("implement me!!"); } std::unique_ptr
USql::create_stmt_result_table(long code, const std::string& text) { std::vector result_tbl_col_defs{}; result_tbl_col_defs.push_back(ColDefNode("code", ColumnType::integer_type, 0, 1, false)); result_tbl_col_defs.push_back(ColDefNode("desc", ColumnType::varchar_type, 1, 255, false)); auto table_def = std::make_unique
("result", result_tbl_col_defs); Row new_row = table_def->createEmptyRow(); new_row.setColumnValue(0, code); new_row.setColumnValue(1, text); table_def->addRow(new_row); return std::move(table_def); } Table *USql::find_table(const std::string &name) { auto name_cmp = [name](const Table& t) { return t.m_name == name; }; auto table_def = std::find_if(begin(m_tables), end(m_tables), name_cmp); if (table_def != std::end(m_tables)) { return table_def.operator->(); } else { throw Exception("table not found (" + name + ")"); } } } // namespace