String CSVRowInputStream::getDiagnosticInfo()
{
if (istr.eof()) /// Buffer has gone, cannot extract information about what has been parsed.
return {};
WriteBufferFromOwnString out;
MutableColumns columns = header.cloneEmptyColumns();
/// It is possible to display detailed diagnostics only if the last and next to last rows are still in the read buffer.
size_t bytes_read_at_start_of_buffer = istr.count() - istr.offset();
if (bytes_read_at_start_of_buffer != bytes_read_at_start_of_buffer_on_prev_row)
{
out << "Could not print diagnostic info because two last rows aren't in buffer (rare case)\n";
return out.str();
}
size_t max_length_of_column_name = 0;
for (size_t i = 0; i < header.columns(); ++i)
if (header.safeGetByPosition(i).name.size() > max_length_of_column_name)
max_length_of_column_name = header.safeGetByPosition(i).name.size();
size_t max_length_of_data_type_name = 0;
for (size_t i = 0; i < header.columns(); ++i)
if (header.safeGetByPosition(i).type->getName().size() > max_length_of_data_type_name)
max_length_of_data_type_name = header.safeGetByPosition(i).type->getName().size();
/// Roll back the cursor to the beginning of the previous or current row and parse all over again. But now we derive detailed information.
if (pos_of_prev_row)
{
istr.position() = pos_of_prev_row;
out << "\nRow " << (row_num - 1) << ":\n";
if (!parseRowAndPrintDiagnosticInfo(columns, out, max_length_of_column_name, max_length_of_data_type_name))
return out.str();
}
else
{
if (!pos_of_current_row)
{
out << "Could not print diagnostic info because parsing of data hasn't started.\n";
return out.str();
}
istr.position() = pos_of_current_row;
}
out << "\nRow " << row_num << ":\n";
parseRowAndPrintDiagnosticInfo(columns, out, max_length_of_column_name, max_length_of_data_type_name);
out << "\n";
return out.str();
}
void TabSeparatedRowInputStream::printDiagnosticInfo(Block & block, WriteBuffer & out)
{
/// Вывести подробную диагностику возможно лишь если последняя и предпоследняя строка ещё находятся в буфере для чтения.
size_t bytes_read_at_start_of_buffer = istr.count() - istr.offset();
if (bytes_read_at_start_of_buffer != bytes_read_at_start_of_buffer_on_prev_row)
{
out << "Could not print diagnostic info because two last rows aren't in buffer (rare case)\n";
return;
}
size_t max_length_of_column_name = 0;
for (size_t i = 0; i < sample.columns(); ++i)
if (sample.getByPosition(i).name.size() > max_length_of_column_name)
max_length_of_column_name = sample.getByPosition(i).name.size();
size_t max_length_of_data_type_name = 0;
for (size_t i = 0; i < sample.columns(); ++i)
if (sample.getByPosition(i).type->getName().size() > max_length_of_data_type_name)
max_length_of_data_type_name = sample.getByPosition(i).type->getName().size();
/// Откатываем курсор для чтения на начало предыдущей или текущей строки и парсим всё заново. Но теперь выводим подробную информацию.
if (pos_of_prev_row)
{
istr.position() = pos_of_prev_row;
out << "\nRow " << (row_num - 1) << ":\n";
if (!parseRowAndPrintDiagnosticInfo(block, out, max_length_of_column_name, max_length_of_data_type_name))
return;
}
else
{
if (!pos_of_current_row)
{
out << "Could not print diagnostic info because parsing of data hasn't started.\n";
return;
}
istr.position() = pos_of_current_row;
}
out << "\nRow " << row_num << ":\n";
parseRowAndPrintDiagnosticInfo(block, out, max_length_of_column_name, max_length_of_data_type_name);
out << "\n";
}
