void GroupByScan::executeGroupBy() {
auto resultTab = createResultTableLayout();
auto groupResults = getInputHashTable();
// Allocate some memory for the result tab and resize the table
resultTab->resize(groupResults->numKeys());
pos_t row = 0;
typename HashTableType::map_const_iterator_t it1, it2, end;
// set iterators: in the sequential case, getInputTable() returns an AggregateHashTable, in the parallel case a HashTableView<>
// Alternatively, a common type could be introduced
if (_count < 1) {
auto aggregateHashTable = std::dynamic_pointer_cast<const HashTableType>(groupResults);
it1 = aggregateHashTable->getMapBegin();
end = aggregateHashTable->getMapEnd();
} else {
auto hashTableView = std::dynamic_pointer_cast<const HashTableView<MapType, KeyType> >(groupResults);
it1 = hashTableView->getMapBegin();
end = hashTableView->getMapEnd();
}
for (it2 = it1; it1 != end; it1 = it2) {
// outer loop over unique keys
auto pos_list = std::make_shared<pos_list_t>();
for (; (it2 != end) && (it1->first == it2->first); ++it2) {
// inner loop, all keys equal to it1->first
pos_list->push_back(it2->second);
}
writeGroupResult(resultTab, pos_list, row);
row++;
}
this->addResult(resultTab);
}
void PipeliningHashProbe::fetchPositions() {
const auto& probeTable = getProbeTable();
const auto& hash_table = std::dynamic_pointer_cast<const HashTable>(getInputHashTable(0));
assert(hash_table != nullptr);
LOG4CXX_DEBUG(logger, hash_table->stats());
LOG4CXX_DEBUG(logger, "Probe Table Size: " << probeTable->size());
LOG4CXX_DEBUG(logger, "Hash Table Size: " << hash_table->size());
for (pos_t probeTableRow = 0; probeTableRow < probeTable->size(); ++probeTableRow) {
pos_list_t matchingRows(hash_table->get(probeTable, _field_definition, probeTableRow));
if (!matchingRows.empty()) {
_buildTablePosList->insert(_buildTablePosList->end(), matchingRows.begin(), matchingRows.end());
_probeTablePosList->insert(_probeTablePosList->end(), matchingRows.size(), probeTableRow);
}
// We can only monitor the size of the output after each row has been probed.
// Each probed row can produce n matches. The output table will increase by those n rows.
// As soon as the current n causes the output table to reach the _chunkSize threshold,
// we will emit a chunk and reset the _buildTablePosList and _probeTablePosList.
if (_buildTablePosList->size() > _chunkSize) {
emitChunk();
}
}
// Emit final results.
if (_buildTablePosList->size() > 0) {
emitChunk();
}
LOG4CXX_DEBUG(logger, "Done Probing");
}
void HashJoinProbe::fetchPositions(storage::pos_list_t* buildTablePosList, storage::pos_list_t* probeTablePosList) {
const auto& probeTable = getProbeTable();
const auto& hash_table = std::dynamic_pointer_cast<const HashTable>(getInputHashTable(0));
assert(hash_table != nullptr);
LOG4CXX_DEBUG(logger, hash_table->stats());
LOG4CXX_DEBUG(logger, "Probe Table Size: " << probeTable->size());
LOG4CXX_DEBUG(logger, "Hash Table Size: " << hash_table->size());
for (pos_t probeTableRow = 0; probeTableRow < probeTable->size(); ++probeTableRow) {
pos_list_t matchingRows(hash_table->get(probeTable, _field_definition, probeTableRow));
if (!matchingRows.empty()) {
buildTablePosList->insert(buildTablePosList->end(), matchingRows.begin(), matchingRows.end());
probeTablePosList->insert(probeTablePosList->end(), matchingRows.size(), probeTableRow);
}
}
LOG4CXX_DEBUG(logger, "Done Probing");
}
void PipeliningHashProbe::setupPlanOperation() {
PlanOperation::setupPlanOperation();
setBuildTable(getInputHashTable()->getTable());
}
void HashJoinProbe::setupPlanOperation() {
PlanOperation::setupPlanOperation();
setBuildTable(getInputHashTable()->getTable());
}