unsigned long Analyzer::getDistinctCardinality(const AttributeID& attr_id) {
LogicalOperator* scan = new LogicalScan(
Catalog::getInstance()->getTable(attr_id.table_id)->getProjectoin(0));
std::vector<Attribute> group_by_attributes;
group_by_attributes.push_back(
Catalog::getInstance()->getTable(attr_id.table_id)->getAttribute(
attr_id.offset));
LogicalOperator* agg = new LogicalAggregation(
group_by_attributes, std::vector<Attribute>(),
std::vector<PhysicalAggregation::State::Aggregation>(), scan);
std::vector<Attribute> aggregation_attributes;
aggregation_attributes.push_back(Attribute(ATTRIBUTE_ANY));
std::vector<PhysicalAggregation::State::Aggregation> aggregation_function;
aggregation_function.push_back(PhysicalAggregation::State::kCount);
LogicalOperator* count_agg =
new LogicalAggregation(std::vector<Attribute>(), aggregation_attributes,
aggregation_function, agg);
LogicalOperator* root = new LogicalQueryPlanRoot(
0, count_agg, LogicalQueryPlanRoot::kResultCollector);
PhysicalOperatorBase* collector =
root->GetPhysicalPlan(1024 * 64 - sizeof(unsigned));
collector->Open();
collector->Next(0);
collector->Close();
ResultSet* resultset = collector->GetResultSet();
ResultSet::Iterator it = resultset->createIterator();
BlockStreamBase::BlockStreamTraverseIterator* b_it =
it.nextBlock()->createIterator();
const unsigned long distinct_cardinality = *(unsigned long*)b_it->nextTuple();
resultset->destory();
collector->~PhysicalOperatorBase();
root->~LogicalOperator();
return distinct_cardinality;
}
void Analyzer::compute_table_stat(const TableID& tab_id) {
TableDescriptor* table = Catalog::getInstance()->getTable(tab_id);
LogicalOperator* scan = new LogicalScan(table->getProjectoin(0));
std::vector<Attribute> group_by_attributes;
std::vector<Attribute> aggregation_attributes;
aggregation_attributes.push_back(Attribute(ATTRIBUTE_ANY));
std::vector<PhysicalAggregation::State::Aggregation> aggregation_function;
aggregation_function.push_back(PhysicalAggregation::State::kCount);
LogicalOperator* agg = new LogicalAggregation(
group_by_attributes, aggregation_attributes, aggregation_function, scan);
LogicalOperator* root =
new LogicalQueryPlanRoot(0, agg, LogicalQueryPlanRoot::kResultCollector);
PhysicalOperatorBase* collector =
root->GetPhysicalPlan(1024 * 64 - sizeof(unsigned));
collector->Open();
collector->Next(0);
collector->Close();
ResultSet* resultset = collector->GetResultSet();
ResultSet::Iterator it = resultset->createIterator();
BlockStreamBase::BlockStreamTraverseIterator* b_it =
it.nextBlock()->createIterator();
const unsigned long tuple_count = *(unsigned long*)b_it->nextTuple();
BlockStreamBase* block;
while (block = it.nextBlock()) {
BlockStreamBase::BlockStreamTraverseIterator* b_it =
block->createIterator();
}
TableStatistic* tab_stat = new TableStatistic();
tab_stat->number_of_tuples_ = tuple_count;
printf("Statistics for table %s is gathered!\n",
Catalog::getInstance()->getTable(tab_id)->getTableName().c_str());
tab_stat->print();
StatManager::getInstance()->setTableStatistic(tab_id, tab_stat);
resultset->destory();
root->~LogicalOperator();
}
void Analyzer::analyse(const AttributeID &attrID) {
Catalog *catalog = Catalog::getInstance();
TableDescriptor* table = catalog->getTable(attrID.table_id);
ProjectionDescriptor * projection = NULL;
unsigned pidSize = table->getNumberOfProjection();
const Attribute attr = table->getAttribute(attrID.offset);
for (unsigned i = 0; i < pidSize; ++i) {
if (table->getProjectoin(i)->hasAttribute(attr)) {
projection = table->getProjectoin(i);
break;
}
}
std::vector<Attribute> group_by_attributes;
std::vector<Attribute> aggregation_attributes;
group_by_attributes.push_back(attr);
aggregation_attributes.push_back(attr);
std::vector<BlockStreamAggregationIterator::State::aggregation> aggregation_function;
aggregation_function.push_back(
BlockStreamAggregationIterator::State::count);
LogicalOperator* sb_payload_scan = new LogicalScan(projection);
LogicalOperator* aggregation = new Aggregation(group_by_attributes,
aggregation_attributes, aggregation_function, sb_payload_scan);
const NodeID collector_node_id = 0;
LogicalOperator* root = new LogicalQueryPlanRoot(collector_node_id,
aggregation, LogicalQueryPlanRoot::RESULTCOLLECTOR);
BlockStreamIteratorBase* collector = root->getIteratorTree(
1024 * 64 - sizeof(unsigned));
collector->open();
collector->next(0);
collector->close();
ResultSet* resultset = collector->getResultSet();
ResultSet::Iterator it = resultset->createIterator();
BlockStreamBase* block;
void* tuple;
BlockStreamBase::BlockStreamTraverseIterator *block_it;
unsigned long valueCount = resultset->getNumberOftuples();
unsigned long tupleCount = 0;
TuplePtr *list = new TuplePtr[valueCount];
unsigned long i = 0;
while (block = (BlockStreamBase*) it.atomicNextBlock()) {
block_it = block->createIterator();
while (tuple = block_it->nextTuple()) {
list[i++] = tuple;
tupleCount += getFrequency(tuple, attr.attrType);
}
}
int magicNumber = 100;
StatisticOnTable *stat = new StatisticOnTable(magicNumber);
stat->setValueCount(valueCount);
stat->setTupleCount(tupleCount);
qsort_r(list, valueCount, sizeof(void *), compare,
(void *) (attr.attrType->operate));
mcvAnalyse(list, valueCount, attr, (Histogram *) stat);
equiDepthAnalyse(list, valueCount, attr, (Histogram *) stat);
// StatManager::getInstance()->addStat(attrID, stat);
StatManager::getInstance()->getTableStatistic(attrID.table_id);
delete list;
resultset->destory();
}
Histogram* Analyzer::computeHistogram(const AttributeID& attr_id,
const unsigned nbuckets) {
printf("Compute for histogram for attribute %s (%d buckets)\n",
Catalog::getInstance()
->getTable(attr_id.table_id)
->getAttribute(attr_id.offset)
.attrName.c_str(),
nbuckets);
Catalog* catalog = Catalog::getInstance();
TableDescriptor* table = catalog->getTable(attr_id.table_id);
ProjectionDescriptor* projection = NULL;
unsigned pidSize = table->getNumberOfProjection();
const Attribute attr = table->getAttribute(attr_id.offset);
for (unsigned i = 0; i < pidSize; ++i) {
if (table->getProjectoin(i)->hasAttribute(attr)) {
projection = table->getProjectoin(i);
break;
}
}
std::vector<Attribute> group_by_attributes;
std::vector<Attribute> aggregation_attributes;
group_by_attributes.push_back(attr);
aggregation_attributes.push_back(attr);
std::vector<PhysicalAggregation::State::Aggregation> aggregation_function;
aggregation_function.push_back(PhysicalAggregation::State::kCount);
LogicalOperator* sb_payload_scan = new LogicalScan(projection);
LogicalOperator* aggregation =
new LogicalAggregation(group_by_attributes, aggregation_attributes,
aggregation_function, sb_payload_scan);
const NodeID collector_node_id = 0;
LogicalOperator* root = new LogicalQueryPlanRoot(
collector_node_id, aggregation, LogicalQueryPlanRoot::kResultCollector);
PhysicalOperatorBase* collector =
root->GetPhysicalPlan(1024 * 64 - sizeof(unsigned));
collector->Open();
collector->Next(0);
collector->Close();
ResultSet* resultset = collector->GetResultSet();
ResultSet::Iterator it = resultset->createIterator();
BlockStreamBase* block;
void* tuple;
BlockStreamBase::BlockStreamTraverseIterator* block_it;
unsigned long valueCount = resultset->getNumberOftuples();
unsigned long tupleCount = 0;
TuplePtr* list = new TuplePtr[valueCount];
unsigned long i = 0;
while (block = (BlockStreamBase*)it.atomicNextBlock()) {
block_it = block->createIterator();
while (tuple = block_it->nextTuple()) {
list[i++] = tuple;
tupleCount += getFrequency(tuple, attr.attrType);
}
}
Histogram* stat = new Histogram(nbuckets);
stat->setValueCount(valueCount);
stat->setTupleCount(tupleCount);
qsort_r(list, valueCount, sizeof(void*), compare,
(void*)(attr.attrType->operate));
mcvAnalyse(list, valueCount, attr, (Histogram*)stat);
equiDepthAnalyse(list, valueCount, attr, (Histogram*)stat);
// StatManager::getInstance()->addStat(attrID, stat);
// StatManager::getInstance()->getTableStatistic(attrID.table_id)
delete list;
resultset->destory();
return stat;
}