bool BlockStreamPerformanceTest::next(BlockStreamBase*){
block_->setEmpty();
if(state_.child1_->next(block_)){
BlockStreamBase::BlockStreamTraverseIterator* it=block_->createIterator();
void* data;
// printf("before while!<<<<</n");
// int count=0;
while(data=it->nextTuple()){
tuplecount1_++;
tuplecount_++;
if(*(int*)data==-1){
tuplecount1_++;
}
// printf("%d\n",count++);
}
return true;
}
else if(state_.child2_->next(block_)) {
BlockStreamBase::BlockStreamTraverseIterator* it = block_->createIterator();
while(it->nextTuple()) {
tuplecount2_++;
tuplecount_++;
}
return true;
}
cout << "performance test return false!\n";
return false;
}
// TODO(wangli): ideally, fetching blocks from child iterator in cases
// that the limit is exhausted is not necessary. However, in the current
// implementation, the child iterator sub-tree leaded by exchange
// lower iterator cannot be closed if not all the blocks are called.
bool PhysicalLimit::Next(BlockStreamBase* block) {
while (state_.child_->Next(block_for_asking_)) {
void* tuple_from_child;
BlockStreamBase::BlockStreamTraverseIterator* it =
block_for_asking_->createIterator();
while (NULL != (tuple_from_child = it->currentTuple())) {
if (!LimitExhausted()) {
if (!ShouldSkip()) {
const unsigned tuple_size =
state_.schema_->getTupleActualSize(tuple_from_child);
void* target_tuple;
if (NULL != (target_tuple = block->allocateTuple(tuple_size))) {
state_.schema_->copyTuple(tuple_from_child, target_tuple);
received_tuples_++;
tuple_cur_++;
it->increase_cur_();
} else {
it->~BlockStreamTraverseIterator();
return true;
}
} else {
tuple_cur_++;
it->increase_cur_();
continue;
}
} else {
break; // maybe the below statement is right ---YuKai
// return !block->Empty(); // to consume the next block;
}
}
}
return !block->Empty();
}
bool InOperator::Open(const PartitionOffset& partition_offset) {
state_.child_set_->Open(partition_offset);
state_.child_in_->Open(partition_offset);
AtomicPushFreeHtBlockStream(BlockStreamBase::createBlock(
state_.schema_child_set_, state_.block_size_));
AtomicPushFreeBlockStream(BlockStreamBase::createBlock(
state_.schema_child_in_, state_.block_size_));
if (sema_open_.try_wait()) {
// initialize hash table, use the child_set to build hash table
hash_func_ =
PartitionFunctionFactory::createBoostHashFunction(state_.ht_nbuckets_);
vector<unsigned> ht_index;
ht_index.push_back(state_.index_child_set_);
hash_table_ = new BasicHashTable(
state_.ht_nbuckets_, state_.ht_bucket_size_,
(state_.schema_child_set_->getSubSchema(ht_index))->getTupleMaxSize());
ht_index.clear();
open_finished_ = true;
} else {
while (!open_finished_) usleep(1);
}
void* cur_tuple = NULL;
void* tuple_in_hashtable = NULL;
unsigned bn = 0;
BlockStreamBase* bsb = AtomicPopFreeHtBlockStream();
while (state_.child_set_->Next(bsb)) {
BlockStreamBase::BlockStreamTraverseIterator* bsti = bsb->createIterator();
bsti->reset();
while (cur_tuple = bsti->nextTuple()) {
bn = state_.schema_child_set_->getcolumn(state_.index_child_set_)
.operate->GetPartitionValue(
state_.schema_child_set_->getColumnAddess(
state_.index_child_set_, cur_tuple),
state_.ht_nbuckets_);
tuple_in_hashtable = hash_table_->atomicAllocate(bn);
state_.schema_child_set_->getcolumn(state_.index_child_set_)
.operate->Assign(state_.schema_child_set_->getColumnAddess(
state_.index_child_set_, cur_tuple),
tuple_in_hashtable);
}
bsb->setEmpty();
}
barrier_->Arrive();
printf("-----------In Iterator Open Successful!-----------\n");
return true;
}
bool bottomLayerSorting::open(const PartitionOffset& partition_offset)
{
if (tryEntryIntoSerializedSection())
{
computeVectorSchema();
const bool child_open_return = state_.child_->open(partition_offset);
setReturnStatus(child_open_return);
}
barrierArrive();
//Construct the PartitionID for the next function to make up the ChunkID
partition_id_.projection_id = state_.projection_id_;
partition_id_.partition_off = partition_offset;
// Open finished. Buffer all the child create dataset in different group according to their ChunkIDs
BlockStreamBase* block_for_asking = BlockStreamBase::createBlock(state_.schema_, state_.block_size_);
block_for_asking->setEmpty();
BlockStreamBase::BlockStreamTraverseIterator* iterator = NULL;
void* current_chunk = new ChunkOffset;
Operate* op_ = state_.schema_->getcolumn(1).operate->duplicateOperator();
while (state_.child_->next(block_for_asking))
{
iterator = block_for_asking->createIterator();
void* current_tuple = NULL;
while((current_tuple = iterator->nextTuple()) != 0)
{
state_.schema_->getColumnValue(0, current_tuple, current_chunk);
if(tuples_in_chunk_.find(*(ChunkOffset*)current_chunk)==tuples_in_chunk_.end()){
vector<compare_node*> tmp;
tuples_in_chunk_[*(ChunkOffset*)current_chunk] = tmp;
}
compare_node* c_node = (compare_node*)malloc(sizeof(compare_node)); //newmalloc
c_node->vector_schema_ = vector_schema_;
c_node->tuple_ = malloc(vector_schema_->getTupleMaxSize()); //newmalloc
vector_schema_->copyTuple((char*)current_tuple+state_.schema_->getcolumn(0).get_length(),c_node->tuple_);
// c_node->tuple_ = current_tuple+state_.schema_->getcolumn(0).get_length();
// c_node->op_ = state_.schema_->getcolumn(1).operate->duplicateOperator();
c_node->op_ = op_;
tuples_in_chunk_.find(*(ChunkOffset*)current_chunk)->second.push_back(c_node);
//for testing begin
// if ((*(ChunkOffset*)current_chunk) == 0)
// {
// cout << "current chunk: " << *(ChunkOffset*)current_chunk << " tuple: ";
// vector_schema_->displayTuple(current_tuple+state_.schema_->getcolumn(0).get_length(), " | ");
// vector_schema_->displayTuple(tuples_in_chunk_.find(*(ChunkOffset*)current_chunk)->second.back()->tuple_, " | ");
// sleep(1);
// }
//for testing end
}
block_for_asking->setEmpty();
}
//for testing begin
// sleep(10000);
// cout << "Chunk Num: " << tuples_in_chunk_.size() << endl;
// sleep(1000);
//for testing end
// Sorting the tuples in each chunk
/*for testing*/ cout << "Chunk num: " << tuples_in_chunk_.size() << endl;
for (std::map<ChunkOffset, vector<compare_node*> >::iterator iter = tuples_in_chunk_.begin(); iter != tuples_in_chunk_.end(); iter++)
{
///*for testing*/ cout << "chunk id: " << *(unsigned short*)iter->first << endl;
//for testing begin
cout << "Chunk size: " << iter->second.size() << endl;
// for (unsigned i = 0; i < iter->second.size(); i++)
// {
// vector_schema_->displayTuple(iter->second[i]->tuple_, "\t");
//// sleep(1);
// }
// sleep(1000);
//for testing end
stable_sort(iter->second.begin(), iter->second.end(), compare);
//for testing begin
// for (unsigned i = 0; i < iter->second.size(); i++)
// {
// vector_schema_->displayTuple(iter->second[i]->tuple_, "\t");
//// sleep(1);
// }
// sleep(1000);
//for testing end
}
return getReturnStatus();
}
