HeapFile *SortFile(HeapFile *S, int len, int offset)
{
Status s;
Scan *scan;
scan = S->OpenScan(s);
if (s != OK)
{
cerr << "ERROR : cannot open scan on the heapfile to sort.\n";
}
//
// Scan the HeapFile S, new a B+Tree and insert the records into B+Tree.
//
BTreeFile *btree;
btree = new BTreeFile (s, "BTree", ATTR_INT, sizeof(int));
char *recPtr = new char[len];
int recLen = len;
RecordID rid;
while (scan->GetNext(rid, recPtr, recLen) == OK)
{
btree->Insert(recPtr + offset, rid);
}
delete scan;
HeapFile *sorted;
sorted = new HeapFile(NULL, s); // create a temp HeapFile
if (s != OK)
{
cerr << "Cannot create new file for sortedS\n";
}
//
// Now scan the B+-Tree and insert the records into a
// new (sorted) HeapFile.
//
BTreeFileScan *btreeScan;
btreeScan = (BTreeFileScan *)btree->OpenScan(NULL, NULL);
int key;
while (btreeScan->GetNext(rid, &key) == OK)
{
S->GetRecord (rid, recPtr, recLen);
sorted->InsertRecord (recPtr, recLen, rid);
}
btree->DestroyFile();
delete btree;
delete btreeScan;
delete [] recPtr;
return sorted;
}
//--------------------------------------------------------------------
// JoinMethod::SortFile
//
// Purpose : Sorts a relation by an integer attribute.
// Input : file - pointer to the HeapFile to be sorted.
// len - length of the records in the file. (assume fixed size).
// offset - offset of the sort attribute from the beginning of the record.
// Method : We create a B+-Tree using that attribute as the key. Then
// we scan the B+-Tree and insert the records into a new
// HeapFile. he HeapFile guarantees that the order of
// insertion will be the same as the order of scan later.
// Return : The new sorted relation/HeapFile.
//--------------------------------------------------------------------
HeapFile* JoinMethod::SortHeapFile(HeapFile *file, int len, int offset) {
Status s;
Scan *scan;
scan = file->OpenScan(s);
if (s != OK) {
std::cerr << "ERROR : cannot open scan on the heapfile to sort." << std::endl;
}
//
// Scan the HeapFile S, create a new B+Tree and insert the records into B+Tree.
//
BTreeFile *btree;
btree = new BTreeFile (s, "BTree");
char* recPtr = new char[len];
int recLen = len;
RecordID rid;
char* recKey = new char[100];
while (scan->GetNext(rid, recPtr, recLen) == OK)
{
int* valPtr = (int*)(recPtr+offset);
int val = *valPtr;
toString(val,recKey);
btree->Insert(recKey, rid);
}
delete scan;
delete [] recKey;
//std::cout << "created B+ tree!" << std::endl;
HeapFile *sorted = new HeapFile(NULL, s); // create a temp HeapFile
if (s != OK)
{
std::cerr << "Cannot create new file for sortedS\n";
}
// Now scan the B+-Tree and insert the records into a
// new (sorted) HeapFile.
BTreeFileScan* btreeScan = btree->OpenScan(NULL, NULL);
//int key;
char* keyPtr;
while (btreeScan->GetNext(rid, keyPtr) == OK)
{
//std::cout << "scanning " << rid << " " << keyPtr << std::endl;
file->GetRecord (rid, recPtr, recLen);
sorted->InsertRecord (recPtr, recLen, rid);
}
btree->DestroyFile();
delete btree;
delete btreeScan;
delete [] recPtr;
return sorted;
}
HeapFile* IndexNestedLoopJoin(JoinSpec specOfR, JoinSpec specOfS)
{
Status status = OK;
// Create a HeapFile for join results
HeapFile* joinedFile = new HeapFile(NULL, status);
if (OK != status)
{
cerr << "ERROR: cannot create a file for the joined relation.\n";
return NULL;
}
int recLenR = specOfR.recLen;
int recLenS = specOfS.recLen;
int recLenJoined = recLenR + recLenS;
char* recR = new char[recLenR];
char* recS = new char[recLenS];
char* recJoined = new char[recLenJoined];
RecordID ridR, ridS, ridJoined;
// Build the B+-tree index on the inner relation (S)
Scan* scanS = specOfS.file->OpenScan(status);
if (OK != status)
{
cerr << "ERROR: cannot open scan on the relation S heap file.\n";
return NULL;
}
BTreeFile* bTree = new BTreeFile(status, "IJBT", ATTR_INT, sizeof(int));
while (OK == scanS->GetNext(ridS, recS, recLenS))
{
bTree->Insert(recS + specOfS.offset, ridS);
}
delete scanS;
// Iterate through the outer relation (R) and join
Scan* scanR = specOfR.file->OpenScan(status);
if (OK != status)
{
cerr << "ERROR: cannot open scan on the relation R heap file.\n";
return NULL;
}
while (OK == scanR->GetNext(ridR, recR, recLenR))
{
int* joinArgR = (int*)&recR[specOfR.offset];
BTreeFileScan* bTreeScan = (BTreeFileScan*)bTree->OpenSearchScan(joinArgR, joinArgR);
int key;
while (OK == bTreeScan->GetNext(ridS, &key))
{
specOfS.file->GetRecord(ridS, recS, recLenS);
MakeNewRecord(recJoined, recR, recS, recLenR, recLenS);
joinedFile->InsertRecord(recJoined, recLenJoined, ridJoined);
}
delete bTreeScan;
}
// Release the allocated resources
delete scanR;
delete[] recR;
delete[] recS;
delete[] recJoined;
delete bTree;
return joinedFile;
}
//---------------------------------------------------------------
// TupleNestedLoop::Execute
//
// Input: left - The left relation to join.
// right - The right relation to join.
// Output: out - The relation to hold the ouptut.
// Return: OK if join completed succesfully. FAIL otherwise.
//
// Purpose: Performs a nested loop join on relations left and right
// a tuple a time. You can assume that left is the outer
// relation and right is the inner relation.
//---------------------------------------------------------------
Status TupleNestedLoops::Execute(JoinSpec& left, JoinSpec& right, JoinSpec& out) {
JoinMethod::Execute(left, right, out);
// Create the temporary heapfile
Status st;
HeapFile *tmpHeap = new HeapFile(NULL, st);
if (st != OK) {
std::cerr << "Failed to create output heapfile." << std::endl;
return FAIL;
}
// Open scan on left relation
Status leftStatus;
Scan *leftScan = left.file->OpenScan(leftStatus);
if (leftStatus != OK) {
std::cerr << "Failed to open scan on left relation." << std::endl;
return FAIL;
}
// Loop over the left relation
char *leftRec = new char[left.recLen];
while (true) {
RecordID leftRid;
leftStatus = leftScan->GetNext(leftRid, leftRec, left.recLen);
if (leftStatus == DONE) break;
if (leftStatus != OK) return FAIL;
// The join attribute on left relation
int *leftJoinValPtr = (int*)(leftRec + left.offset);
// Open scan on right relation
Status rightStatus;
Scan *rightScan = right.file->OpenScan(rightStatus);
if (rightStatus != OK) {
std::cerr << "Failed to open scan on right relation." << std::endl;
return FAIL;
}
// Loop over right relation
char *rightRec = new char[right.recLen];
while (true) {
RecordID rightRid;
rightStatus = rightScan->GetNext(rightRid, rightRec, right.recLen);
if (rightStatus == DONE) break;
if (rightStatus != OK) return FAIL;
// Compare join attribute
int *rightJoinValPtr = (int*)(rightRec + right.offset);
if (*leftJoinValPtr == *rightJoinValPtr) {
// Create the record and insert into tmpHeap...
char *joinedRec = new char[out.recLen];
MakeNewRecord(joinedRec, leftRec, rightRec, left, right);
RecordID insertedRid;
Status tmpStatus = tmpHeap->InsertRecord(joinedRec, out.recLen, insertedRid);
if (tmpStatus != OK) {
std::cerr << "Failed to insert tuple into output heapfile." << std::endl;
return FAIL;
}
delete [] joinedRec;
}
}
delete [] rightRec;
delete rightScan;
}
out.file = tmpHeap;
delete leftScan;
delete [] leftRec;
return OK;
}
HeapFile* BlockNestedLoopJoin(JoinSpec specOfR, JoinSpec specOfS, int B)
{
Status status = OK;
// Create a HeapFile for join results
HeapFile* joinedFile = new HeapFile(NULL, status);
if (OK != status)
{
cerr << "ERROR: cannot create a file for the joined relation.\n";
return NULL;
}
int recLenR = specOfR.recLen;
int recLenS = specOfS.recLen;
int recLenJoined = recLenR + recLenS;
char* recBlockR = new char[B]; // Allocate memory for the block
char* recS = new char[recLenS];
char* recJoined = new char[recLenJoined];
RecordID ridR, ridS, ridJoined;
Scan* scanR = specOfR.file->OpenScan(status);
if (OK != status)
{
cerr << "ERROR: cannot open scan on the relation R heap file.\n";
return NULL;
}
const int recordsPerBlock = B / recLenR;
bool lastBlock = false;
while (!lastBlock)
{
// Fill the block
int i;
for (i = 0; i < recordsPerBlock; i++)
{
if (OK != scanR->GetNext(ridR, recBlockR + i*recLenR, recLenR))
{
lastBlock = true;
break;
}
}
int lastRecordIndex = i;
Scan* scanS = specOfS.file->OpenScan(status);
if (OK != status)
{
cerr << "ERROR: cannot open scan on the relation S heap file.\n";
return NULL;
}
while (OK == scanS->GetNext(ridS, recS, recLenS))
{
int* joinArgS = (int*)&recS[specOfS.offset];
for (int currentRecordIndex = 0; currentRecordIndex < lastRecordIndex; currentRecordIndex++)
{
char* currentRecordPtr = recBlockR + (currentRecordIndex * recLenR);
int* joinArgR = (int*)(currentRecordPtr + specOfR.offset);
if (*joinArgR == *joinArgS)
{
MakeNewRecord(recJoined, currentRecordPtr, recS, recLenR, recLenS);
joinedFile->InsertRecord(recJoined, recLenJoined, ridJoined);
}
}
}
delete scanS;
}
// Release the allocated resources
delete scanR;
delete[] recBlockR;
delete[] recS;
delete[] recJoined;
return joinedFile;
}
//---------------------------------------------------------------
// BlockNestedLoop::Execute
//
// Input: left - The left relation to join.
// right - The right relation to join.
// Output: out - The relation to hold the ouptut.
// Return: OK if join completed succesfully. FAIL otherwise.
//
// Purpose: Performs a block nested loops join on the specified relations.
// You can find a specification of this algorithm on page 455. You should
// choose the smaller of the two relations to be the outer relation, but you
// should make sure to concatenate the tuples in order <left, right> when
// producing output. The block size can be specified in the constructor,
// and is stored in the variable blockSize.
//---------------------------------------------------------------
Status BlockNestedLoops::Execute(JoinSpec& left, JoinSpec& right, JoinSpec& out) {
JoinMethod::Execute(left, right, out);
Status s;
HeapFile* tmpHeap = new HeapFile(NULL, s);
if (s != OK) {
std::cout << "Creating new Heap File Failed" << std::endl;
return FAIL;
}
Scan * leftScan = left.file->OpenScan(s);
if (s != OK) {
std::cout << "Open scan left failed" << std::endl;
return FAIL;
}
Scan * rightScan = right.file->OpenScan(s);
if (s != OK) {
std::cout << "Open scan left failed" << std::endl;
return FAIL;
}
RecordID leftRid, rightRid, rightFirstRid, outRid;
// array to hold the "block" in memory
char* blockArray = new char[left.recLen * blockSize];
int blockArraySize = 0; // size in case of half full block
int* leftCurrRec = (int *)blockArray;
int* leftRec = new int[left.numOfAttr];
int* rightRec = new int[right.numOfAttr];
int leftRecLen = left.recLen;
int rightRecLen = right.recLen;
char* newRec = new char[left.recLen + right.recLen];
rightFirstRid = rightScan->currRid;
Status st = OK;
while (true) {
// fill the block with as many records as possible
if (blockArraySize < blockSize) {
st = leftScan->GetNext(leftRid, (char *)leftRec, leftRecLen);
if (st != DONE) {
memcpy(blockArray + left.recLen * blockArraySize, leftRec, left.recLen);
blockArraySize++;
continue;
}
}
// scan through the right, and scan the block in memory for joins
while (rightScan->GetNext(rightRid, (char *)rightRec, rightRecLen) != DONE) {
for (int j = 0; j < blockSize; j++) {
if (j >= blockArraySize) {
break;
}
leftCurrRec = (int *) (blockArray + left.recLen * j);
if (leftCurrRec[left.joinAttr] == rightRec[right.joinAttr]) {
MakeNewRecord(newRec, (char *)leftCurrRec, (char *)rightRec, left, right);
tmpHeap->InsertRecord(newRec, left.recLen + right.recLen, outRid);
}
}
}
rightScan->MoveTo(rightFirstRid);
blockArraySize = 0;
if (st == DONE) {
break;
}
}
out.file = tmpHeap;
//std::cout << "NUM BNL: " << tmpHeap->GetNumOfRecords() << std::endl;
delete leftScan;
delete rightScan;
delete blockArray;
delete leftRec;
delete rightRec;
delete newRec;
return OK;
}
//-------------------------------------------------------------------
// Sort::PassZero
//
// Input : None
// Output : Number of temp files generated after this pass
// Return : OK if Pass 0 succeeds, FAIL otherwise
//-------------------------------------------------------------------
Status Sort::PassZero(int &numTempFiles) {
// Get input file
Status status;
HeapFile inputFile(_inFile, status);
if (status != OK) return ReturnFAIL("Opening input file in PassZero function failed.");
int numRecords = inputFile.GetNumOfRecords();
int recCounter = 0, globalRecCounter = 0;
// Allocate memory
int areaSize = MINIBASE_PAGESIZE * _numBufPages;
char *area = (char *)malloc(areaSize);
char *areaPtr = area;
RecordID rid; char *recPtr = (char *)malloc(_recLength); int recLen = _recLength;
int numRecForSort = std::min(areaSize/_recLength,numRecords); // number of rec in sorting area at once
// Open Scan
Scan *scan = inputFile.OpenScan(status);
if (status != OK) return ReturnFAIL("Opening scan in PassZero function failed.");
// Sort
passZeroRuns = 0;
if (areaSize >= _recLength) { // can fit at least one record
while (scan->GetNext(rid,recPtr,recLen) == OK) {
recCounter++; globalRecCounter++;
// add to memory
if (memcpy(areaPtr,recPtr,recLen) != areaPtr)
return ReturnFAIL("Reading records to memory in PassZero function failed.");
areaPtr += recLen;
areaSize -= recLen;
if (areaSize < _recLength || globalRecCounter == numRecords) { // can't fit another rec or all recs have been added
// sort
switch (_sortType) {
case attrInteger:
std::qsort(area,recCounter,_recLength,CompareInt);
break;
case attrString:
std::qsort(area,recCounter,_recLength,CompareString);
default:
break;
}
// write out
char *fileName = CreateTempFilename(_outFile,0,passZeroRuns);
passZeroRuns++;
HeapFile *tempFile = new HeapFile(fileName,status);
if (status != OK) return ReturnFAIL("Opening temp file in PassZero function failed.");
areaPtr = area;
while (recCounter > 0) { // insert
tempFile->InsertRecord(areaPtr,_recLength,rid);
recCounter--;
areaPtr += _recLength;
}
numTempFiles++;
areaSize = MINIBASE_PAGESIZE * _numBufPages;
areaPtr = area; // reset
delete fileName;
delete tempFile;
}
}
}
free(area);
free(recPtr);
delete scan;
return OK;
}