int BlockMgr::findBlockedOverlaps(RecordKeyVector &keyList, RecordKeyVector &hitList, RecordKeyVector &resultList)
{
bool deleteKeyBlocks = false;
if (keyList.empty()) {
//get all the blocks for the query record, put them in it's list.
getBlocks(keyList, deleteKeyBlocks);
}
_overlapBases.clear();
int keyBlocksSumLength = getTotalBlockLength(keyList);
//Loop through every database record the query intersected with
for (RecordKeyVector::const_iterator_type hitListIter = hitList.begin(); hitListIter != hitList.end(); hitListIter = hitList.next()) {
RecordKeyVector hitBlocks(*hitListIter);
bool deleteHitBlocks = false;
getBlocks(hitBlocks, deleteHitBlocks); //get all blocks for the hit record.
int hitBlockSumLength = getTotalBlockLength(hitBlocks); //get total length of the bocks for the hitRecord.
int totalHitOverlap = 0;
bool hitHasOverlap = false;
//loop through every block of the database record.
for (RecordKeyVector::const_iterator_type hitBlockIter = hitBlocks.begin(); hitBlockIter != hitBlocks.end(); hitBlockIter = hitBlocks.next()) {
//loop through every block of the query record.
for (RecordKeyVector::const_iterator_type keyListIter = keyList.begin(); keyListIter != keyList.end(); keyListIter = keyList.next()) {
const Record *keyBlock = *keyListIter;
const Record *hitBlock = *hitBlockIter;
int maxStart = max(keyBlock->getStartPos(), hitBlock->getStartPos());
int minEnd = min(keyBlock->getEndPos(), hitBlock->getEndPos());
int overlap = minEnd - maxStart;
if (overlap > 0) {
hitHasOverlap = true;
totalHitOverlap += overlap;
}
}
}
if (hitHasOverlap) {
if ((float) totalHitOverlap / (float)keyBlocksSumLength >= _overlapFraction) {
if (_hasReciprocal &&
((float)totalHitOverlap / (float)hitBlockSumLength >= _overlapFraction)) {
_overlapBases.push_back(totalHitOverlap);
resultList.push_back(*hitListIter);
} else if (!_hasReciprocal) {
_overlapBases.push_back(totalHitOverlap);
resultList.push_back(*hitListIter);
}
}
}
if (deleteHitBlocks) {
deleteBlocks(hitBlocks);
}
}
if (deleteKeyBlocks) {
deleteBlocks(keyList);
}
resultList.setKey(keyList.getKey());
return (int)resultList.size();
}
void RecordOutputMgr::printClosest(RecordKeyVector &keyList, const vector<int> *dists) {
//The first time we print a record is when we print any header, because the header
//hasn't been read from the query file until after the first record has also been read.
checkForHeader();
const ContextClosest *context = static_cast<const ContextClosest *>(_context);
bool deleteBlocks = false;
const Record *keyRec = keyList.getKey();
RecordKeyVector blockList(keyRec);
if (keyRec->getType() == FileRecordTypeChecker::BAM_RECORD_TYPE) {
_bamBlockMgr->getBlocks(blockList, deleteBlocks);
_currBamBlockList = &blockList;
}
if (!keyList.empty()) {
int distCount = 0;
for (RecordKeyVector::const_iterator_type iter = keyList.begin(); iter != keyList.end(); iter = keyList.next()) {
const Record *hitRec = *iter;
printKey(keyRec, keyRec->getStartPosStr(), keyRec->getEndPosStr());
tab();
addDbFileId(hitRec->getFileIdx());
printKey(hitRec, hitRec->getStartPosStr(), hitRec->getEndPosStr());
if (dists != NULL) {
tab();
int dist = (*dists)[distCount];
//if not using sign distance, use absolute value instead.
dist = context->signDistance() ? dist : abs(dist);
_outBuf.append(dist);
distCount++;
}
newline();
if (needsFlush()) flush();
}
} else {
printKey(keyRec, keyRec->getStartPosStr(), keyRec->getEndPosStr());
tab();
// need to add a dummy file id if multiple DB files are used
if (_context->getNumInputFiles() > 2) {
_outBuf.append('.');
tab();
}
null(false, true);
if (context->reportDistance()) {
tab();
_outBuf.append(-1);
}
newline();
}
if (deleteBlocks) {
_bamBlockMgr->deleteBlocks(blockList);
_currBamBlockList = NULL;
}
return;
}
void RecordOutputMgr::printClosest(RecordKeyVector &keyList, const vector<int> *dists) {
const ContextClosest *context = static_cast<const ContextClosest *>(_context);
bool deleteBlocks = false;
RecordKeyVector blockList(keyList.getKey());
if (keyList.getKey()->getType() == FileRecordTypeChecker::BAM_RECORD_TYPE) {
_bamBlockMgr->getBlocks(blockList, deleteBlocks);
_currBamBlockList = &blockList;
}
if (!keyList.empty()) {
int distCount = 0;
for (RecordKeyVector::const_iterator_type iter = keyList.begin(); iter != keyList.end(); iter = keyList.next()) {
printKey(keyList.getKey());
tab();
addDbFileId((*iter)->getFileIdx());
(*iter)->print(_outBuf);
if (dists != NULL) {
tab();
_outBuf.append((*dists)[distCount]);
distCount++;
}
newline();
if (needsFlush()) flush();
}
} else {
printKey(keyList.getKey());
tab();
null(true, false);
if (context->reportDistance()) {
tab();
_outBuf.append(-1);
}
newline();
}
if (deleteBlocks) {
_bamBlockMgr->deleteBlocks(blockList);
_currBamBlockList = NULL;
}
return;
}
void RecordOutputMgr::printRecord(RecordKeyVector &keyList, RecordKeyVector *blockList)
{
if (needsFlush()) {
flush();
}
//The first time we print a record is when we print any header, because the header
//hasn't been read from the query file until after the first record has also been read.
checkForHeader();
const_cast<Record *>(keyList.getKey())->undoZeroLength();
_currBamBlockList = blockList;
if (_context->getProgram() == ContextBase::INTERSECT) {
if (_printable) {
if (keyList.empty()) {
if ((static_cast<ContextIntersect *>(_context))->getWriteAllOverlap()) {
// -wao the user wants to force the reporting of 0 overlap
if (printKeyAndTerminate(keyList)) {
_currBamBlockList = NULL;
return;
}
tab();
null(false, true);
tab();
_outBuf.append('0');
newline();
if (needsFlush()) flush();
}
else if ((static_cast<ContextIntersect *>(_context))->getLeftJoin()) {
if (printKeyAndTerminate(keyList)) {
_currBamBlockList = NULL;
return;
}
tab();
null(false, true);
newline();
if (needsFlush()) flush();
_currBamBlockList = NULL;
return;
}
} else {
if (printBamRecord(keyList, true) == BAM_AS_BAM) {
_currBamBlockList = NULL;
return;
}
int hitIdx = 0;
for (RecordKeyVector::const_iterator_type iter = keyList.begin(); iter != keyList.end(); iter = keyList.next()) {
reportOverlapDetail(keyList.getKey(), *iter, hitIdx);
hitIdx++;
}
}
} else { // not printable
reportOverlapSummary(keyList);
}
_currBamBlockList = NULL;
} else if (_context->getProgram() == ContextBase::SAMPLE) {
if (!printKeyAndTerminate(keyList)) {
newline();
}
_currBamBlockList = NULL;
return;
} else if (_context->getProgram() == ContextBase::MAP) {
printKeyAndTerminate(keyList);
_currBamBlockList = NULL;
return;
} else if (_context->getProgram() == ContextBase::MERGE) {
printKeyAndTerminate(keyList);
_currBamBlockList = NULL;
return;
}
}
void SubtractFile::subtractHits(RecordKeyVector &hits) {
if (hits.empty()) {
// no intersection, nothing to subtract.
// just copy key to hits as if it were a
// self-intersection. This is just for reporting
// purposes.
hits.push_back(hits.getKey());
return;
}
if (upCast(_context)->getRemoveAll() && upCast(_context)->getSubtractFraction() == 0.0) {
// hits aren't empty, meaning there is intersection,
// so we want to not report the hit.
_dontReport = true;
return;
}
//loop through hits. Track which bases in query were covered
Record *keyRec = hits.getKey();
int keyStart = keyRec->getStartPos();
int keyEnd = keyRec->getEndPos();
//this vector of bools will represent the bases of the query.
//for each base, true means uncovered, false means covered.
//they begin as all uncovered.
vector<bool> keyBases(keyEnd - keyStart, true);
//now loop through the hits, and cover corresponding query bases
//by setting them to false.
bool basesRemoved = false;
for (RecordKeyVector::iterator_type iter = hits.begin(); iter != hits.end(); iter = hits.next()) {
Record *hitRec = *iter;
int hitStart = hitRec->getStartPos();
int hitEnd = hitRec->getEndPos();
int startIdx = max(keyStart, hitStart) - keyStart;
int endIdx = min(keyEnd, hitEnd) - keyStart;
int keyLen = keyEnd - keyStart;
int coveredLen = endIdx - startIdx;
float coveragePct = (float)coveredLen / (float)keyLen;
//for each base in the hit, set the base in the query to false.
//this effectively "erases" the covered bits. Only do
if (upCast(_context)->getRemoveSum() || coveragePct >= upCast(_context)->getSubtractFraction()) {
std::fill(keyBases.begin() + startIdx, keyBases.begin() + endIdx, false);
basesRemoved = true;
}
}
if (!basesRemoved) {
//treat as if there were no intersection
hits.clearVector();
hits.push_back(hits.getKey());
return;
} else if (upCast(_context)->getRemoveAll()) {
_dontReport = true;
return;
}
// if the -N option is used ( removeSum), do not report if the percentage of
// uniquely covered bases exceeds the overlap fraction.
if (upCast(_context)->getRemoveSum()) {
//determine how many bases are left uncovered.
int numBasesUncovered = std::accumulate(keyBases.begin(), keyBases.end(), 0);
//determine percentage that are covered.
float pctCovered = 1.0 - (float)numBasesUncovered / (float)(keyEnd - keyStart);
if (pctCovered > upCast(_context)->getSubtractFraction()) {
_dontReport = true;
return;
} else {
hits.clearVector();
hits.push_back(hits.getKey());
}
return;
}
//now make "blocks" out of the query's remaining stretches of
//uncovered bases.
hits.clearVector();
for (int i = 0; i < (int)keyBases.size(); i++) {
if (keyBases[i] == true) {
int blockStart = keyStart + i;
while (keyBases[i] == true && i < (int)keyBases.size()) {
i++;
}
int blockEnd = min(keyStart + i, keyEnd);
hits.push_back(_tmpBlocksMgr->allocateAndAssignRecord(keyRec, blockStart, blockEnd));
}
}
_deleteTmpBlocks = true;
}
