SubtractFile::SubtractFile(ContextSubtract *context)
: IntersectFile(context),
_tmpBlocksMgr(NULL),
_deleteTmpBlocks(false),
_dontReport(false)
{
_tmpBlocksMgr = new BlockMgr(upCast(_context)->getOverlapFractionA(), upCast(_context)->getReciprocalFraction());
}
bool Fisher::init(void)
{
if(!(upCast(_context)->getExcludeFile().empty())){
string ex = upCast(_context)->getExcludeFile();
_excludeFile = new BedFile(ex);
_excludeFile->loadBedFileIntoMergedMap();
_haveExclude = true;
}
return Jaccard::init();
}
bool IntersectFile::finalizeCalculations()
{
if (upCast(_context)->getSortedInput() && !upCast(_context)->hasGenomeFile())
{
if (_context->getNameCheckDisabled())
_sweep->closeOut(false);
else
_sweep->closeOut(true);
}
return true;
}
bool SampleFile::strandComplies(const Record * record) {
if (!upCast(_context)->getSameStrand()) {
return true;
}
if (upCast(_context)->getForwardOnly() && record->getStrandVal() == Record::FORWARD) {
return true;
}
if (upCast(_context)->getReverseOnly() && record->getStrandVal() == Record::REVERSE) {
return true;
}
return false;
}
bool IntersectFile::init() {
_queryFRM = upCast(_context)->getFile(upCast(_context)->getQueryFileIdx());
if (upCast(_context)->getSortedInput()) {
makeSweep();
return _sweep->init();
} else {
_binTree = new BinTree( upCast(_context));
_binTree->loadDB();
}
return true;
}
bool GroupBy::init()
{
Tokenizer groupColsTokens;
groupColsTokens.tokenize(upCast(_context)->getGroupCols(), ',');
int numElems = groupColsTokens.getNumValidElems();
for (int i=0; i < numElems; i++) {
//if the item is a range, such as 3-5,
//must split that as well.
const QuickString &elem = groupColsTokens.getElem(i);
if (strchr(elem.c_str(), '-')) {
Tokenizer rangeElems;
rangeElems.tokenize(elem, '-');
int startNum = str2chrPos(rangeElems.getElem(0));
int endNum = str2chrPos(rangeElems.getElem(1));
for (int i=startNum; i <= endNum; i++) {
_groupCols.push_back(i);
}
} else {
_groupCols.push_back(str2chrPos(elem));
}
}
_queryFRM = _context->getFile(0);
_prevFields.resize(_groupCols.size());
_prevRecord = getNextRecord();
return true;
}
unsigned long Fisher::getTotalIntersection(RecordKeyVector &recList)
{
unsigned long intersection = 0;
Record *key = recList.getKey();
CHRPOS keyStart = key->getStartPos();
CHRPOS keyEnd = key->getEndPos();
_overlapCounts += recList.size();
// note that we truncate to a max size of 2.1GB
_qsizes.push_back((int)(keyEnd - keyStart));
int hitIdx = 0;
for (RecordKeyVector::iterator_type iter = recList.begin(); iter != recList.end(); iter = recList.next()) {
CHRPOS maxStart = max((*iter)->getStartPos(), keyStart);
CHRPOS minEnd = min((*iter)->getEndPos(), keyEnd);
_qsizes.push_back((int)(minEnd - maxStart));
if (_context->getObeySplits()) {
intersection += upCast(_context)->getSplitBlockInfo()->getOverlapBases(hitIdx);
hitIdx++;
} else {
intersection += (unsigned long)(minEnd - maxStart);
}
}
_numIntersections += (int)recList.size();
return intersection;
}
void CoverageFile::giveFinalReport(RecordOutputMgr *outputMgr) {
//only give report for histogram option
if (upCast(_context)->getCoverageType() != ContextCoverage::HIST) {
return;
}
for (depthMapType::iterator iter = _finalDepthMap.begin(); iter != _finalDepthMap.end(); iter++) {
size_t depth = iter->first;
size_t basesAtDepth = iter->second;
float depthPct = (float)basesAtDepth / (float)_totalQueryLen;
_finalOutput = "all\t";
_finalOutput.append(depth);
_finalOutput.append("\t");
_finalOutput.append(basesAtDepth);
_finalOutput.append("\t");
_finalOutput.append(_totalQueryLen);
_finalOutput.append("\t");
format(depthPct);
outputMgr->printRecord(NULL, _finalOutput);
}
}
void CoverageFile::processHits(RecordOutputMgr *outputMgr, RecordKeyVector &hits) {
makeDepthCount(hits);
_finalOutput.clear();
switch(upCast(_context)->getCoverageType()) {
case ContextCoverage::COUNT:
doCounts(outputMgr, hits);
break;
case ContextCoverage::PER_BASE:
doPerBase(outputMgr, hits);
break;
case ContextCoverage::MEAN:
doMean(outputMgr, hits);
break;
case ContextCoverage::HIST:
doHist(outputMgr, hits);
break;
case ContextCoverage::DEFAULT:
default:
doDefault(outputMgr, hits);
break;
}
}
IntersectFile::IntersectFile(ContextIntersect *context)
: ToolBase(upCast(context)),
_sweep(NULL),
_binTree(NULL),
_queryFRM(NULL)
{
}
void IntersectFile::checkSplits(RecordKeyVector &hitSet)
{
if (upCast(_context)->getObeySplits()) {
RecordKeyVector keySet(hitSet.getKey());
RecordKeyVector resultSet(hitSet.getKey());
RecordKeyVector overlapSet(hitSet.getKey());
upCast(_context)->getSplitBlockInfo()->findBlockedOverlaps(keySet, hitSet, resultSet, overlapSet);
// when using coverage, we need a list of the sub-intervals of coverage
// so that per-base depth can be properly calculated when obeying splits
if (_context->getProgram() == ContextBase::COVERAGE)
{
hitSet.swap(overlapSet);
}
else {
hitSet.swap(resultSet);
}
}
}
bool IntersectFile::findNext(RecordKeyVector &hits)
{
bool retVal = false;
if (upCast(_context)->getSortedInput()) {
retVal = nextSortedFind(hits);
}
else {
retVal = nextUnsortedFind(hits);
}
if (retVal) {
checkSplits(hits);
}
return retVal;
}
bool GroupBy::canGroup(const Record *newRecord) {
for (int i=0; i < (int)_groupCols.size(); i++) {
int fieldNum = _groupCols[i];
const QuickString &newField = newRecord->getField(fieldNum);
const QuickString &oldField = _prevFields[i];
if (upCast(_context)->ignoreCase()) {
if (oldField.stricmp(newField)) return false;
} else {
if (oldField != newField) return false;
}
}
return true;
}
void GroupBy::processHits(RecordOutputMgr *outputMgr, RecordKeyVector &hits)
{
const Record *rec = hits.getKey();
const QuickString &opVal = _context->getColumnOpsVal(hits);
if (upCast(_context)->printFullCols()) {
outputMgr->printRecord(rec, opVal);
} else {
QuickString outBuf;
for (int i=0; i < (int)_groupCols.size(); i++) {
outBuf.append(rec->getField(_groupCols[i]));
outBuf.append('\t');
}
outBuf.append(opVal);
outputMgr->printRecord(NULL, outBuf);
}
}
unsigned long Jaccard::getTotalIntersection(RecordKeyVector &hits)
{
unsigned long intersection = 0;
Record *key = hits.getKey();
CHRPOS keyStart = key->getStartPos();
CHRPOS keyEnd = key->getEndPos();
int hitIdx = 0;
for (RecordKeyVector::iterator_type iter = hits.begin(); iter != hits.end(); iter = hits.next()) {
Record *currRec = *iter;
CHRPOS maxStart = max(currRec->getStartPos(), keyStart);
CHRPOS minEnd = min(currRec->getEndPos(), keyEnd);
if (_context->getObeySplits()) {
intersection += upCast(_context)->getSplitBlockInfo()->getOverlapBases(hitIdx);
hitIdx++;
} else {
intersection += (unsigned long)(minEnd - maxStart);
}
}
_numIntersections += (int)hits.size();
return intersection;
}
SkOpSpanBase* SkOpSpanBase::active() {
SkOpSpanBase* result = fPrev ? fPrev->next() : upCast()->next()->prev();
SkASSERT(this == result || fDebugDeleted);
return result;
}
void IntersectFile::makeSweep() {
_sweep = new NewChromSweep(upCast(_context));
}
bool ComplementFile::init()
{
_frm = static_cast<FileRecordMergeMgr *>(upCast(_context)->getFile(0));
return true;
}
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;
}
Array<T>* createScalarNode(const dim4 &size, const T val)
{
JIT::ScalarNode *node = NULL;
node = new JIT::ScalarNode(upCast(val), isDouble<T>());
return createNodeArray<T>(size, reinterpret_cast<JIT::Node *>(node));
}
