1 // ***************************************************************************
2 // BamStandardIndex.cpp (c) 2010 Derek Barnett
3 // Marth Lab, Department of Biology, Boston College
4 // All rights reserved.
5 // ---------------------------------------------------------------------------
6 // Last modified: 24 June 2011 (DB)
7 // ---------------------------------------------------------------------------
8 // Provides index operations for the standardized BAM index format (".bai")
9 // ***************************************************************************
11 #include <api/BamAlignment.h>
12 #include <api/internal/BamReader_p.h>
13 #include <api/internal/BamStandardIndex_p.h>
14 using namespace BamTools;
15 using namespace BamTools::Internal;
24 // static BamStandardIndex constants
25 const int BamStandardIndex::MAX_BIN = 37450; // =(8^6-1)/7+1
26 const int BamStandardIndex::BAM_LIDX_SHIFT = 14;
27 const string BamStandardIndex::BAI_EXTENSION = ".bai";
28 const char* const BamStandardIndex::BAI_MAGIC = "BAI\1";
29 const int BamStandardIndex::SIZEOF_ALIGNMENTCHUNK = sizeof(uint64_t)*2;
30 const int BamStandardIndex::SIZEOF_BINCORE = sizeof(uint32_t) + sizeof(int32_t);
31 const int BamStandardIndex::SIZEOF_LINEAROFFSET = sizeof(uint64_t);
34 BamStandardIndex::BamStandardIndex(Internal::BamReaderPrivate* reader)
37 , m_cacheMode(BamIndex::LimitedIndexCaching)
41 m_isBigEndian = BamTools::SystemIsBigEndian();
45 BamStandardIndex::~BamStandardIndex(void) {
49 bool BamStandardIndex::AdjustRegion(const BamRegion& region, uint32_t& begin, uint32_t& end) {
51 // retrieve references from reader
52 const RefVector& references = m_reader->GetReferenceData();
54 // make sure left-bound position is valid
55 if ( region.LeftPosition > references.at(region.LeftRefID).RefLength )
59 begin = (unsigned int)region.LeftPosition;
61 // if right bound specified AND left&right bounds are on same reference
62 // OK to use right bound position as region 'end'
63 if ( region.isRightBoundSpecified() && ( region.LeftRefID == region.RightRefID ) )
64 end = (unsigned int)region.RightPosition;
66 // otherwise, set region 'end' to last reference base
67 else end = (unsigned int)references.at(region.LeftRefID).RefLength - 1;
73 void BamStandardIndex::CalculateCandidateBins(const uint32_t& begin,
75 set<uint16_t>& candidateBins)
77 // initialize list, bin '0' is always a valid bin
78 candidateBins.insert(0);
80 // get rest of bins that contain this region
82 for (k = 1 + (begin>>26); k <= 1 + (end>>26); ++k) { candidateBins.insert(k); }
83 for (k = 9 + (begin>>23); k <= 9 + (end>>23); ++k) { candidateBins.insert(k); }
84 for (k = 73 + (begin>>20); k <= 73 + (end>>20); ++k) { candidateBins.insert(k); }
85 for (k = 585 + (begin>>17); k <= 585 + (end>>17); ++k) { candidateBins.insert(k); }
86 for (k = 4681 + (begin>>14); k <= 4681 + (end>>14); ++k) { candidateBins.insert(k); }
89 bool BamStandardIndex::CalculateCandidateOffsets(const BaiReferenceSummary& refSummary,
90 const uint64_t& minOffset,
91 set<uint16_t>& candidateBins,
92 vector<int64_t>& offsets)
94 // attempt seek to first bin
95 if ( !Seek(refSummary.FirstBinFilePosition, SEEK_SET) )
98 // iterate over reference bins
100 int32_t numAlignmentChunks;
101 set<uint16_t>::iterator candidateBinIter;
102 for ( int i = 0; i < refSummary.NumBins; ++i ) {
104 // read bin contents (if successful, alignment chunks are now in m_buffer)
105 if ( !ReadBinIntoBuffer(binId, numAlignmentChunks) )
108 // see if bin is a 'candidate bin'
109 candidateBinIter = candidateBins.find(binId);
111 // if not, move on to next bin
112 if ( candidateBinIter == candidateBins.end() )
115 // otherwise, check bin's contents against for overlap
118 unsigned int offset = 0;
122 // iterate over alignment chunks
123 for (int j = 0; j < numAlignmentChunks; ++j ) {
125 // read chunk start & stop from buffer
126 memcpy((char*)&chunkStart, m_buffer+offset, sizeof(uint64_t));
127 offset += sizeof(uint64_t);
128 memcpy((char*)&chunkStop, m_buffer+offset, sizeof(uint64_t));
129 offset += sizeof(uint64_t);
131 // swap endian-ness if necessary
132 if ( m_isBigEndian ) {
133 SwapEndian_64(chunkStart);
134 SwapEndian_64(chunkStop);
137 // store alignment chunk's start offset
138 // if its stop offset is larger than our 'minOffset'
139 if ( chunkStop >= minOffset )
140 offsets.push_back(chunkStart);
143 // 'pop' bin ID from candidate bins set
144 candidateBins.erase(candidateBinIter);
146 // quit if no more candidates
147 if ( candidateBins.empty() )
156 uint64_t BamStandardIndex::CalculateMinOffset(const BaiReferenceSummary& refSummary,
157 const uint32_t& begin)
159 // if no linear offsets exist, return 0
160 if ( refSummary.NumLinearOffsets == 0 )
163 // if 'begin' starts beyond last linear offset, use the last linear offset as minimum
164 // else use the offset corresponding to the requested start position
165 const int shiftedBegin = begin>>BamStandardIndex::BAM_LIDX_SHIFT;
166 if ( shiftedBegin >= refSummary.NumLinearOffsets )
167 return LookupLinearOffset( refSummary, refSummary.NumLinearOffsets-1 );
169 return LookupLinearOffset( refSummary, shiftedBegin );
172 void BamStandardIndex::CheckBufferSize(char*& buffer,
173 unsigned int& bufferLength,
174 const unsigned int& requestedBytes)
177 if ( requestedBytes > bufferLength ) {
178 bufferLength = requestedBytes + 10;
180 buffer = new char[bufferLength];
182 } catch ( std::bad_alloc ) {
183 cerr << "BamStandardIndex ERROR: out of memory when allocating "
184 << requestedBytes << " byes" << endl;
189 void BamStandardIndex::CheckBufferSize(unsigned char*& buffer,
190 unsigned int& bufferLength,
191 const unsigned int& requestedBytes)
194 if ( requestedBytes > bufferLength ) {
195 bufferLength = requestedBytes + 10;
197 buffer = new unsigned char[bufferLength];
199 } catch ( std::bad_alloc ) {
200 cerr << "BamStandardIndex ERROR: out of memory when allocating "
201 << requestedBytes << " byes" << endl;
206 bool BamStandardIndex::CheckMagicNumber(void) {
208 // check 'magic number' to see if file is BAI index
210 size_t elementsRead = fread(magic, sizeof(char), 4, m_indexStream);
211 if ( elementsRead != 4 ) {
212 cerr << "BamStandardIndex ERROR: could not read format 'magic number'" << endl;
216 // compare to expected value
217 if ( strncmp(magic, BamStandardIndex::BAI_MAGIC, 4) != 0 ) {
218 cerr << "BamStandardIndex ERROR: invalid format" << endl;
226 void BamStandardIndex::ClearReferenceEntry(BaiReferenceEntry& refEntry) {
228 refEntry.Bins.clear();
229 refEntry.LinearOffsets.clear();
232 void BamStandardIndex::CloseFile(void) {
236 fclose(m_indexStream);
238 // clear index file summary data
239 m_indexFileSummary.clear();
241 // clean up I/O buffer
247 // builds index from associated BAM file & writes out to index file
248 bool BamStandardIndex::Create(void) {
250 // return false if BamReader is invalid or not open
251 if ( m_reader == 0 || !m_reader->IsOpen() ) {
252 cerr << "BamStandardIndex ERROR: BamReader is not open"
253 << ", aborting index creation" << endl;
258 if ( !m_reader->Rewind() ) {
259 cerr << "BamStandardIndex ERROR: could not rewind BamReader to create index"
260 << ", aborting index creation" << endl;
264 // open new index file (read & write)
265 string indexFilename = m_reader->Filename() + Extension();
266 if ( !OpenFile(indexFilename, "w+b") ) {
267 cerr << "BamStandardIndex ERROR: could not open ouput index file: " << indexFilename
268 << ", aborting index creation" << endl;
272 // initialize BaiFileSummary with number of references
273 const int& numReferences = m_reader->GetReferenceCount();
274 ReserveForSummary(numReferences);
276 // initialize output file
277 bool createdOk = true;
278 createdOk &= WriteHeader();
280 // set up bin, ID, offset, & coordinate markers
281 const uint32_t defaultValue = 0xffffffffu;
282 uint32_t currentBin = defaultValue;
283 uint32_t lastBin = defaultValue;
284 int32_t currentRefID = defaultValue;
285 int32_t lastRefID = defaultValue;
286 uint64_t currentOffset = (uint64_t)m_reader->Tell();
287 uint64_t lastOffset = currentOffset;
288 int32_t lastPosition = defaultValue;
290 // iterate through alignments in BAM file
292 BaiReferenceEntry refEntry;
293 while ( m_reader->LoadNextAlignment(al) ) {
295 // changed to new reference
296 if ( lastRefID != al.RefID ) {
298 // if not first reference, save previous reference data
299 if ( lastRefID != (int32_t)defaultValue ) {
301 SaveAlignmentChunkToBin(refEntry.Bins, currentBin, currentOffset, lastOffset);
302 createdOk &= WriteReferenceEntry(refEntry);
303 ClearReferenceEntry(refEntry);
305 // write any empty references between (but *NOT* including) lastRefID & al.RefID
306 for ( int i = lastRefID+1; i < al.RefID; ++i ) {
307 BaiReferenceEntry emptyEntry(i);
308 createdOk &= WriteReferenceEntry(emptyEntry);
311 // update bin markers
312 currentOffset = lastOffset;
315 currentRefID = al.RefID;
319 // write any empty references up to (but *NOT* including) al.RefID
321 for ( int i = 0; i < al.RefID; ++i ) {
322 BaiReferenceEntry emptyEntry(i);
323 createdOk &= WriteReferenceEntry(emptyEntry);
327 // update reference markers
328 refEntry.ID = al.RefID;
329 lastRefID = al.RefID;
330 lastBin = defaultValue;
333 // if lastPosition greater than current alignment position - file not sorted properly
334 else if ( lastPosition > al.Position ) {
335 cerr << "BamStandardIndex ERROR: BAM file is not properly sorted by coordinate"
336 << ", aborting index creation"
338 << "At alignment: " << al.Name
339 << " : previous position " << lastPosition
340 << " > this alignment position " << al.Position
341 << " on reference id: " << al.RefID << endl;
345 // if alignment's ref ID is valid & its bin is not a 'leaf'
346 if ( (al.RefID >= 0) && (al.Bin < 4681) )
347 SaveLinearOffsetEntry(refEntry.LinearOffsets, al.Position, al.GetEndPosition(), lastOffset);
349 // changed to new BAI bin
350 if ( al.Bin != lastBin ) {
352 // if not first bin on reference, save previous bin data
353 if ( currentBin != defaultValue )
354 SaveAlignmentChunkToBin(refEntry.Bins, currentBin, currentOffset, lastOffset);
357 currentOffset = lastOffset;
360 currentRefID = al.RefID;
362 // if invalid RefID, break out
363 if ( currentRefID < 0 )
367 // make sure that current file pointer is beyond lastOffset
368 if ( m_reader->Tell() <= (int64_t)lastOffset ) {
369 cerr << "BamStandardIndex ERROR: calculating offsets failed"
370 << ", aborting index creation" << endl;
374 // update lastOffset & lastPosition
375 lastOffset = m_reader->Tell();
376 lastPosition = al.Position;
379 // after finishing alignments, if any data was read, check:
380 if ( currentRefID >= 0 ) {
382 // store last alignment chunk to its bin, then write last reference entry with data
383 SaveAlignmentChunkToBin(refEntry.Bins, currentBin, currentOffset, lastOffset);
384 createdOk &= WriteReferenceEntry(refEntry);
386 // then write any empty references remaining at end of file
387 for ( int i = currentRefID+1; i < numReferences; ++i ) {
388 BaiReferenceEntry emptyEntry(i);
389 createdOk &= WriteReferenceEntry(emptyEntry);
393 // rewind reader now that we're done building
394 createdOk &= m_reader->Rewind();
400 // returns format's file extension
401 const string BamStandardIndex::Extension(void) {
402 return BamStandardIndex::BAI_EXTENSION;
405 bool BamStandardIndex::GetOffset(const BamRegion& region, int64_t& offset, bool* hasAlignmentsInRegion) {
407 // cannot calculate offsets if unknown/invalid reference ID requested
408 if ( region.LeftRefID < 0 || region.LeftRefID >= (int)m_indexFileSummary.size() )
411 // retrieve index summary for left bound reference
412 const BaiReferenceSummary& refSummary = m_indexFileSummary.at(region.LeftRefID);
414 // set up region boundaries based on actual BamReader data
417 if ( !AdjustRegion(region, begin, end) ) {
418 cerr << "BamStandardIndex ERROR: cannot calculate offsets on invalid region" << endl;
422 // retrieve all candidate bin IDs for region
423 set<uint16_t> candidateBins;
424 CalculateCandidateBins(begin, end, candidateBins);
426 // use reference's linear offsets to calculate the minimum offset
427 // that must be considered to find overlap
428 const uint64_t& minOffset = CalculateMinOffset(refSummary, begin);
430 // attempt to use reference summary, minOffset, & candidateBins to calculate offsets
431 // no data should not be error
432 vector<int64_t> offsets;
433 if ( !CalculateCandidateOffsets(refSummary, minOffset, candidateBins, offsets) ) {
434 cerr << "BamStandardIndex ERROR: could not calculate candidate offsets for requested region" << endl;
438 // ensure that offsets are sorted before processing
439 sort( offsets.begin(), offsets.end() );
441 // binary search for an overlapping block (may not be first one though)
443 typedef vector<int64_t>::const_iterator OffsetConstIterator;
444 OffsetConstIterator offsetFirst = offsets.begin();
445 OffsetConstIterator offsetIter = offsetFirst;
446 OffsetConstIterator offsetLast = offsets.end();
447 iterator_traits<OffsetConstIterator>::difference_type count = distance(offsetFirst, offsetLast);
448 iterator_traits<OffsetConstIterator>::difference_type step;
449 while ( count > 0 ) {
450 offsetIter = offsetFirst;
452 advance(offsetIter, step);
454 // attempt seek to candidate offset
455 const int64_t& candidateOffset = (*offsetIter);
456 if ( !m_reader->Seek(candidateOffset) ) {
457 cerr << "BamStandardIndex ERROR: could not jump"
458 << ", there was a problem seeking in BAM file" << endl;
462 // load first available alignment, setting flag to true if data exists
463 *hasAlignmentsInRegion = m_reader->LoadNextAlignment(al);
465 // check alignment against region
466 if ( al.GetEndPosition() < region.LeftPosition ) {
467 offsetFirst = ++offsetIter;
472 // seek back to the offset before the 'current offset' (to cover overlaps)
473 if ( offsetIter != offsets.begin() )
475 offset = (*offsetIter);
481 // returns whether reference has alignments or no
482 bool BamStandardIndex::HasAlignments(const int& referenceID) const {
483 if ( referenceID < 0 || referenceID >= (int)m_indexFileSummary.size() )
485 const BaiReferenceSummary& refSummary = m_indexFileSummary.at(referenceID);
486 return ( refSummary.NumBins > 0 );
489 bool BamStandardIndex::IsFileOpen(void) const {
490 return ( m_indexStream != 0 );
493 // attempts to use index data to jump to @region, returns success/fail
494 // a "successful" jump indicates no error, but not whether this region has data
495 // * thus, the method sets a flag to indicate whether there are alignments
496 // available after the jump position
497 bool BamStandardIndex::Jump(const BamRegion& region, bool* hasAlignmentsInRegion) {
500 *hasAlignmentsInRegion = false;
502 // skip if reader is not valid or is not open
503 if ( m_reader == 0 || !m_reader->IsOpen() )
506 // calculate nearest offset to jump to
508 if ( !GetOffset(region, offset, hasAlignmentsInRegion) ) {
509 cerr << "BamStandardIndex ERROR: could not jump"
510 << ", unable to calculate offset for specified region" << endl;
514 // if region has alignments, return success/fail of seeking there
515 if ( *hasAlignmentsInRegion )
516 return m_reader->Seek(offset);
518 // otherwise, simply return true (but hasAlignmentsInRegion flag has been set to false)
519 // (this is OK, BamReader will check this flag before trying to load data)
523 // loads existing data from file into memory
524 bool BamStandardIndex::Load(const std::string& filename) {
526 // attempt open index file (read-only)
527 if ( !OpenFile(filename, "rb") ) {
528 cerr << "BamStandardIndex ERROR: could not open input index file: " << filename
529 << ", aborting index load" << endl;
533 // if invalid format 'magic number', close & return failure
534 if ( !CheckMagicNumber() ) {
535 cerr << "BamStandardIndex ERROR: unexpected format for index file: " << filename
536 << ", aborting index load" << endl;
541 // attempt to load index file summary, return success/failure
542 if ( !SummarizeIndexFile() ) {
543 cerr << "BamStandardIndex ERROR: could not generate a summary of index file " << filename
544 << ", aborting index load" << endl;
549 // if we get here, index summary is loaded OK
553 uint64_t BamStandardIndex::LookupLinearOffset(const BaiReferenceSummary& refSummary, const int& index) {
555 // attempt seek to proper index file position
556 const int64_t linearOffsetFilePosition = (int64_t)refSummary.FirstLinearOffsetFilePosition +
557 index*BamStandardIndex::SIZEOF_LINEAROFFSET;
558 if ( !Seek(linearOffsetFilePosition, SEEK_SET) )
561 // read linear offset from BAI file
562 uint64_t linearOffset(0);
563 if ( !ReadLinearOffset(linearOffset) )
568 void BamStandardIndex::MergeAlignmentChunks(BaiAlignmentChunkVector& chunks) {
570 // skip if chunks are empty, nothing to merge
571 if ( chunks.empty() )
574 // set up merged alignment chunk container
575 BaiAlignmentChunkVector mergedChunks;
576 mergedChunks.push_back( chunks[0] );
578 // iterate over chunks
580 BaiAlignmentChunkVector::iterator chunkIter = chunks.begin();
581 BaiAlignmentChunkVector::iterator chunkEnd = chunks.end();
582 for ( ++chunkIter; chunkIter != chunkEnd; ++chunkIter) {
584 // get 'currentMergeChunk' based on numeric index
585 BaiAlignmentChunk& currentMergeChunk = mergedChunks[i];
587 // get sourceChunk based on source vector iterator
588 BaiAlignmentChunk& sourceChunk = (*chunkIter);
590 // if currentMergeChunk ends where sourceChunk starts, then merge the two
591 if ( currentMergeChunk.Stop>>16 == sourceChunk.Start>>16 )
592 currentMergeChunk.Stop = sourceChunk.Stop;
596 // append sourceChunk after currentMergeChunk
597 mergedChunks.push_back(sourceChunk);
599 // update i, so the next iteration will consider the
600 // recently-appended sourceChunk as new mergeChunk candidate
605 // saved newly-merged chunks into (parameter) chunks
606 chunks = mergedChunks;
609 bool BamStandardIndex::OpenFile(const std::string& filename, const char* mode) {
611 // make sure any previous index file is closed
614 // attempt to open file
615 m_indexStream = fopen(filename.c_str(), mode);
619 bool BamStandardIndex::ReadBinID(uint32_t& binId) {
620 size_t elementsRead = 0;
621 elementsRead += fread(&binId, sizeof(binId), 1, m_indexStream);
622 if ( m_isBigEndian ) SwapEndian_32(binId);
623 return ( elementsRead == 1 );
626 bool BamStandardIndex::ReadBinIntoBuffer(uint32_t& binId, int32_t& numAlignmentChunks) {
631 readOk &= ReadBinID(binId);
632 readOk &= ReadNumAlignmentChunks(numAlignmentChunks);
635 const unsigned int bytesRequested = numAlignmentChunks*BamStandardIndex::SIZEOF_ALIGNMENTCHUNK;
636 readOk &= ReadIntoBuffer(bytesRequested);
638 // return success/failure
642 bool BamStandardIndex::ReadIntoBuffer(const unsigned int& bytesRequested) {
644 // ensure that our buffer is big enough for request
645 BamStandardIndex::CheckBufferSize(m_buffer, m_bufferLength, bytesRequested);
647 // read from BAI file stream
648 size_t bytesRead = fread( m_buffer, sizeof(char), bytesRequested, m_indexStream );
649 return ( bytesRead == (size_t)bytesRequested );
652 bool BamStandardIndex::ReadLinearOffset(uint64_t& linearOffset) {
653 size_t elementsRead = 0;
654 elementsRead += fread(&linearOffset, sizeof(linearOffset), 1, m_indexStream);
655 if ( m_isBigEndian ) SwapEndian_64(linearOffset);
656 return ( elementsRead == 1 );
659 bool BamStandardIndex::ReadNumAlignmentChunks(int& numAlignmentChunks) {
660 size_t elementsRead = 0;
661 elementsRead += fread(&numAlignmentChunks, sizeof(numAlignmentChunks), 1, m_indexStream);
662 if ( m_isBigEndian ) SwapEndian_32(numAlignmentChunks);
663 return ( elementsRead == 1 );
666 bool BamStandardIndex::ReadNumBins(int& numBins) {
667 size_t elementsRead = 0;
668 elementsRead += fread(&numBins, sizeof(numBins), 1, m_indexStream);
669 if ( m_isBigEndian ) SwapEndian_32(numBins);
670 return ( elementsRead == 1 );
673 bool BamStandardIndex::ReadNumLinearOffsets(int& numLinearOffsets) {
674 size_t elementsRead = 0;
675 elementsRead += fread(&numLinearOffsets, sizeof(numLinearOffsets), 1, m_indexStream);
676 if ( m_isBigEndian ) SwapEndian_32(numLinearOffsets);
677 return ( elementsRead == 1 );
680 bool BamStandardIndex::ReadNumReferences(int& numReferences) {
681 size_t elementsRead = 0;
682 elementsRead += fread(&numReferences, sizeof(numReferences), 1, m_indexStream);
683 if ( m_isBigEndian ) SwapEndian_32(numReferences);
684 return ( elementsRead == 1 );
687 void BamStandardIndex::ReserveForSummary(const int& numReferences) {
688 m_indexFileSummary.clear();
689 m_indexFileSummary.assign( numReferences, BaiReferenceSummary() );
692 void BamStandardIndex::SaveAlignmentChunkToBin(BaiBinMap& binMap,
693 const uint32_t& currentBin,
694 const uint64_t& currentOffset,
695 const uint64_t& lastOffset)
697 // create new alignment chunk
698 BaiAlignmentChunk newChunk(currentOffset, lastOffset);
702 // if no entry exists yet for this bin, create one and store alignment chunk
703 BaiBinMap::iterator binIter = binMap.find(currentBin);
704 if ( binIter == binMap.end() ) {
705 BaiAlignmentChunkVector newChunks;
706 newChunks.push_back(newChunk);
707 binMap.insert( pair<uint32_t, BaiAlignmentChunkVector>(currentBin, newChunks));
710 // otherwise, just append alignment chunk
712 BaiAlignmentChunkVector& binChunks = (*binIter).second;
713 binChunks.push_back( newChunk );
717 void BamStandardIndex::SaveBinsSummary(const int& refId, const int& numBins) {
718 BaiReferenceSummary& refSummary = m_indexFileSummary.at(refId);
719 refSummary.NumBins = numBins;
720 refSummary.FirstBinFilePosition = Tell();
723 void BamStandardIndex::SaveLinearOffsetEntry(BaiLinearOffsetVector& offsets,
724 const int& alignmentStartPosition,
725 const int& alignmentStopPosition,
726 const uint64_t& lastOffset)
728 // get converted offsets
729 const int beginOffset = alignmentStartPosition >> BamStandardIndex::BAM_LIDX_SHIFT;
730 const int endOffset = (alignmentStopPosition - 1) >> BamStandardIndex::BAM_LIDX_SHIFT;
732 // resize vector if necessary
733 int oldSize = offsets.size();
734 int newSize = endOffset + 1;
735 if ( oldSize < newSize )
736 offsets.resize(newSize, 0);
739 for( int i = beginOffset + 1; i <= endOffset; ++i ) {
740 if ( offsets[i] == 0 )
741 offsets[i] = lastOffset;
745 void BamStandardIndex::SaveLinearOffsetsSummary(const int& refId, const int& numLinearOffsets) {
746 BaiReferenceSummary& refSummary = m_indexFileSummary.at(refId);
747 refSummary.NumLinearOffsets = numLinearOffsets;
748 refSummary.FirstLinearOffsetFilePosition = Tell();
751 // seek to position in index file stream
752 bool BamStandardIndex::Seek(const int64_t& position, const int& origin) {
753 return ( fseek64(m_indexStream, position, origin) == 0 );
756 // change the index caching behavior
757 void BamStandardIndex::SetCacheMode(const BamIndex::IndexCacheMode& mode) {
759 // do nothing else here ? cache mode will be ignored from now on, most likely
762 bool BamStandardIndex::SkipBins(const int& numBins) {
764 int32_t numAlignmentChunks;
765 bool skippedOk = true;
766 for (int i = 0; i < numBins; ++i)
767 skippedOk &= ReadBinIntoBuffer(binId, numAlignmentChunks); // results & buffer ignored
771 bool BamStandardIndex::SkipLinearOffsets(const int& numLinearOffsets) {
772 const unsigned int bytesRequested = numLinearOffsets*BamStandardIndex::SIZEOF_LINEAROFFSET;
773 return ReadIntoBuffer(bytesRequested);
776 void BamStandardIndex::SortLinearOffsets(BaiLinearOffsetVector& linearOffsets) {
777 sort( linearOffsets.begin(), linearOffsets.end() );
780 bool BamStandardIndex::SummarizeBins(BaiReferenceSummary& refSummary) {
782 // load number of bins
784 if ( !ReadNumBins(numBins) )
787 // store bins summary for this reference
788 refSummary.NumBins = numBins;
789 refSummary.FirstBinFilePosition = Tell();
791 // attempt skip reference bins, return success/failure
792 if ( !SkipBins(numBins) )
795 // if we get here, bin summarized OK
799 bool BamStandardIndex::SummarizeIndexFile(void) {
801 // load number of reference sequences
803 if ( !ReadNumReferences(numReferences) )
806 // initialize file summary data
807 ReserveForSummary(numReferences);
809 // iterate over reference entries
810 bool loadedOk = true;
811 BaiFileSummary::iterator summaryIter = m_indexFileSummary.begin();
812 BaiFileSummary::iterator summaryEnd = m_indexFileSummary.end();
813 for ( int i = 0; summaryIter != summaryEnd; ++summaryIter, ++i )
814 loadedOk &= SummarizeReference(*summaryIter);
820 bool BamStandardIndex::SummarizeLinearOffsets(BaiReferenceSummary& refSummary) {
822 // load number of linear offsets
823 int numLinearOffsets;
824 if ( !ReadNumLinearOffsets(numLinearOffsets) )
827 // store bin summary data for this reference
828 refSummary.NumLinearOffsets = numLinearOffsets;
829 refSummary.FirstLinearOffsetFilePosition = Tell();
831 // skip linear offsets in index file
832 if ( !SkipLinearOffsets(numLinearOffsets) )
835 // if get here, linear offsets summarized OK
839 bool BamStandardIndex::SummarizeReference(BaiReferenceSummary& refSummary) {
841 bool loadedOk = true;
842 loadedOk &= SummarizeBins(refSummary);
843 loadedOk &= SummarizeLinearOffsets(refSummary);
847 // return position of file pointer in index file stream
848 int64_t BamStandardIndex::Tell(void) const {
849 return ftell64(m_indexStream);
852 bool BamStandardIndex::WriteAlignmentChunk(const BaiAlignmentChunk& chunk) {
854 size_t elementsWritten = 0;
856 // localize alignment chunk offsets
857 uint64_t start = chunk.Start;
858 uint64_t stop = chunk.Stop;
860 // swap endian-ness if necessary
861 if ( m_isBigEndian ) {
862 SwapEndian_64(start);
866 // write to index file
867 elementsWritten += fwrite(&start, sizeof(start), 1, m_indexStream);
868 elementsWritten += fwrite(&stop, sizeof(stop), 1, m_indexStream);
870 // return success/failure of write
871 return ( elementsWritten == 2 );
874 bool BamStandardIndex::WriteAlignmentChunks(BaiAlignmentChunkVector& chunks) {
876 // make sure chunks are merged (simplified) before writing & saving summary
877 MergeAlignmentChunks(chunks);
879 size_t elementsWritten = 0;
882 int32_t chunkCount = chunks.size();
883 if ( m_isBigEndian ) SwapEndian_32(chunkCount);
884 elementsWritten += fwrite(&chunkCount, sizeof(chunkCount), 1, m_indexStream);
886 // iterate over chunks
887 bool chunksOk = true;
888 BaiAlignmentChunkVector::const_iterator chunkIter = chunks.begin();
889 BaiAlignmentChunkVector::const_iterator chunkEnd = chunks.end();
890 for ( ; chunkIter != chunkEnd; ++chunkIter )
891 chunksOk &= WriteAlignmentChunk( (*chunkIter) );
893 // return success/failure of write
894 return ( (elementsWritten == 1) && chunksOk );
897 bool BamStandardIndex::WriteBin(const uint32_t& binId, BaiAlignmentChunkVector& chunks) {
899 size_t elementsWritten = 0;
902 uint32_t binKey = binId;
903 if ( m_isBigEndian ) SwapEndian_32(binKey);
904 elementsWritten += fwrite(&binKey, sizeof(binKey), 1, m_indexStream);
906 // write bin's alignment chunks
907 bool chunksOk = WriteAlignmentChunks(chunks);
909 // return success/failure of write
910 return ( (elementsWritten == 1) && chunksOk );
913 bool BamStandardIndex::WriteBins(const int& refId, BaiBinMap& bins) {
915 size_t elementsWritten = 0;
917 // write number of bins
918 int32_t binCount = bins.size();
919 if ( m_isBigEndian ) SwapEndian_32(binCount);
920 elementsWritten += fwrite(&binCount, sizeof(binCount), 1, m_indexStream);
922 // save summary for reference's bins
923 SaveBinsSummary(refId, bins.size());
927 BaiBinMap::iterator binIter = bins.begin();
928 BaiBinMap::iterator binEnd = bins.end();
929 for ( ; binIter != binEnd; ++binIter )
930 binsOk &= WriteBin( (*binIter).first, (*binIter).second );
932 // return success/failure of write
933 return ( (elementsWritten == 1) && binsOk );
936 bool BamStandardIndex::WriteHeader(void) {
938 size_t elementsWritten = 0;
940 // write magic number
941 elementsWritten += fwrite(BamStandardIndex::BAI_MAGIC, sizeof(char), 4, m_indexStream);
943 // write number of reference sequences
944 int32_t numReferences = m_indexFileSummary.size();
945 if ( m_isBigEndian ) SwapEndian_32(numReferences);
946 elementsWritten += fwrite(&numReferences, sizeof(numReferences), 1, m_indexStream);
948 // return success/failure of write
949 return (elementsWritten == 5);
952 bool BamStandardIndex::WriteLinearOffsets(const int& refId, BaiLinearOffsetVector& linearOffsets) {
954 // make sure linear offsets are sorted before writing & saving summary
955 SortLinearOffsets(linearOffsets);
957 size_t elementsWritten = 0;
959 // write number of linear offsets
960 int32_t offsetCount = linearOffsets.size();
961 if ( m_isBigEndian ) SwapEndian_32(offsetCount);
962 elementsWritten += fwrite(&offsetCount, sizeof(offsetCount), 1, m_indexStream);
964 // save summary for reference's linear offsets
965 SaveLinearOffsetsSummary(refId, linearOffsets.size());
967 // iterate over linear offsets
968 BaiLinearOffsetVector::const_iterator offsetIter = linearOffsets.begin();
969 BaiLinearOffsetVector::const_iterator offsetEnd = linearOffsets.end();
970 for ( ; offsetIter != offsetEnd; ++offsetIter ) {
972 // write linear offset
973 uint64_t linearOffset = (*offsetIter);
974 if ( m_isBigEndian ) SwapEndian_64(linearOffset);
975 elementsWritten += fwrite(&linearOffset, sizeof(linearOffset), 1, m_indexStream);
978 // return success/failure of write
979 return ( elementsWritten == (size_t)(linearOffsets.size() + 1) );
982 bool BamStandardIndex::WriteReferenceEntry(BaiReferenceEntry& refEntry) {
984 refOk &= WriteBins(refEntry.ID, refEntry.Bins);
985 refOk &= WriteLinearOffsets(refEntry.ID, refEntry.LinearOffsets);