1 // ***************************************************************************
\r
2 // BamReader.cpp (c) 2009 Derek Barnett, Michael Str�mberg
\r
3 // Marth Lab, Department of Biology, Boston College
\r
4 // All rights reserved.
\r
5 // ---------------------------------------------------------------------------
\r
6 // Last modified: 6 April 2010 (DB)
\r
7 // ---------------------------------------------------------------------------
\r
8 // Uses BGZF routines were adapted from the bgzf.c code developed at the Broad
\r
10 // ---------------------------------------------------------------------------
\r
11 // Provides the basic functionality for reading BAM files
\r
12 // ***************************************************************************
\r
15 #include <algorithm>
\r
20 // BamTools includes
\r
22 #include "BamReader.h"
\r
23 using namespace BamTools;
\r
24 using namespace std;
\r
26 namespace BamTools {
\r
27 struct BamAlignmentSupportData {
\r
29 uint32_t BlockLength;
\r
30 uint32_t NumCigarOperations;
\r
31 uint32_t QueryNameLength;
\r
32 uint32_t QuerySequenceLength;
\r
34 } // namespace BamTools
\r
36 struct BamReader::BamReaderPrivate {
\r
38 // -------------------------------
\r
40 // -------------------------------
\r
46 RefVector References;
\r
48 int64_t AlignmentsBeginOffset;
\r
50 string IndexFilename;
\r
54 // user-specified region values
\r
55 bool IsRegionSpecified;
\r
59 // BAM character constants
\r
60 const char* DNA_LOOKUP;
\r
61 const char* CIGAR_LOOKUP;
\r
63 // -------------------------------
\r
64 // constructor & destructor
\r
65 // -------------------------------
\r
66 BamReaderPrivate(void);
\r
67 ~BamReaderPrivate(void);
\r
69 // -------------------------------
\r
70 // "public" interface
\r
71 // -------------------------------
\r
75 bool Jump(int refID, int position = 0);
\r
76 void Open(const string& filename, const string& indexFilename = "");
\r
79 // access alignment data
\r
80 bool GetNextAlignment(BamAlignment& bAlignment);
\r
82 // access auxiliary data
\r
83 int GetReferenceID(const string& refName) const;
\r
86 bool CreateIndex(void);
\r
88 // -------------------------------
\r
90 // -------------------------------
\r
92 // *** reading alignments and auxiliary data *** //
\r
94 // calculate bins that overlap region ( left to reference end for now )
\r
95 int BinsFromRegion(int refID, int left, uint16_t[MAX_BIN]);
\r
96 // fills out character data for BamAlignment data
\r
97 bool BuildCharData(BamAlignment& bAlignment, const BamAlignmentSupportData& supportData);
\r
98 // calculates alignment end position based on starting position and provided CIGAR operations
\r
99 int CalculateAlignmentEnd(const int& position, const std::vector<CigarOp>& cigarData);
\r
100 // calculate file offset for first alignment chunk overlapping 'left'
\r
101 int64_t GetOffset(int refID, int left);
\r
102 // checks to see if alignment overlaps current region
\r
103 bool IsOverlap(BamAlignment& bAlignment);
\r
104 // retrieves header text from BAM file
\r
105 void LoadHeaderData(void);
\r
106 // retrieves BAM alignment under file pointer
\r
107 bool LoadNextAlignment(BamAlignment& bAlignment, BamAlignmentSupportData& supportData);
\r
108 // builds reference data structure from BAM file
\r
109 void LoadReferenceData(void);
\r
111 // *** index file handling *** //
\r
113 // calculates index for BAM file
\r
114 bool BuildIndex(void);
\r
115 // clear out inernal index data structure
\r
116 void ClearIndex(void);
\r
117 // saves BAM bin entry for index
\r
118 void InsertBinEntry(BamBinMap& binMap, const uint32_t& saveBin, const uint64_t& saveOffset, const uint64_t& lastOffset);
\r
119 // saves linear offset entry for index
\r
120 void InsertLinearOffset(LinearOffsetVector& offsets, const BamAlignment& bAlignment, const uint64_t& lastOffset);
\r
121 // loads index from BAM index file
\r
122 bool LoadIndex(void);
\r
123 // simplifies index by merging 'chunks'
\r
124 void MergeChunks(void);
\r
125 // saves index to BAM index file
\r
126 bool WriteIndex(void);
\r
129 // -----------------------------------------------------
\r
130 // BamReader implementation (wrapper around BRPrivate)
\r
131 // -----------------------------------------------------
\r
134 BamReader::BamReader(void) {
\r
135 d = new BamReaderPrivate;
\r
139 BamReader::~BamReader(void) {
\r
145 void BamReader::Close(void) { d->Close(); }
\r
146 bool BamReader::Jump(int refID, int position) { return d->Jump(refID, position); }
\r
147 void BamReader::Open(const string& filename, const string& indexFilename) { d->Open(filename, indexFilename); }
\r
148 bool BamReader::Rewind(void) { return d->Rewind(); }
\r
150 // access alignment data
\r
151 bool BamReader::GetNextAlignment(BamAlignment& bAlignment) { return d->GetNextAlignment(bAlignment); }
\r
153 // access auxiliary data
\r
154 const string BamReader::GetHeaderText(void) const { return d->HeaderText; }
\r
155 int BamReader::GetReferenceCount(void) const { return d->References.size(); }
\r
156 const RefVector BamReader::GetReferenceData(void) const { return d->References; }
\r
157 int BamReader::GetReferenceID(const string& refName) const { return d->GetReferenceID(refName); }
\r
159 // index operations
\r
160 bool BamReader::CreateIndex(void) { return d->CreateIndex(); }
\r
162 // -----------------------------------------------------
\r
163 // BamReaderPrivate implementation
\r
164 // -----------------------------------------------------
\r
167 BamReader::BamReaderPrivate::BamReaderPrivate(void)
\r
168 : IsIndexLoaded(false)
\r
169 , AlignmentsBeginOffset(0)
\r
170 , IsRegionSpecified(false)
\r
173 , DNA_LOOKUP("=ACMGRSVTWYHKDBN")
\r
174 , CIGAR_LOOKUP("MIDNSHP")
\r
176 IsBigEndian = SystemIsBigEndian();
\r
180 BamReader::BamReaderPrivate::~BamReaderPrivate(void) {
\r
184 // calculate bins that overlap region ( left to reference end for now )
\r
185 int BamReader::BamReaderPrivate::BinsFromRegion(int refID, int left, uint16_t list[MAX_BIN]) {
\r
187 // get region boundaries
\r
188 uint32_t begin = (unsigned int)left;
\r
189 uint32_t end = (unsigned int)References.at(refID).RefLength - 1;
\r
191 // initialize list, bin '0' always a valid bin
\r
195 // get rest of bins that contain this region
\r
197 for (k = 1 + (begin>>26); k <= 1 + (end>>26); ++k) { list[i++] = k; }
\r
198 for (k = 9 + (begin>>23); k <= 9 + (end>>23); ++k) { list[i++] = k; }
\r
199 for (k = 73 + (begin>>20); k <= 73 + (end>>20); ++k) { list[i++] = k; }
\r
200 for (k = 585 + (begin>>17); k <= 585 + (end>>17); ++k) { list[i++] = k; }
\r
201 for (k = 4681 + (begin>>14); k <= 4681 + (end>>14); ++k) { list[i++] = k; }
\r
203 // return number of bins stored
\r
207 bool BamReader::BamReaderPrivate::BuildCharData(BamAlignment& bAlignment, const BamAlignmentSupportData& supportData) {
\r
209 // calculate character lengths/offsets
\r
210 const unsigned int dataLength = supportData.BlockLength - BAM_CORE_SIZE;
\r
211 const unsigned int seqDataOffset = supportData.QueryNameLength + (supportData.NumCigarOperations * 4);
\r
212 const unsigned int qualDataOffset = seqDataOffset + (supportData.QuerySequenceLength+1)/2;
\r
213 const unsigned int tagDataOffset = qualDataOffset + supportData.QuerySequenceLength;
\r
214 const unsigned int tagDataLength = dataLength - tagDataOffset;
\r
216 // set up char buffers
\r
217 const char* allCharData = supportData.AllCharData.data();
\r
218 const char* seqData = ((const char*)allCharData) + seqDataOffset;
\r
219 const char* qualData = ((const char*)allCharData) + qualDataOffset;
\r
220 char* tagData = ((char*)allCharData) + tagDataOffset;
\r
222 // save query sequence
\r
223 bAlignment.QueryBases.clear();
\r
224 bAlignment.QueryBases.reserve(supportData.QuerySequenceLength);
\r
225 for (unsigned int i = 0; i < supportData.QuerySequenceLength; ++i) {
\r
226 char singleBase = DNA_LOOKUP[ ( ( seqData[(i/2)] >> (4*(1-(i%2)))) & 0xf ) ];
\r
227 bAlignment.QueryBases.append(1, singleBase);
\r
230 // save qualities, converting from numeric QV to 'FASTQ-style' ASCII character
\r
231 bAlignment.Qualities.clear();
\r
232 bAlignment.Qualities.reserve(supportData.QuerySequenceLength);
\r
233 for (unsigned int i = 0; i < supportData.QuerySequenceLength; ++i) {
\r
234 char singleQuality = (char)(qualData[i]+33);
\r
235 bAlignment.Qualities.append(1, singleQuality);
\r
238 // parse CIGAR to build 'AlignedBases'
\r
239 bAlignment.AlignedBases.clear();
\r
240 bAlignment.AlignedBases.reserve(supportData.QuerySequenceLength);
\r
243 vector<CigarOp>::const_iterator cigarIter = bAlignment.CigarData.begin();
\r
244 vector<CigarOp>::const_iterator cigarEnd = bAlignment.CigarData.end();
\r
245 for ( ; cigarIter != cigarEnd; ++cigarIter ) {
\r
247 const CigarOp& op = (*cigarIter);
\r
252 bAlignment.AlignedBases.append(bAlignment.QueryBases.substr(k, op.Length)); // for 'M', 'I' - write bases
\r
256 k += op.Length; // for 'S' - soft clip, skip over query bases
\r
260 bAlignment.AlignedBases.append(op.Length, '-'); // for 'D' - write gap character
\r
264 bAlignment.AlignedBases.append( op.Length, '*' ); // for 'P' - write padding character
\r
268 bAlignment.AlignedBases.append( op.Length, 'N' ); // for 'N' - write N's, skip bases in original query sequence
\r
273 break; // for 'H' - hard clip, do nothing to AlignedBases, move to next op
\r
276 printf("ERROR: Invalid Cigar op type\n"); // shouldn't get here
\r
281 // -----------------------
\r
282 // Added: 3-25-2010 DWB
\r
283 // Fixed: endian-correctness for tag data
\r
284 // -----------------------
\r
285 if ( IsBigEndian ) {
\r
287 while ( (unsigned int)i < tagDataLength ) {
\r
289 i += 2; // skip tag type (e.g. "RG", "NM", etc)
\r
290 uint8_t type = toupper(tagData[i]); // lower & upper case letters have same meaning
\r
291 ++i; // skip value type
\r
301 SwapEndian_16p(&tagData[i]);
\r
302 i+=2; // sizeof(uint16_t)
\r
307 SwapEndian_32p(&tagData[i]);
\r
308 i+=4; // sizeof(uint32_t)
\r
312 SwapEndian_64p(&tagData[i]);
\r
313 i+=8; // sizeof(uint64_t)
\r
318 while (tagData[i]) { ++i; }
\r
319 ++i; // increment one more for null terminator
\r
323 printf("ERROR: Invalid tag value type\n"); // shouldn't get here
\r
330 bAlignment.TagData.clear();
\r
331 bAlignment.TagData.resize(tagDataLength);
\r
332 memcpy((char*)bAlignment.TagData.data(), tagData, tagDataLength);
\r
338 // populates BAM index data structure from BAM file data
\r
339 bool BamReader::BamReaderPrivate::BuildIndex(void) {
\r
341 // check to be sure file is open
\r
342 if (!mBGZF.IsOpen) { return false; }
\r
344 // move file pointer to beginning of alignments
\r
347 // get reference count, reserve index space
\r
348 int numReferences = References.size();
\r
349 for ( int i = 0; i < numReferences; ++i ) {
\r
350 Index.push_back(ReferenceIndex());
\r
353 // sets default constant for bin, ID, offset, coordinate variables
\r
354 const uint32_t defaultValue = 0xffffffffu;
\r
357 uint32_t saveBin(defaultValue);
\r
358 uint32_t lastBin(defaultValue);
\r
360 // reference ID data
\r
361 int32_t saveRefID(defaultValue);
\r
362 int32_t lastRefID(defaultValue);
\r
365 uint64_t saveOffset = mBGZF.Tell();
\r
366 uint64_t lastOffset = saveOffset;
\r
369 int32_t lastCoordinate = defaultValue;
\r
371 BamAlignment bAlignment;
\r
372 while( GetNextAlignment(bAlignment) ) {
\r
374 // change of chromosome, save ID, reset bin
\r
375 if ( lastRefID != bAlignment.RefID ) {
\r
376 lastRefID = bAlignment.RefID;
\r
377 lastBin = defaultValue;
\r
380 // if lastCoordinate greater than BAM position - file not sorted properly
\r
381 else if ( lastCoordinate > bAlignment.Position ) {
\r
382 printf("BAM file not properly sorted:\n");
\r
383 printf("Alignment %s : %d > %d on reference (id = %d)", bAlignment.Name.c_str(), lastCoordinate, bAlignment.Position, bAlignment.RefID);
\r
387 // if valid reference && BAM bin spans some minimum cutoff (smaller bin ids span larger regions)
\r
388 if ( (bAlignment.RefID >= 0) && (bAlignment.Bin < 4681) ) {
\r
390 // save linear offset entry (matched to BAM entry refID)
\r
391 ReferenceIndex& refIndex = Index.at(bAlignment.RefID);
\r
392 LinearOffsetVector& offsets = refIndex.Offsets;
\r
393 InsertLinearOffset(offsets, bAlignment, lastOffset);
\r
396 // if current BamAlignment bin != lastBin, "then possibly write the binning index"
\r
397 if ( bAlignment.Bin != lastBin ) {
\r
399 // if not first time through
\r
400 if ( saveBin != defaultValue ) {
\r
402 // save Bam bin entry
\r
403 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
404 BamBinMap& binMap = refIndex.Bins;
\r
405 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
408 // update saveOffset
\r
409 saveOffset = lastOffset;
\r
411 // update bin values
\r
412 saveBin = bAlignment.Bin;
\r
413 lastBin = bAlignment.Bin;
\r
415 // update saveRefID
\r
416 saveRefID = bAlignment.RefID;
\r
418 // if invalid RefID, break out (why?)
\r
419 if ( saveRefID < 0 ) { break; }
\r
422 // make sure that current file pointer is beyond lastOffset
\r
423 if ( mBGZF.Tell() <= (int64_t)lastOffset ) {
\r
424 printf("Error in BGZF offsets.\n");
\r
428 // update lastOffset
\r
429 lastOffset = mBGZF.Tell();
\r
431 // update lastCoordinate
\r
432 lastCoordinate = bAlignment.Position;
\r
435 // save any leftover BAM data (as long as refID is valid)
\r
436 if ( saveRefID >= 0 ) {
\r
437 // save Bam bin entry
\r
438 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
439 BamBinMap& binMap = refIndex.Bins;
\r
440 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
443 // simplify index by merging chunks
\r
446 // iterate over references
\r
447 BamIndex::iterator indexIter = Index.begin();
\r
448 BamIndex::iterator indexEnd = Index.end();
\r
449 for ( int i = 0; indexIter != indexEnd; ++indexIter, ++i ) {
\r
451 // get reference index data
\r
452 ReferenceIndex& refIndex = (*indexIter);
\r
453 BamBinMap& binMap = refIndex.Bins;
\r
454 LinearOffsetVector& offsets = refIndex.Offsets;
\r
456 // store whether reference has alignments or no
\r
457 References[i].RefHasAlignments = ( binMap.size() > 0 );
\r
459 // sort linear offsets
\r
460 sort(offsets.begin(), offsets.end());
\r
464 // rewind file pointer to beginning of alignments, return success/fail
\r
468 // calculates alignment end position based on starting position and provided CIGAR operations
\r
469 int BamReader::BamReaderPrivate::CalculateAlignmentEnd(const int& position, const vector<CigarOp>& cigarData) {
\r
471 // initialize alignment end to starting position
\r
472 int alignEnd = position;
\r
474 // iterate over cigar operations
\r
475 vector<CigarOp>::const_iterator cigarIter = cigarData.begin();
\r
476 vector<CigarOp>::const_iterator cigarEnd = cigarData.end();
\r
477 for ( ; cigarIter != cigarEnd; ++cigarIter) {
\r
478 char cigarType = (*cigarIter).Type;
\r
479 if ( cigarType == 'M' || cigarType == 'D' || cigarType == 'N' ) {
\r
480 alignEnd += (*cigarIter).Length;
\r
487 // clear index data structure
\r
488 void BamReader::BamReaderPrivate::ClearIndex(void) {
\r
489 Index.clear(); // sufficient ??
\r
492 // closes the BAM file
\r
493 void BamReader::BamReaderPrivate::Close(void) {
\r
496 HeaderText.clear();
\r
497 IsRegionSpecified = false;
\r
500 // create BAM index from BAM file (keep structure in memory) and write to default index output file
\r
501 bool BamReader::BamReaderPrivate::CreateIndex(void) {
\r
506 // build (& save) index from BAM file
\r
508 ok &= BuildIndex();
\r
509 ok &= WriteIndex();
\r
511 // return success/fail
\r
515 // returns RefID for given RefName (returns References.size() if not found)
\r
516 int BamReader::BamReaderPrivate::GetReferenceID(const string& refName) const {
\r
518 // retrieve names from reference data
\r
519 vector<string> refNames;
\r
520 RefVector::const_iterator refIter = References.begin();
\r
521 RefVector::const_iterator refEnd = References.end();
\r
522 for ( ; refIter != refEnd; ++refIter) {
\r
523 refNames.push_back( (*refIter).RefName );
\r
526 // return 'index-of' refName ( if not found, returns refNames.size() )
\r
527 return distance(refNames.begin(), find(refNames.begin(), refNames.end(), refName));
\r
530 // get next alignment (from specified region, if given)
\r
531 bool BamReader::BamReaderPrivate::GetNextAlignment(BamAlignment& bAlignment) {
\r
533 BamAlignmentSupportData supportData;
\r
535 // if valid alignment available
\r
536 if ( LoadNextAlignment(bAlignment, supportData) ) {
\r
538 // if region not specified, return success
\r
539 if ( !IsRegionSpecified ) {
\r
540 bool ok = BuildCharData(bAlignment, supportData);
\r
544 // load next alignment until region overlap is found
\r
545 while ( !IsOverlap(bAlignment) ) {
\r
546 // if no valid alignment available (likely EOF) return failure
\r
547 if ( !LoadNextAlignment(bAlignment, supportData) ) { return false; }
\r
550 // return success (alignment found that overlaps region)
\r
551 bool ok = BuildCharData(bAlignment, supportData);
\r
555 // no valid alignment
\r
556 else { return false; }
\r
559 // calculate closest indexed file offset for region specified
\r
560 int64_t BamReader::BamReaderPrivate::GetOffset(int refID, int left) {
\r
562 // calculate which bins overlap this region
\r
563 uint16_t* bins = (uint16_t*)calloc(MAX_BIN, 2);
\r
564 int numBins = BinsFromRegion(refID, left, bins);
\r
566 // get bins for this reference
\r
567 const ReferenceIndex& refIndex = Index.at(refID);
\r
568 const BamBinMap& binMap = refIndex.Bins;
\r
570 // get minimum offset to consider
\r
571 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
572 uint64_t minOffset = ( (unsigned int)(left>>BAM_LIDX_SHIFT) >= offsets.size() ) ? 0 : offsets.at(left>>BAM_LIDX_SHIFT);
\r
574 // store offsets to beginning of alignment 'chunks'
\r
575 std::vector<int64_t> chunkStarts;
\r
577 // store all alignment 'chunk' starts for bins in this region
\r
578 for (int i = 0; i < numBins; ++i ) {
\r
579 uint16_t binKey = bins[i];
\r
581 map<uint32_t, ChunkVector>::const_iterator binIter = binMap.find(binKey);
\r
582 if ( (binIter != binMap.end()) && ((*binIter).first == binKey) ) {
\r
584 const ChunkVector& chunks = (*binIter).second;
\r
585 std::vector<Chunk>::const_iterator chunksIter = chunks.begin();
\r
586 std::vector<Chunk>::const_iterator chunksEnd = chunks.end();
\r
587 for ( ; chunksIter != chunksEnd; ++chunksIter) {
\r
588 const Chunk& chunk = (*chunksIter);
\r
589 if ( chunk.Stop > minOffset ) {
\r
590 chunkStarts.push_back( chunk.Start );
\r
599 // if no alignments found, else return smallest offset for alignment starts
\r
600 if ( chunkStarts.size() == 0 ) { return -1; }
\r
601 else { return *min_element(chunkStarts.begin(), chunkStarts.end()); }
\r
604 // saves BAM bin entry for index
\r
605 void BamReader::BamReaderPrivate::InsertBinEntry(BamBinMap& binMap,
\r
606 const uint32_t& saveBin,
\r
607 const uint64_t& saveOffset,
\r
608 const uint64_t& lastOffset)
\r
611 BamBinMap::iterator binIter = binMap.find(saveBin);
\r
613 // create new chunk
\r
614 Chunk newChunk(saveOffset, lastOffset);
\r
616 // if entry doesn't exist
\r
617 if ( binIter == binMap.end() ) {
\r
618 ChunkVector newChunks;
\r
619 newChunks.push_back(newChunk);
\r
620 binMap.insert( pair<uint32_t, ChunkVector>(saveBin, newChunks));
\r
625 ChunkVector& binChunks = (*binIter).second;
\r
626 binChunks.push_back( newChunk );
\r
630 // saves linear offset entry for index
\r
631 void BamReader::BamReaderPrivate::InsertLinearOffset(LinearOffsetVector& offsets,
\r
632 const BamAlignment& bAlignment,
\r
633 const uint64_t& lastOffset)
\r
635 // get converted offsets
\r
636 int beginOffset = bAlignment.Position >> BAM_LIDX_SHIFT;
\r
637 int endOffset = ( CalculateAlignmentEnd(bAlignment.Position, bAlignment.CigarData) - 1) >> BAM_LIDX_SHIFT;
\r
639 // resize vector if necessary
\r
640 int oldSize = offsets.size();
\r
641 int newSize = endOffset + 1;
\r
642 if ( oldSize < newSize ) { offsets.resize(newSize, 0); }
\r
645 for(int i = beginOffset + 1; i <= endOffset ; ++i) {
\r
646 if ( offsets[i] == 0) {
\r
647 offsets[i] = lastOffset;
\r
652 // returns whether alignment overlaps currently specified region (refID, leftBound)
\r
653 bool BamReader::BamReaderPrivate::IsOverlap(BamAlignment& bAlignment) {
\r
655 // if on different reference sequence, quit
\r
656 if ( bAlignment.RefID != CurrentRefID ) { return false; }
\r
658 // read starts after left boundary
\r
659 if ( bAlignment.Position >= CurrentLeft) { return true; }
\r
661 // return whether alignment end overlaps left boundary
\r
662 return ( CalculateAlignmentEnd(bAlignment.Position, bAlignment.CigarData) >= CurrentLeft );
\r
665 // jumps to specified region(refID, leftBound) in BAM file, returns success/fail
\r
666 bool BamReader::BamReaderPrivate::Jump(int refID, int position) {
\r
668 // if data exists for this reference and position is valid
\r
669 if ( References.at(refID).RefHasAlignments && (position <= References.at(refID).RefLength) ) {
\r
671 // set current region
\r
672 CurrentRefID = refID;
\r
673 CurrentLeft = position;
\r
674 IsRegionSpecified = true;
\r
676 // calculate offset
\r
677 int64_t offset = GetOffset(CurrentRefID, CurrentLeft);
\r
679 // if in valid offset, return failure
\r
680 if ( offset == -1 ) { return false; }
\r
682 // otherwise return success of seek operation
\r
683 else { return mBGZF.Seek(offset); }
\r
686 // invalid jump request parameters, return failure
\r
690 // load BAM header data
\r
691 void BamReader::BamReaderPrivate::LoadHeaderData(void) {
\r
693 // check to see if proper BAM header
\r
695 if (mBGZF.Read(buffer, 4) != 4) {
\r
696 printf("Could not read header type\n");
\r
700 if (strncmp(buffer, "BAM\001", 4)) {
\r
701 printf("wrong header type!\n");
\r
705 // get BAM header text length
\r
706 mBGZF.Read(buffer, 4);
\r
707 unsigned int headerTextLength = BgzfData::UnpackUnsignedInt(buffer);
\r
708 if ( IsBigEndian ) { SwapEndian_32(headerTextLength); }
\r
710 // get BAM header text
\r
711 char* headerText = (char*)calloc(headerTextLength + 1, 1);
\r
712 mBGZF.Read(headerText, headerTextLength);
\r
713 HeaderText = (string)((const char*)headerText);
\r
715 // clean up calloc-ed temp variable
\r
719 // load existing index data from BAM index file (".bai"), return success/fail
\r
720 bool BamReader::BamReaderPrivate::LoadIndex(void) {
\r
722 // clear out index data
\r
725 // skip if index file empty
\r
726 if ( IndexFilename.empty() ) { return false; }
\r
728 // open index file, abort on error
\r
729 FILE* indexStream = fopen(IndexFilename.c_str(), "rb");
\r
731 printf("ERROR: Unable to open the BAM index file %s for reading.\n", IndexFilename.c_str() );
\r
735 size_t elementsRead = 0;
\r
737 // see if index is valid BAM index
\r
739 elementsRead = fread(magic, 1, 4, indexStream);
\r
740 if (strncmp(magic, "BAI\1", 4)) {
\r
741 printf("Problem with index file - invalid format.\n");
\r
742 fclose(indexStream);
\r
746 // get number of reference sequences
\r
747 uint32_t numRefSeqs;
\r
748 elementsRead = fread(&numRefSeqs, 4, 1, indexStream);
\r
749 if ( IsBigEndian ) { SwapEndian_32(numRefSeqs); }
\r
751 // intialize space for BamIndex data structure
\r
752 Index.reserve(numRefSeqs);
\r
754 // iterate over reference sequences
\r
755 for (unsigned int i = 0; i < numRefSeqs; ++i) {
\r
757 // get number of bins for this reference sequence
\r
759 elementsRead = fread(&numBins, 4, 1, indexStream);
\r
760 if ( IsBigEndian ) { SwapEndian_32(numBins); }
\r
763 RefData& refEntry = References[i];
\r
764 refEntry.RefHasAlignments = true;
\r
767 // intialize BinVector
\r
770 // iterate over bins for that reference sequence
\r
771 for (int j = 0; j < numBins; ++j) {
\r
775 elementsRead = fread(&binID, 4, 1, indexStream);
\r
777 // get number of regionChunks in this bin
\r
778 uint32_t numChunks;
\r
779 elementsRead = fread(&numChunks, 4, 1, indexStream);
\r
781 if ( IsBigEndian ) {
\r
782 SwapEndian_32(binID);
\r
783 SwapEndian_32(numChunks);
\r
786 // intialize ChunkVector
\r
787 ChunkVector regionChunks;
\r
788 regionChunks.reserve(numChunks);
\r
790 // iterate over regionChunks in this bin
\r
791 for (unsigned int k = 0; k < numChunks; ++k) {
\r
793 // get chunk boundaries (left, right)
\r
796 elementsRead = fread(&left, 8, 1, indexStream);
\r
797 elementsRead = fread(&right, 8, 1, indexStream);
\r
799 if ( IsBigEndian ) {
\r
800 SwapEndian_64(left);
\r
801 SwapEndian_64(right);
\r
805 regionChunks.push_back( Chunk(left, right) );
\r
808 // sort chunks for this bin
\r
809 sort( regionChunks.begin(), regionChunks.end(), ChunkLessThan );
\r
811 // save binID, chunkVector for this bin
\r
812 binMap.insert( pair<uint32_t, ChunkVector>(binID, regionChunks) );
\r
815 // load linear index for this reference sequence
\r
817 // get number of linear offsets
\r
818 int32_t numLinearOffsets;
\r
819 elementsRead = fread(&numLinearOffsets, 4, 1, indexStream);
\r
820 if ( IsBigEndian ) { SwapEndian_32(numLinearOffsets); }
\r
822 // intialize LinearOffsetVector
\r
823 LinearOffsetVector offsets;
\r
824 offsets.reserve(numLinearOffsets);
\r
826 // iterate over linear offsets for this reference sequeence
\r
827 uint64_t linearOffset;
\r
828 for (int j = 0; j < numLinearOffsets; ++j) {
\r
829 // read a linear offset & store
\r
830 elementsRead = fread(&linearOffset, 8, 1, indexStream);
\r
831 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
832 offsets.push_back(linearOffset);
\r
835 // sort linear offsets
\r
836 sort( offsets.begin(), offsets.end() );
\r
838 // store index data for that reference sequence
\r
839 Index.push_back( ReferenceIndex(binMap, offsets) );
\r
842 // close index file (.bai) and return
\r
843 fclose(indexStream);
\r
847 // populates BamAlignment with alignment data under file pointer, returns success/fail
\r
848 bool BamReader::BamReaderPrivate::LoadNextAlignment(BamAlignment& bAlignment, BamAlignmentSupportData& supportData) {
\r
850 // read in the 'block length' value, make sure it's not zero
\r
852 mBGZF.Read(buffer, 4);
\r
853 supportData.BlockLength = BgzfData::UnpackUnsignedInt(buffer);
\r
854 if ( IsBigEndian ) { SwapEndian_32(supportData.BlockLength); }
\r
855 if ( supportData.BlockLength == 0 ) { return false; }
\r
857 // read in core alignment data, make sure the right size of data was read
\r
858 char x[BAM_CORE_SIZE];
\r
859 if ( mBGZF.Read(x, BAM_CORE_SIZE) != BAM_CORE_SIZE ) { return false; }
\r
861 if ( IsBigEndian ) {
\r
862 for ( int i = 0; i < BAM_CORE_SIZE; i+=sizeof(uint32_t) ) {
\r
863 SwapEndian_32p(&x[i]);
\r
867 // set BamAlignment 'core' and 'support' data
\r
868 bAlignment.RefID = BgzfData::UnpackSignedInt(&x[0]);
\r
869 bAlignment.Position = BgzfData::UnpackSignedInt(&x[4]);
\r
871 unsigned int tempValue = BgzfData::UnpackUnsignedInt(&x[8]);
\r
872 bAlignment.Bin = tempValue >> 16;
\r
873 bAlignment.MapQuality = tempValue >> 8 & 0xff;
\r
874 supportData.QueryNameLength = tempValue & 0xff;
\r
876 tempValue = BgzfData::UnpackUnsignedInt(&x[12]);
\r
877 bAlignment.AlignmentFlag = tempValue >> 16;
\r
878 supportData.NumCigarOperations = tempValue & 0xffff;
\r
880 supportData.QuerySequenceLength = BgzfData::UnpackUnsignedInt(&x[16]);
\r
881 bAlignment.MateRefID = BgzfData::UnpackSignedInt(&x[20]);
\r
882 bAlignment.MatePosition = BgzfData::UnpackSignedInt(&x[24]);
\r
883 bAlignment.InsertSize = BgzfData::UnpackSignedInt(&x[28]);
\r
885 // store 'all char data' and cigar ops
\r
886 const unsigned int dataLength = supportData.BlockLength - BAM_CORE_SIZE;
\r
887 const unsigned int cigarDataOffset = supportData.QueryNameLength;
\r
889 char* allCharData = (char*)calloc(sizeof(char), dataLength);
\r
890 uint32_t* cigarData = (uint32_t*)(allCharData + cigarDataOffset);
\r
892 // read in character data - make sure proper data size was read
\r
893 if ( mBGZF.Read(allCharData, dataLength) != (signed int)dataLength) { return false; }
\r
896 // store alignment name and length
\r
897 // bAlignment.Name.clear();
\r
898 bAlignment.Name.assign((const char*)(allCharData));
\r
899 bAlignment.Length = supportData.QuerySequenceLength;
\r
901 // store remaining 'allCharData' in supportData structure
\r
902 // supportData.AllCharData.clear();
\r
903 supportData.AllCharData.assign((const char*)allCharData, dataLength);
\r
905 // save CigarOps for BamAlignment
\r
906 bAlignment.CigarData.clear();
\r
907 for (unsigned int i = 0; i < supportData.NumCigarOperations; ++i) {
\r
909 // swap if necessary
\r
910 if ( IsBigEndian ) { SwapEndian_32(cigarData[i]); }
\r
912 // build CigarOp structure
\r
914 op.Length = (cigarData[i] >> BAM_CIGAR_SHIFT);
\r
915 op.Type = CIGAR_LOOKUP[ (cigarData[i] & BAM_CIGAR_MASK) ];
\r
918 bAlignment.CigarData.push_back(op);
\r
926 // loads reference data from BAM file
\r
927 void BamReader::BamReaderPrivate::LoadReferenceData(void) {
\r
929 // get number of reference sequences
\r
931 mBGZF.Read(buffer, 4);
\r
932 unsigned int numberRefSeqs = BgzfData::UnpackUnsignedInt(buffer);
\r
933 if ( IsBigEndian ) { SwapEndian_32(numberRefSeqs); }
\r
934 if (numberRefSeqs == 0) { return; }
\r
935 References.reserve((int)numberRefSeqs);
\r
937 // iterate over all references in header
\r
938 for (unsigned int i = 0; i != numberRefSeqs; ++i) {
\r
940 // get length of reference name
\r
941 mBGZF.Read(buffer, 4);
\r
942 unsigned int refNameLength = BgzfData::UnpackUnsignedInt(buffer);
\r
943 if ( IsBigEndian ) { SwapEndian_32(refNameLength); }
\r
944 char* refName = (char*)calloc(refNameLength, 1);
\r
946 // get reference name and reference sequence length
\r
947 mBGZF.Read(refName, refNameLength);
\r
948 mBGZF.Read(buffer, 4);
\r
949 int refLength = BgzfData::UnpackSignedInt(buffer);
\r
950 if ( IsBigEndian ) { SwapEndian_32(refLength); }
\r
952 // store data for reference
\r
953 RefData aReference;
\r
954 aReference.RefName = (string)((const char*)refName);
\r
955 aReference.RefLength = refLength;
\r
956 References.push_back(aReference);
\r
958 // clean up calloc-ed temp variable
\r
963 // merges 'alignment chunks' in BAM bin (used for index building)
\r
964 void BamReader::BamReaderPrivate::MergeChunks(void) {
\r
966 // iterate over reference enties
\r
967 BamIndex::iterator indexIter = Index.begin();
\r
968 BamIndex::iterator indexEnd = Index.end();
\r
969 for ( ; indexIter != indexEnd; ++indexIter ) {
\r
971 // get BAM bin map for this reference
\r
972 ReferenceIndex& refIndex = (*indexIter);
\r
973 BamBinMap& bamBinMap = refIndex.Bins;
\r
975 // iterate over BAM bins
\r
976 BamBinMap::iterator binIter = bamBinMap.begin();
\r
977 BamBinMap::iterator binEnd = bamBinMap.end();
\r
978 for ( ; binIter != binEnd; ++binIter ) {
\r
980 // get chunk vector for this bin
\r
981 ChunkVector& binChunks = (*binIter).second;
\r
982 if ( binChunks.size() == 0 ) { continue; }
\r
984 ChunkVector mergedChunks;
\r
985 mergedChunks.push_back( binChunks[0] );
\r
987 // iterate over chunks
\r
989 ChunkVector::iterator chunkIter = binChunks.begin();
\r
990 ChunkVector::iterator chunkEnd = binChunks.end();
\r
991 for ( ++chunkIter; chunkIter != chunkEnd; ++chunkIter) {
\r
993 // get 'currentChunk' based on numeric index
\r
994 Chunk& currentChunk = mergedChunks[i];
\r
996 // get iteratorChunk based on vector iterator
\r
997 Chunk& iteratorChunk = (*chunkIter);
\r
999 // if currentChunk.Stop(shifted) == iterator Chunk.Start(shifted)
\r
1000 if ( currentChunk.Stop>>16 == iteratorChunk.Start>>16 ) {
\r
1002 // set currentChunk.Stop to iteratorChunk.Stop
\r
1003 currentChunk.Stop = iteratorChunk.Stop;
\r
1008 // set currentChunk + 1 to iteratorChunk
\r
1009 mergedChunks.push_back(iteratorChunk);
\r
1014 // saved merged chunk vector
\r
1015 (*binIter).second = mergedChunks;
\r
1020 // opens BAM file (and index)
\r
1021 void BamReader::BamReaderPrivate::Open(const string& filename, const string& indexFilename) {
\r
1023 Filename = filename;
\r
1024 IndexFilename = indexFilename;
\r
1026 // open the BGZF file for reading, retrieve header text & reference data
\r
1027 mBGZF.Open(filename, "rb");
\r
1029 LoadReferenceData();
\r
1031 // store file offset of first alignment
\r
1032 AlignmentsBeginOffset = mBGZF.Tell();
\r
1034 // open index file & load index data (if exists)
\r
1035 if ( !IndexFilename.empty() ) {
\r
1040 // returns BAM file pointer to beginning of alignment data
\r
1041 bool BamReader::BamReaderPrivate::Rewind(void) {
\r
1043 // find first reference that has alignments in the BAM file
\r
1045 int refCount = References.size();
\r
1046 for ( ; refID < refCount; ++refID ) {
\r
1047 if ( References.at(refID).RefHasAlignments ) { break; }
\r
1050 // store default bounds for first alignment
\r
1051 CurrentRefID = refID;
\r
1053 IsRegionSpecified = false;
\r
1055 // return success/failure of seek
\r
1056 return mBGZF.Seek(AlignmentsBeginOffset);
\r
1059 // saves index data to BAM index file (".bai"), returns success/fail
\r
1060 bool BamReader::BamReaderPrivate::WriteIndex(void) {
\r
1062 IndexFilename = Filename + ".bai";
\r
1063 FILE* indexStream = fopen(IndexFilename.c_str(), "wb");
\r
1064 if ( indexStream == 0 ) {
\r
1065 printf("ERROR: Could not open file to save index\n");
\r
1069 // write BAM index header
\r
1070 fwrite("BAI\1", 1, 4, indexStream);
\r
1072 // write number of reference sequences
\r
1073 int32_t numReferenceSeqs = Index.size();
\r
1074 if ( IsBigEndian ) { SwapEndian_32(numReferenceSeqs); }
\r
1075 fwrite(&numReferenceSeqs, 4, 1, indexStream);
\r
1077 // iterate over reference sequences
\r
1078 BamIndex::const_iterator indexIter = Index.begin();
\r
1079 BamIndex::const_iterator indexEnd = Index.end();
\r
1080 for ( ; indexIter != indexEnd; ++ indexIter ) {
\r
1082 // get reference index data
\r
1083 const ReferenceIndex& refIndex = (*indexIter);
\r
1084 const BamBinMap& binMap = refIndex.Bins;
\r
1085 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
1087 // write number of bins
\r
1088 int32_t binCount = binMap.size();
\r
1089 if ( IsBigEndian ) { SwapEndian_32(binCount); }
\r
1090 fwrite(&binCount, 4, 1, indexStream);
\r
1092 // iterate over bins
\r
1093 BamBinMap::const_iterator binIter = binMap.begin();
\r
1094 BamBinMap::const_iterator binEnd = binMap.end();
\r
1095 for ( ; binIter != binEnd; ++binIter ) {
\r
1097 // get bin data (key and chunk vector)
\r
1098 uint32_t binKey = (*binIter).first;
\r
1099 const ChunkVector& binChunks = (*binIter).second;
\r
1101 // save BAM bin key
\r
1102 if ( IsBigEndian ) { SwapEndian_32(binKey); }
\r
1103 fwrite(&binKey, 4, 1, indexStream);
\r
1105 // save chunk count
\r
1106 int32_t chunkCount = binChunks.size();
\r
1107 if ( IsBigEndian ) { SwapEndian_32(chunkCount); }
\r
1108 fwrite(&chunkCount, 4, 1, indexStream);
\r
1110 // iterate over chunks
\r
1111 ChunkVector::const_iterator chunkIter = binChunks.begin();
\r
1112 ChunkVector::const_iterator chunkEnd = binChunks.end();
\r
1113 for ( ; chunkIter != chunkEnd; ++chunkIter ) {
\r
1115 // get current chunk data
\r
1116 const Chunk& chunk = (*chunkIter);
\r
1117 uint64_t start = chunk.Start;
\r
1118 uint64_t stop = chunk.Stop;
\r
1120 if ( IsBigEndian ) {
\r
1121 SwapEndian_64(start);
\r
1122 SwapEndian_64(stop);
\r
1125 // save chunk offsets
\r
1126 fwrite(&start, 8, 1, indexStream);
\r
1127 fwrite(&stop, 8, 1, indexStream);
\r
1131 // write linear offsets size
\r
1132 int32_t offsetSize = offsets.size();
\r
1133 if ( IsBigEndian ) { SwapEndian_32(offsetSize); }
\r
1134 fwrite(&offsetSize, 4, 1, indexStream);
\r
1136 // iterate over linear offsets
\r
1137 LinearOffsetVector::const_iterator offsetIter = offsets.begin();
\r
1138 LinearOffsetVector::const_iterator offsetEnd = offsets.end();
\r
1139 for ( ; offsetIter != offsetEnd; ++offsetIter ) {
\r
1141 // write linear offset value
\r
1142 uint64_t linearOffset = (*offsetIter);
\r
1143 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
1144 fwrite(&linearOffset, 8, 1, indexStream);
\r
1148 // flush buffer, close file, and return success
\r
1149 fflush(indexStream);
\r
1150 fclose(indexStream);
\r