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: 8 June 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 struct BamReader::BamReaderPrivate {
\r
28 // -------------------------------
\r
30 // -------------------------------
\r
32 // general file data
\r
36 RefVector References;
\r
38 int64_t AlignmentsBeginOffset;
\r
40 string IndexFilename;
\r
45 // user-specified region values
\r
46 bool IsRegionSpecified;
\r
50 // BAM character constants
\r
51 const char* DNA_LOOKUP;
\r
52 const char* CIGAR_LOOKUP;
\r
54 // -------------------------------
\r
55 // constructor & destructor
\r
56 // -------------------------------
\r
57 BamReaderPrivate(void);
\r
58 ~BamReaderPrivate(void);
\r
60 // -------------------------------
\r
61 // "public" interface
\r
62 // -------------------------------
\r
66 bool Jump(int refID, int position = 0);
\r
67 void Open(const string& filename, const string& indexFilename = "");
\r
70 // access alignment data
\r
71 bool GetNextAlignment(BamAlignment& bAlignment);
\r
72 bool GetNextAlignmentCore(BamAlignment& bAlignment);
\r
74 // access auxiliary data
\r
75 int GetReferenceID(const string& refName) const;
\r
78 bool CreateIndex(void);
\r
80 // -------------------------------
\r
82 // -------------------------------
\r
84 // *** reading alignments and auxiliary data *** //
\r
86 // calculate bins that overlap region ( left to reference end for now )
\r
87 int BinsFromRegion(int refID, int left, uint16_t[MAX_BIN]);
\r
88 // fills out character data for BamAlignment data
\r
89 bool BuildCharData(BamAlignment& bAlignment);
\r
90 // calculate file offset for first alignment chunk overlapping 'left'
\r
91 int64_t GetOffset(int refID, int left);
\r
92 // checks to see if alignment overlaps current region
\r
93 bool IsOverlap(BamAlignment& bAlignment);
\r
94 // retrieves header text from BAM file
\r
95 void LoadHeaderData(void);
\r
96 // retrieves BAM alignment under file pointer
\r
97 bool LoadNextAlignment(BamAlignment& bAlignment);
\r
98 // builds reference data structure from BAM file
\r
99 void LoadReferenceData(void);
\r
101 // *** index file handling *** //
\r
103 // calculates index for BAM file
\r
104 bool BuildIndex(void);
\r
105 // clear out inernal index data structure
\r
106 void ClearIndex(void);
\r
107 // saves BAM bin entry for index
\r
108 void InsertBinEntry(BamBinMap& binMap, const uint32_t& saveBin, const uint64_t& saveOffset, const uint64_t& lastOffset);
\r
109 // saves linear offset entry for index
\r
110 void InsertLinearOffset(LinearOffsetVector& offsets, const BamAlignment& bAlignment, const uint64_t& lastOffset);
\r
111 // loads index from BAM index file
\r
112 bool LoadIndex(void);
\r
113 // simplifies index by merging 'chunks'
\r
114 void MergeChunks(void);
\r
115 // saves index to BAM index file
\r
116 bool WriteIndex(void);
\r
119 // -----------------------------------------------------
\r
120 // BamReader implementation (wrapper around BRPrivate)
\r
121 // -----------------------------------------------------
\r
124 BamReader::BamReader(void) {
\r
125 d = new BamReaderPrivate;
\r
129 BamReader::~BamReader(void) {
\r
135 void BamReader::Close(void) { d->Close(); }
\r
136 bool BamReader::Jump(int refID, int position) { return d->Jump(refID, position); }
\r
137 void BamReader::Open(const string& filename, const string& indexFilename) { d->Open(filename, indexFilename); }
\r
138 bool BamReader::Rewind(void) { return d->Rewind(); }
\r
140 // access alignment data
\r
141 bool BamReader::GetNextAlignment(BamAlignment& bAlignment) { return d->GetNextAlignment(bAlignment); }
\r
142 bool BamReader::GetNextAlignmentCore(BamAlignment& bAlignment) { return d->GetNextAlignmentCore(bAlignment); }
\r
144 // access auxiliary data
\r
145 const string BamReader::GetHeaderText(void) const { return d->HeaderText; }
\r
146 int BamReader::GetReferenceCount(void) const { return d->References.size(); }
\r
147 const RefVector BamReader::GetReferenceData(void) const { return d->References; }
\r
148 int BamReader::GetReferenceID(const string& refName) const { return d->GetReferenceID(refName); }
\r
149 const std::string BamReader::GetFilename(void) const { return d->Filename; }
\r
151 // index operations
\r
152 bool BamReader::CreateIndex(void) { return d->CreateIndex(); }
\r
154 // -----------------------------------------------------
\r
155 // BamReaderPrivate implementation
\r
156 // -----------------------------------------------------
\r
159 BamReader::BamReaderPrivate::BamReaderPrivate(void)
\r
160 : IsIndexLoaded(false)
\r
161 , AlignmentsBeginOffset(0)
\r
162 , IsRegionSpecified(false)
\r
165 , DNA_LOOKUP("=ACMGRSVTWYHKDBN")
\r
166 , CIGAR_LOOKUP("MIDNSHP")
\r
168 IsBigEndian = SystemIsBigEndian();
\r
172 BamReader::BamReaderPrivate::~BamReaderPrivate(void) {
\r
176 // calculate bins that overlap region ( left to reference end for now )
\r
177 int BamReader::BamReaderPrivate::BinsFromRegion(int refID, int left, uint16_t list[MAX_BIN]) {
\r
179 // get region boundaries
\r
180 uint32_t begin = (unsigned int)left;
\r
181 uint32_t end = (unsigned int)References.at(refID).RefLength - 1;
\r
183 // initialize list, bin '0' always a valid bin
\r
187 // get rest of bins that contain this region
\r
189 for (k = 1 + (begin>>26); k <= 1 + (end>>26); ++k) { list[i++] = k; }
\r
190 for (k = 9 + (begin>>23); k <= 9 + (end>>23); ++k) { list[i++] = k; }
\r
191 for (k = 73 + (begin>>20); k <= 73 + (end>>20); ++k) { list[i++] = k; }
\r
192 for (k = 585 + (begin>>17); k <= 585 + (end>>17); ++k) { list[i++] = k; }
\r
193 for (k = 4681 + (begin>>14); k <= 4681 + (end>>14); ++k) { list[i++] = k; }
\r
195 // return number of bins stored
\r
199 bool BamReader::BamReaderPrivate::BuildCharData(BamAlignment& bAlignment) {
\r
201 // calculate character lengths/offsets
\r
202 const unsigned int dataLength = bAlignment.SupportData.BlockLength - BAM_CORE_SIZE;
\r
203 const unsigned int seqDataOffset = bAlignment.SupportData.QueryNameLength + (bAlignment.SupportData.NumCigarOperations * 4);
\r
204 const unsigned int qualDataOffset = seqDataOffset + (bAlignment.SupportData.QuerySequenceLength+1)/2;
\r
205 const unsigned int tagDataOffset = qualDataOffset + bAlignment.SupportData.QuerySequenceLength;
\r
206 const unsigned int tagDataLength = dataLength - tagDataOffset;
\r
208 // set up char buffers
\r
209 const char* allCharData = bAlignment.SupportData.AllCharData.data();
\r
210 const char* seqData = ((const char*)allCharData) + seqDataOffset;
\r
211 const char* qualData = ((const char*)allCharData) + qualDataOffset;
\r
212 char* tagData = ((char*)allCharData) + tagDataOffset;
\r
214 // save query sequence
\r
215 bAlignment.QueryBases.clear();
\r
216 bAlignment.QueryBases.reserve(bAlignment.SupportData.QuerySequenceLength);
\r
217 for (unsigned int i = 0; i < bAlignment.SupportData.QuerySequenceLength; ++i) {
\r
218 char singleBase = DNA_LOOKUP[ ( ( seqData[(i/2)] >> (4*(1-(i%2)))) & 0xf ) ];
\r
219 bAlignment.QueryBases.append(1, singleBase);
\r
222 // save qualities, converting from numeric QV to 'FASTQ-style' ASCII character
\r
223 bAlignment.Qualities.clear();
\r
224 bAlignment.Qualities.reserve(bAlignment.SupportData.QuerySequenceLength);
\r
225 for (unsigned int i = 0; i < bAlignment.SupportData.QuerySequenceLength; ++i) {
\r
226 char singleQuality = (char)(qualData[i]+33);
\r
227 bAlignment.Qualities.append(1, singleQuality);
\r
230 // parse CIGAR to build 'AlignedBases'
\r
231 bAlignment.AlignedBases.clear();
\r
232 bAlignment.AlignedBases.reserve(bAlignment.SupportData.QuerySequenceLength);
\r
235 vector<CigarOp>::const_iterator cigarIter = bAlignment.CigarData.begin();
\r
236 vector<CigarOp>::const_iterator cigarEnd = bAlignment.CigarData.end();
\r
237 for ( ; cigarIter != cigarEnd; ++cigarIter ) {
\r
239 const CigarOp& op = (*cigarIter);
\r
244 bAlignment.AlignedBases.append(bAlignment.QueryBases.substr(k, op.Length)); // for 'M', 'I' - write bases
\r
248 k += op.Length; // for 'S' - soft clip, skip over query bases
\r
252 bAlignment.AlignedBases.append(op.Length, '-'); // for 'D' - write gap character
\r
256 bAlignment.AlignedBases.append( op.Length, '*' ); // for 'P' - write padding character
\r
260 bAlignment.AlignedBases.append( op.Length, 'N' ); // for 'N' - write N's, skip bases in original query sequence
\r
265 break; // for 'H' - hard clip, do nothing to AlignedBases, move to next op
\r
268 printf("ERROR: Invalid Cigar op type\n"); // shouldn't get here
\r
273 // -----------------------
\r
274 // Added: 3-25-2010 DWB
\r
275 // Fixed: endian-correctness for tag data
\r
276 // -----------------------
\r
277 if ( IsBigEndian ) {
\r
279 while ( (unsigned int)i < tagDataLength ) {
\r
281 i += 2; // skip tag type (e.g. "RG", "NM", etc)
\r
282 uint8_t type = toupper(tagData[i]); // lower & upper case letters have same meaning
\r
283 ++i; // skip value type
\r
293 SwapEndian_16p(&tagData[i]);
\r
294 i+=2; // sizeof(uint16_t)
\r
299 SwapEndian_32p(&tagData[i]);
\r
300 i+=4; // sizeof(uint32_t)
\r
304 SwapEndian_64p(&tagData[i]);
\r
305 i+=8; // sizeof(uint64_t)
\r
310 while (tagData[i]) { ++i; }
\r
311 ++i; // increment one more for null terminator
\r
315 printf("ERROR: Invalid tag value type\n"); // shouldn't get here
\r
322 bAlignment.TagData.clear();
\r
323 bAlignment.TagData.resize(tagDataLength);
\r
324 memcpy((char*)bAlignment.TagData.data(), tagData, tagDataLength);
\r
326 // set support data parsed flag
\r
327 bAlignment.SupportData.IsParsed = true;
\r
333 // populates BAM index data structure from BAM file data
\r
334 bool BamReader::BamReaderPrivate::BuildIndex(void) {
\r
336 // check to be sure file is open
\r
337 if (!mBGZF.IsOpen) { return false; }
\r
339 // move file pointer to beginning of alignments
\r
342 // get reference count, reserve index space
\r
343 int numReferences = References.size();
\r
344 for ( int i = 0; i < numReferences; ++i ) {
\r
345 Index.push_back(ReferenceIndex());
\r
348 // sets default constant for bin, ID, offset, coordinate variables
\r
349 const uint32_t defaultValue = 0xffffffffu;
\r
352 uint32_t saveBin(defaultValue);
\r
353 uint32_t lastBin(defaultValue);
\r
355 // reference ID data
\r
356 int32_t saveRefID(defaultValue);
\r
357 int32_t lastRefID(defaultValue);
\r
360 uint64_t saveOffset = mBGZF.Tell();
\r
361 uint64_t lastOffset = saveOffset;
\r
364 int32_t lastCoordinate = defaultValue;
\r
366 BamAlignment bAlignment;
\r
367 while( GetNextAlignment(bAlignment) ) {
\r
369 // change of chromosome, save ID, reset bin
\r
370 if ( lastRefID != bAlignment.RefID ) {
\r
371 lastRefID = bAlignment.RefID;
\r
372 lastBin = defaultValue;
\r
375 // if lastCoordinate greater than BAM position - file not sorted properly
\r
376 else if ( lastCoordinate > bAlignment.Position ) {
\r
377 printf("BAM file not properly sorted:\n");
\r
378 printf("Alignment %s : %d > %d on reference (id = %d)", bAlignment.Name.c_str(), lastCoordinate, bAlignment.Position, bAlignment.RefID);
\r
382 // if valid reference && BAM bin spans some minimum cutoff (smaller bin ids span larger regions)
\r
383 if ( (bAlignment.RefID >= 0) && (bAlignment.Bin < 4681) ) {
\r
385 // save linear offset entry (matched to BAM entry refID)
\r
386 ReferenceIndex& refIndex = Index.at(bAlignment.RefID);
\r
387 LinearOffsetVector& offsets = refIndex.Offsets;
\r
388 InsertLinearOffset(offsets, bAlignment, lastOffset);
\r
391 // if current BamAlignment bin != lastBin, "then possibly write the binning index"
\r
392 if ( bAlignment.Bin != lastBin ) {
\r
394 // if not first time through
\r
395 if ( saveBin != defaultValue ) {
\r
397 // save Bam bin entry
\r
398 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
399 BamBinMap& binMap = refIndex.Bins;
\r
400 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
403 // update saveOffset
\r
404 saveOffset = lastOffset;
\r
406 // update bin values
\r
407 saveBin = bAlignment.Bin;
\r
408 lastBin = bAlignment.Bin;
\r
410 // update saveRefID
\r
411 saveRefID = bAlignment.RefID;
\r
413 // if invalid RefID, break out (why?)
\r
414 if ( saveRefID < 0 ) { break; }
\r
417 // make sure that current file pointer is beyond lastOffset
\r
418 if ( mBGZF.Tell() <= (int64_t)lastOffset ) {
\r
419 printf("Error in BGZF offsets.\n");
\r
423 // update lastOffset
\r
424 lastOffset = mBGZF.Tell();
\r
426 // update lastCoordinate
\r
427 lastCoordinate = bAlignment.Position;
\r
430 // save any leftover BAM data (as long as refID is valid)
\r
431 if ( saveRefID >= 0 ) {
\r
432 // save Bam bin entry
\r
433 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
434 BamBinMap& binMap = refIndex.Bins;
\r
435 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
438 // simplify index by merging chunks
\r
441 // iterate over references
\r
442 BamIndex::iterator indexIter = Index.begin();
\r
443 BamIndex::iterator indexEnd = Index.end();
\r
444 for ( int i = 0; indexIter != indexEnd; ++indexIter, ++i ) {
\r
446 // get reference index data
\r
447 ReferenceIndex& refIndex = (*indexIter);
\r
448 BamBinMap& binMap = refIndex.Bins;
\r
449 LinearOffsetVector& offsets = refIndex.Offsets;
\r
451 // store whether reference has alignments or no
\r
452 References[i].RefHasAlignments = ( binMap.size() > 0 );
\r
454 // sort linear offsets
\r
455 sort(offsets.begin(), offsets.end());
\r
459 // rewind file pointer to beginning of alignments, return success/fail
\r
464 // clear index data structure
\r
465 void BamReader::BamReaderPrivate::ClearIndex(void) {
\r
466 Index.clear(); // sufficient ??
\r
469 // closes the BAM file
\r
470 void BamReader::BamReaderPrivate::Close(void) {
\r
473 HeaderText.clear();
\r
474 IsRegionSpecified = false;
\r
477 // create BAM index from BAM file (keep structure in memory) and write to default index output file
\r
478 bool BamReader::BamReaderPrivate::CreateIndex(void) {
\r
483 // build (& save) index from BAM file
\r
485 ok &= BuildIndex();
\r
486 ok &= WriteIndex();
\r
488 // return success/fail
\r
492 // returns RefID for given RefName (returns References.size() if not found)
\r
493 int BamReader::BamReaderPrivate::GetReferenceID(const string& refName) const {
\r
495 // retrieve names from reference data
\r
496 vector<string> refNames;
\r
497 RefVector::const_iterator refIter = References.begin();
\r
498 RefVector::const_iterator refEnd = References.end();
\r
499 for ( ; refIter != refEnd; ++refIter) {
\r
500 refNames.push_back( (*refIter).RefName );
\r
503 // return 'index-of' refName ( if not found, returns refNames.size() )
\r
504 return distance(refNames.begin(), find(refNames.begin(), refNames.end(), refName));
\r
507 // get next alignment (from specified region, if given)
\r
508 bool BamReader::BamReaderPrivate::GetNextAlignment(BamAlignment& bAlignment) {
\r
510 // if valid alignment available
\r
511 if ( LoadNextAlignment(bAlignment) ) {
\r
513 // if region not specified, return success
\r
514 if ( !IsRegionSpecified ) {
\r
515 bool ok = BuildCharData(bAlignment);
\r
519 // load next alignment until region overlap is found
\r
520 while ( !IsOverlap(bAlignment) ) {
\r
521 // if no valid alignment available (likely EOF) return failure
\r
522 if ( !LoadNextAlignment(bAlignment) ) return false;
\r
525 // return success (alignment found that overlaps region)
\r
526 bool ok = BuildCharData(bAlignment);
\r
530 // no valid alignment
\r
535 // retrieves next available alignment core data (returns success/fail)
\r
536 // ** DOES NOT parse any character data (bases, qualities, tag data)
\r
537 // these can be accessed, if necessary, from the supportData
\r
538 // useful for operations requiring ONLY positional or other alignment-related information
\r
539 bool BamReader::BamReaderPrivate::GetNextAlignmentCore(BamAlignment& bAlignment) {
\r
541 // if valid alignment available
\r
542 if ( LoadNextAlignment(bAlignment) ) {
\r
544 // if region not specified, return success
\r
545 if ( !IsRegionSpecified ) return true;
\r
547 // load next alignment until region overlap is found
\r
548 while ( !IsOverlap(bAlignment) ) {
\r
549 // if no valid alignment available (likely EOF) return failure
\r
550 if ( !LoadNextAlignment(bAlignment) ) return false;
\r
553 // return success (alignment found that overlaps region)
\r
557 // no valid alignment
\r
562 // calculate closest indexed file offset for region specified
\r
563 int64_t BamReader::BamReaderPrivate::GetOffset(int refID, int left) {
\r
565 // calculate which bins overlap this region
\r
566 uint16_t* bins = (uint16_t*)calloc(MAX_BIN, 2);
\r
567 int numBins = BinsFromRegion(refID, left, bins);
\r
569 // get bins for this reference
\r
570 const ReferenceIndex& refIndex = Index.at(refID);
\r
571 const BamBinMap& binMap = refIndex.Bins;
\r
573 // get minimum offset to consider
\r
574 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
575 uint64_t minOffset = ( (unsigned int)(left>>BAM_LIDX_SHIFT) >= offsets.size() ) ? 0 : offsets.at(left>>BAM_LIDX_SHIFT);
\r
577 // store offsets to beginning of alignment 'chunks'
\r
578 std::vector<int64_t> chunkStarts;
\r
580 // store all alignment 'chunk' starts for bins in this region
\r
581 for (int i = 0; i < numBins; ++i ) {
\r
582 uint16_t binKey = bins[i];
\r
584 map<uint32_t, ChunkVector>::const_iterator binIter = binMap.find(binKey);
\r
585 if ( (binIter != binMap.end()) && ((*binIter).first == binKey) ) {
\r
587 const ChunkVector& chunks = (*binIter).second;
\r
588 std::vector<Chunk>::const_iterator chunksIter = chunks.begin();
\r
589 std::vector<Chunk>::const_iterator chunksEnd = chunks.end();
\r
590 for ( ; chunksIter != chunksEnd; ++chunksIter) {
\r
591 const Chunk& chunk = (*chunksIter);
\r
592 if ( chunk.Stop > minOffset ) {
\r
593 chunkStarts.push_back( chunk.Start );
\r
602 // if no alignments found, else return smallest offset for alignment starts
\r
603 if ( chunkStarts.size() == 0 ) { return -1; }
\r
604 else { return *min_element(chunkStarts.begin(), chunkStarts.end()); }
\r
607 // saves BAM bin entry for index
\r
608 void BamReader::BamReaderPrivate::InsertBinEntry(BamBinMap& binMap,
\r
609 const uint32_t& saveBin,
\r
610 const uint64_t& saveOffset,
\r
611 const uint64_t& lastOffset)
\r
614 BamBinMap::iterator binIter = binMap.find(saveBin);
\r
616 // create new chunk
\r
617 Chunk newChunk(saveOffset, lastOffset);
\r
619 // if entry doesn't exist
\r
620 if ( binIter == binMap.end() ) {
\r
621 ChunkVector newChunks;
\r
622 newChunks.push_back(newChunk);
\r
623 binMap.insert( pair<uint32_t, ChunkVector>(saveBin, newChunks));
\r
628 ChunkVector& binChunks = (*binIter).second;
\r
629 binChunks.push_back( newChunk );
\r
633 // saves linear offset entry for index
\r
634 void BamReader::BamReaderPrivate::InsertLinearOffset(LinearOffsetVector& offsets,
\r
635 const BamAlignment& bAlignment,
\r
636 const uint64_t& lastOffset)
\r
638 // get converted offsets
\r
639 int beginOffset = bAlignment.Position >> BAM_LIDX_SHIFT;
\r
640 int endOffset = (bAlignment.GetEndPosition() - 1) >> BAM_LIDX_SHIFT;
\r
642 // resize vector if necessary
\r
643 int oldSize = offsets.size();
\r
644 int newSize = endOffset + 1;
\r
645 if ( oldSize < newSize ) { offsets.resize(newSize, 0); }
\r
648 for(int i = beginOffset + 1; i <= endOffset ; ++i) {
\r
649 if ( offsets[i] == 0) {
\r
650 offsets[i] = lastOffset;
\r
655 // returns whether alignment overlaps currently specified region (refID, leftBound)
\r
656 bool BamReader::BamReaderPrivate::IsOverlap(BamAlignment& bAlignment) {
\r
658 // if on different reference sequence, quit
\r
659 if ( bAlignment.RefID != CurrentRefID ) { return false; }
\r
661 // read starts after left boundary
\r
662 if ( bAlignment.Position >= CurrentLeft) { return true; }
\r
664 // return whether alignment end overlaps left boundary
\r
665 return ( bAlignment.GetEndPosition() >= CurrentLeft );
\r
668 // jumps to specified region(refID, leftBound) in BAM file, returns success/fail
\r
669 bool BamReader::BamReaderPrivate::Jump(int refID, int position) {
\r
671 // if data exists for this reference and position is valid
\r
672 if ( References.at(refID).RefHasAlignments && (position <= References.at(refID).RefLength) ) {
\r
674 // set current region
\r
675 CurrentRefID = refID;
\r
676 CurrentLeft = position;
\r
677 IsRegionSpecified = true;
\r
679 // calculate offset
\r
680 int64_t offset = GetOffset(CurrentRefID, CurrentLeft);
\r
682 // if in valid offset, return failure
\r
683 if ( offset == -1 ) { return false; }
\r
685 // otherwise return success of seek operation
\r
686 else { return mBGZF.Seek(offset); }
\r
689 // invalid jump request parameters, return failure
\r
693 // load BAM header data
\r
694 void BamReader::BamReaderPrivate::LoadHeaderData(void) {
\r
696 // check to see if proper BAM header
\r
698 if (mBGZF.Read(buffer, 4) != 4) {
\r
699 printf("Could not read header type\n");
\r
703 if (strncmp(buffer, "BAM\001", 4)) {
\r
704 printf("wrong header type!\n");
\r
708 // get BAM header text length
\r
709 mBGZF.Read(buffer, 4);
\r
710 unsigned int headerTextLength = BgzfData::UnpackUnsignedInt(buffer);
\r
711 if ( IsBigEndian ) { SwapEndian_32(headerTextLength); }
\r
713 // get BAM header text
\r
714 char* headerText = (char*)calloc(headerTextLength + 1, 1);
\r
715 mBGZF.Read(headerText, headerTextLength);
\r
716 HeaderText = (string)((const char*)headerText);
\r
718 // clean up calloc-ed temp variable
\r
722 // load existing index data from BAM index file (".bai"), return success/fail
\r
723 bool BamReader::BamReaderPrivate::LoadIndex(void) {
\r
725 // clear out index data
\r
728 // skip if index file empty
\r
729 if ( IndexFilename.empty() ) { return false; }
\r
731 // open index file, abort on error
\r
732 FILE* indexStream = fopen(IndexFilename.c_str(), "rb");
\r
734 printf("ERROR: Unable to open the BAM index file %s for reading.\n", IndexFilename.c_str() );
\r
738 size_t elementsRead = 0;
\r
740 // see if index is valid BAM index
\r
742 elementsRead = fread(magic, 1, 4, indexStream);
\r
743 if (strncmp(magic, "BAI\1", 4)) {
\r
744 printf("Problem with index file - invalid format.\n");
\r
745 fclose(indexStream);
\r
749 // get number of reference sequences
\r
750 uint32_t numRefSeqs;
\r
751 elementsRead = fread(&numRefSeqs, 4, 1, indexStream);
\r
752 if ( IsBigEndian ) { SwapEndian_32(numRefSeqs); }
\r
754 // intialize space for BamIndex data structure
\r
755 Index.reserve(numRefSeqs);
\r
757 // iterate over reference sequences
\r
758 for (unsigned int i = 0; i < numRefSeqs; ++i) {
\r
760 // get number of bins for this reference sequence
\r
762 elementsRead = fread(&numBins, 4, 1, indexStream);
\r
763 if ( IsBigEndian ) { SwapEndian_32(numBins); }
\r
766 RefData& refEntry = References[i];
\r
767 refEntry.RefHasAlignments = true;
\r
770 // intialize BinVector
\r
773 // iterate over bins for that reference sequence
\r
774 for (int j = 0; j < numBins; ++j) {
\r
778 elementsRead = fread(&binID, 4, 1, indexStream);
\r
780 // get number of regionChunks in this bin
\r
781 uint32_t numChunks;
\r
782 elementsRead = fread(&numChunks, 4, 1, indexStream);
\r
784 if ( IsBigEndian ) {
\r
785 SwapEndian_32(binID);
\r
786 SwapEndian_32(numChunks);
\r
789 // intialize ChunkVector
\r
790 ChunkVector regionChunks;
\r
791 regionChunks.reserve(numChunks);
\r
793 // iterate over regionChunks in this bin
\r
794 for (unsigned int k = 0; k < numChunks; ++k) {
\r
796 // get chunk boundaries (left, right)
\r
799 elementsRead = fread(&left, 8, 1, indexStream);
\r
800 elementsRead = fread(&right, 8, 1, indexStream);
\r
802 if ( IsBigEndian ) {
\r
803 SwapEndian_64(left);
\r
804 SwapEndian_64(right);
\r
808 regionChunks.push_back( Chunk(left, right) );
\r
811 // sort chunks for this bin
\r
812 sort( regionChunks.begin(), regionChunks.end(), ChunkLessThan );
\r
814 // save binID, chunkVector for this bin
\r
815 binMap.insert( pair<uint32_t, ChunkVector>(binID, regionChunks) );
\r
818 // load linear index for this reference sequence
\r
820 // get number of linear offsets
\r
821 int32_t numLinearOffsets;
\r
822 elementsRead = fread(&numLinearOffsets, 4, 1, indexStream);
\r
823 if ( IsBigEndian ) { SwapEndian_32(numLinearOffsets); }
\r
825 // intialize LinearOffsetVector
\r
826 LinearOffsetVector offsets;
\r
827 offsets.reserve(numLinearOffsets);
\r
829 // iterate over linear offsets for this reference sequeence
\r
830 uint64_t linearOffset;
\r
831 for (int j = 0; j < numLinearOffsets; ++j) {
\r
832 // read a linear offset & store
\r
833 elementsRead = fread(&linearOffset, 8, 1, indexStream);
\r
834 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
835 offsets.push_back(linearOffset);
\r
838 // sort linear offsets
\r
839 sort( offsets.begin(), offsets.end() );
\r
841 // store index data for that reference sequence
\r
842 Index.push_back( ReferenceIndex(binMap, offsets) );
\r
845 // close index file (.bai) and return
\r
846 fclose(indexStream);
\r
850 // populates BamAlignment with alignment data under file pointer, returns success/fail
\r
851 bool BamReader::BamReaderPrivate::LoadNextAlignment(BamAlignment& bAlignment) {
\r
853 // read in the 'block length' value, make sure it's not zero
\r
855 mBGZF.Read(buffer, 4);
\r
856 bAlignment.SupportData.BlockLength = BgzfData::UnpackUnsignedInt(buffer);
\r
857 if ( IsBigEndian ) { SwapEndian_32(bAlignment.SupportData.BlockLength); }
\r
858 if ( bAlignment.SupportData.BlockLength == 0 ) { return false; }
\r
860 // read in core alignment data, make sure the right size of data was read
\r
861 char x[BAM_CORE_SIZE];
\r
862 if ( mBGZF.Read(x, BAM_CORE_SIZE) != BAM_CORE_SIZE ) { return false; }
\r
864 if ( IsBigEndian ) {
\r
865 for ( int i = 0; i < BAM_CORE_SIZE; i+=sizeof(uint32_t) ) {
\r
866 SwapEndian_32p(&x[i]);
\r
870 // set BamAlignment 'core' and 'support' data
\r
871 bAlignment.RefID = BgzfData::UnpackSignedInt(&x[0]);
\r
872 bAlignment.Position = BgzfData::UnpackSignedInt(&x[4]);
\r
874 unsigned int tempValue = BgzfData::UnpackUnsignedInt(&x[8]);
\r
875 bAlignment.Bin = tempValue >> 16;
\r
876 bAlignment.MapQuality = tempValue >> 8 & 0xff;
\r
877 bAlignment.SupportData.QueryNameLength = tempValue & 0xff;
\r
879 tempValue = BgzfData::UnpackUnsignedInt(&x[12]);
\r
880 bAlignment.AlignmentFlag = tempValue >> 16;
\r
881 bAlignment.SupportData.NumCigarOperations = tempValue & 0xffff;
\r
883 bAlignment.SupportData.QuerySequenceLength = BgzfData::UnpackUnsignedInt(&x[16]);
\r
884 bAlignment.MateRefID = BgzfData::UnpackSignedInt(&x[20]);
\r
885 bAlignment.MatePosition = BgzfData::UnpackSignedInt(&x[24]);
\r
886 bAlignment.InsertSize = BgzfData::UnpackSignedInt(&x[28]);
\r
888 // store 'all char data' and cigar ops
\r
889 const unsigned int dataLength = bAlignment.SupportData.BlockLength - BAM_CORE_SIZE;
\r
890 const unsigned int cigarDataOffset = bAlignment.SupportData.QueryNameLength;
\r
892 char* allCharData = (char*)calloc(sizeof(char), dataLength);
\r
893 uint32_t* cigarData = (uint32_t*)(allCharData + cigarDataOffset);
\r
895 // read in character data - make sure proper data size was read
\r
896 if ( mBGZF.Read(allCharData, dataLength) != (signed int)dataLength) { return false; }
\r
899 // store alignment name and length
\r
900 bAlignment.Name.assign((const char*)(allCharData));
\r
901 bAlignment.Length = bAlignment.SupportData.QuerySequenceLength;
\r
903 // store remaining 'allCharData' in supportData structure
\r
904 bAlignment.SupportData.AllCharData.assign((const char*)allCharData, dataLength);
\r
906 // save CigarOps for BamAlignment
\r
907 bAlignment.CigarData.clear();
\r
908 for (unsigned int i = 0; i < bAlignment.SupportData.NumCigarOperations; ++i) {
\r
910 // swap if necessary
\r
911 if ( IsBigEndian ) { SwapEndian_32(cigarData[i]); }
\r
913 // build CigarOp structure
\r
915 op.Length = (cigarData[i] >> BAM_CIGAR_SHIFT);
\r
916 op.Type = CIGAR_LOOKUP[ (cigarData[i] & BAM_CIGAR_MASK) ];
\r
919 bAlignment.CigarData.push_back(op);
\r
927 // loads reference data from BAM file
\r
928 void BamReader::BamReaderPrivate::LoadReferenceData(void) {
\r
930 // get number of reference sequences
\r
932 mBGZF.Read(buffer, 4);
\r
933 unsigned int numberRefSeqs = BgzfData::UnpackUnsignedInt(buffer);
\r
934 if ( IsBigEndian ) { SwapEndian_32(numberRefSeqs); }
\r
935 if (numberRefSeqs == 0) { return; }
\r
936 References.reserve((int)numberRefSeqs);
\r
938 // iterate over all references in header
\r
939 for (unsigned int i = 0; i != numberRefSeqs; ++i) {
\r
941 // get length of reference name
\r
942 mBGZF.Read(buffer, 4);
\r
943 unsigned int refNameLength = BgzfData::UnpackUnsignedInt(buffer);
\r
944 if ( IsBigEndian ) { SwapEndian_32(refNameLength); }
\r
945 char* refName = (char*)calloc(refNameLength, 1);
\r
947 // get reference name and reference sequence length
\r
948 mBGZF.Read(refName, refNameLength);
\r
949 mBGZF.Read(buffer, 4);
\r
950 int refLength = BgzfData::UnpackSignedInt(buffer);
\r
951 if ( IsBigEndian ) { SwapEndian_32(refLength); }
\r
953 // store data for reference
\r
954 RefData aReference;
\r
955 aReference.RefName = (string)((const char*)refName);
\r
956 aReference.RefLength = refLength;
\r
957 References.push_back(aReference);
\r
959 // clean up calloc-ed temp variable
\r
964 // merges 'alignment chunks' in BAM bin (used for index building)
\r
965 void BamReader::BamReaderPrivate::MergeChunks(void) {
\r
967 // iterate over reference enties
\r
968 BamIndex::iterator indexIter = Index.begin();
\r
969 BamIndex::iterator indexEnd = Index.end();
\r
970 for ( ; indexIter != indexEnd; ++indexIter ) {
\r
972 // get BAM bin map for this reference
\r
973 ReferenceIndex& refIndex = (*indexIter);
\r
974 BamBinMap& bamBinMap = refIndex.Bins;
\r
976 // iterate over BAM bins
\r
977 BamBinMap::iterator binIter = bamBinMap.begin();
\r
978 BamBinMap::iterator binEnd = bamBinMap.end();
\r
979 for ( ; binIter != binEnd; ++binIter ) {
\r
981 // get chunk vector for this bin
\r
982 ChunkVector& binChunks = (*binIter).second;
\r
983 if ( binChunks.size() == 0 ) { continue; }
\r
985 ChunkVector mergedChunks;
\r
986 mergedChunks.push_back( binChunks[0] );
\r
988 // iterate over chunks
\r
990 ChunkVector::iterator chunkIter = binChunks.begin();
\r
991 ChunkVector::iterator chunkEnd = binChunks.end();
\r
992 for ( ++chunkIter; chunkIter != chunkEnd; ++chunkIter) {
\r
994 // get 'currentChunk' based on numeric index
\r
995 Chunk& currentChunk = mergedChunks[i];
\r
997 // get iteratorChunk based on vector iterator
\r
998 Chunk& iteratorChunk = (*chunkIter);
\r
1000 // if currentChunk.Stop(shifted) == iterator Chunk.Start(shifted)
\r
1001 if ( currentChunk.Stop>>16 == iteratorChunk.Start>>16 ) {
\r
1003 // set currentChunk.Stop to iteratorChunk.Stop
\r
1004 currentChunk.Stop = iteratorChunk.Stop;
\r
1009 // set currentChunk + 1 to iteratorChunk
\r
1010 mergedChunks.push_back(iteratorChunk);
\r
1015 // saved merged chunk vector
\r
1016 (*binIter).second = mergedChunks;
\r
1021 // opens BAM file (and index)
\r
1022 void BamReader::BamReaderPrivate::Open(const string& filename, const string& indexFilename) {
\r
1024 Filename = filename;
\r
1025 IndexFilename = indexFilename;
\r
1027 // open the BGZF file for reading, retrieve header text & reference data
\r
1028 mBGZF.Open(filename, "rb");
\r
1030 LoadReferenceData();
\r
1032 // store file offset of first alignment
\r
1033 AlignmentsBeginOffset = mBGZF.Tell();
\r
1035 // open index file & load index data (if exists)
\r
1036 if ( !IndexFilename.empty() ) {
\r
1041 // returns BAM file pointer to beginning of alignment data
\r
1042 bool BamReader::BamReaderPrivate::Rewind(void) {
\r
1044 // find first reference that has alignments in the BAM file
\r
1046 int refCount = References.size();
\r
1047 for ( ; refID < refCount; ++refID ) {
\r
1048 if ( References.at(refID).RefHasAlignments ) { break; }
\r
1051 // store default bounds for first alignment
\r
1052 CurrentRefID = refID;
\r
1054 IsRegionSpecified = false;
\r
1056 // return success/failure of seek
\r
1057 return mBGZF.Seek(AlignmentsBeginOffset);
\r
1060 // saves index data to BAM index file (".bai"), returns success/fail
\r
1061 bool BamReader::BamReaderPrivate::WriteIndex(void) {
\r
1063 IndexFilename = Filename + ".bai";
\r
1064 FILE* indexStream = fopen(IndexFilename.c_str(), "wb");
\r
1065 if ( indexStream == 0 ) {
\r
1066 printf("ERROR: Could not open file to save index\n");
\r
1070 // write BAM index header
\r
1071 fwrite("BAI\1", 1, 4, indexStream);
\r
1073 // write number of reference sequences
\r
1074 int32_t numReferenceSeqs = Index.size();
\r
1075 if ( IsBigEndian ) { SwapEndian_32(numReferenceSeqs); }
\r
1076 fwrite(&numReferenceSeqs, 4, 1, indexStream);
\r
1078 // iterate over reference sequences
\r
1079 BamIndex::const_iterator indexIter = Index.begin();
\r
1080 BamIndex::const_iterator indexEnd = Index.end();
\r
1081 for ( ; indexIter != indexEnd; ++ indexIter ) {
\r
1083 // get reference index data
\r
1084 const ReferenceIndex& refIndex = (*indexIter);
\r
1085 const BamBinMap& binMap = refIndex.Bins;
\r
1086 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
1088 // write number of bins
\r
1089 int32_t binCount = binMap.size();
\r
1090 if ( IsBigEndian ) { SwapEndian_32(binCount); }
\r
1091 fwrite(&binCount, 4, 1, indexStream);
\r
1093 // iterate over bins
\r
1094 BamBinMap::const_iterator binIter = binMap.begin();
\r
1095 BamBinMap::const_iterator binEnd = binMap.end();
\r
1096 for ( ; binIter != binEnd; ++binIter ) {
\r
1098 // get bin data (key and chunk vector)
\r
1099 uint32_t binKey = (*binIter).first;
\r
1100 const ChunkVector& binChunks = (*binIter).second;
\r
1102 // save BAM bin key
\r
1103 if ( IsBigEndian ) { SwapEndian_32(binKey); }
\r
1104 fwrite(&binKey, 4, 1, indexStream);
\r
1106 // save chunk count
\r
1107 int32_t chunkCount = binChunks.size();
\r
1108 if ( IsBigEndian ) { SwapEndian_32(chunkCount); }
\r
1109 fwrite(&chunkCount, 4, 1, indexStream);
\r
1111 // iterate over chunks
\r
1112 ChunkVector::const_iterator chunkIter = binChunks.begin();
\r
1113 ChunkVector::const_iterator chunkEnd = binChunks.end();
\r
1114 for ( ; chunkIter != chunkEnd; ++chunkIter ) {
\r
1116 // get current chunk data
\r
1117 const Chunk& chunk = (*chunkIter);
\r
1118 uint64_t start = chunk.Start;
\r
1119 uint64_t stop = chunk.Stop;
\r
1121 if ( IsBigEndian ) {
\r
1122 SwapEndian_64(start);
\r
1123 SwapEndian_64(stop);
\r
1126 // save chunk offsets
\r
1127 fwrite(&start, 8, 1, indexStream);
\r
1128 fwrite(&stop, 8, 1, indexStream);
\r
1132 // write linear offsets size
\r
1133 int32_t offsetSize = offsets.size();
\r
1134 if ( IsBigEndian ) { SwapEndian_32(offsetSize); }
\r
1135 fwrite(&offsetSize, 4, 1, indexStream);
\r
1137 // iterate over linear offsets
\r
1138 LinearOffsetVector::const_iterator offsetIter = offsets.begin();
\r
1139 LinearOffsetVector::const_iterator offsetEnd = offsets.end();
\r
1140 for ( ; offsetIter != offsetEnd; ++offsetIter ) {
\r
1142 // write linear offset value
\r
1143 uint64_t linearOffset = (*offsetIter);
\r
1144 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
1145 fwrite(&linearOffset, 8, 1, indexStream);
\r
1149 // flush buffer, close file, and return success
\r
1150 fflush(indexStream);
\r
1151 fclose(indexStream);
\r