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: 30 March 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
36 RefVector References;
\r
38 int64_t AlignmentsBeginOffset;
\r
40 string IndexFilename;
\r
44 // user-specified region values
\r
45 bool IsRegionSpecified;
\r
49 // BAM character constants
\r
50 const char* DNA_LOOKUP;
\r
51 const char* CIGAR_LOOKUP;
\r
53 // -------------------------------
\r
54 // constructor & destructor
\r
55 // -------------------------------
\r
56 BamReaderPrivate(void);
\r
57 ~BamReaderPrivate(void);
\r
59 // -------------------------------
\r
60 // "public" interface
\r
61 // -------------------------------
\r
65 bool Jump(int refID, int position = 0);
\r
66 void Open(const string& filename, const string& indexFilename = "");
\r
69 // access alignment data
\r
70 bool GetNextAlignment(BamAlignment& bAlignment);
\r
72 // access auxiliary data
\r
73 int GetReferenceID(const string& refName) const;
\r
76 bool CreateIndex(void);
\r
78 // -------------------------------
\r
80 // -------------------------------
\r
82 // *** reading alignments and auxiliary data *** //
\r
84 // calculate bins that overlap region ( left to reference end for now )
\r
85 int BinsFromRegion(int refID, int left, uint16_t[MAX_BIN]);
\r
86 // calculates alignment end position based on starting position and provided CIGAR operations
\r
87 int CalculateAlignmentEnd(const int& position, const std::vector<CigarOp>& cigarData);
\r
88 // calculate file offset for first alignment chunk overlapping 'left'
\r
89 int64_t GetOffset(int refID, int left);
\r
90 // checks to see if alignment overlaps current region
\r
91 bool IsOverlap(BamAlignment& bAlignment);
\r
92 // retrieves header text from BAM file
\r
93 void LoadHeaderData(void);
\r
94 // retrieves BAM alignment under file pointer
\r
95 bool LoadNextAlignment(BamAlignment& bAlignment);
\r
96 // builds reference data structure from BAM file
\r
97 void LoadReferenceData(void);
\r
99 // *** index file handling *** //
\r
101 // calculates index for BAM file
\r
102 bool BuildIndex(void);
\r
103 // clear out inernal index data structure
\r
104 void ClearIndex(void);
\r
105 // saves BAM bin entry for index
\r
106 void InsertBinEntry(BamBinMap& binMap, const uint32_t& saveBin, const uint64_t& saveOffset, const uint64_t& lastOffset);
\r
107 // saves linear offset entry for index
\r
108 void InsertLinearOffset(LinearOffsetVector& offsets, const BamAlignment& bAlignment, const uint64_t& lastOffset);
\r
109 // loads index from BAM index file
\r
110 bool LoadIndex(void);
\r
111 // simplifies index by merging 'chunks'
\r
112 void MergeChunks(void);
\r
113 // round-up 32-bit integer to next power-of-2
\r
114 void Roundup32(int& value);
\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
143 // access auxiliary data
\r
144 const string BamReader::GetHeaderText(void) const { return d->HeaderText; }
\r
145 int BamReader::GetReferenceCount(void) const { return d->References.size(); }
\r
146 const RefVector BamReader::GetReferenceData(void) const { return d->References; }
\r
147 int BamReader::GetReferenceID(const string& refName) const { return d->GetReferenceID(refName); }
\r
149 // index operations
\r
150 bool BamReader::CreateIndex(void) { return d->CreateIndex(); }
\r
152 // -----------------------------------------------------
\r
153 // BamReaderPrivate implementation
\r
154 // -----------------------------------------------------
\r
157 BamReader::BamReaderPrivate::BamReaderPrivate(void)
\r
158 : IsIndexLoaded(false)
\r
159 , AlignmentsBeginOffset(0)
\r
160 , IsRegionSpecified(false)
\r
163 , DNA_LOOKUP("=ACMGRSVTWYHKDBN")
\r
164 , CIGAR_LOOKUP("MIDNSHP")
\r
166 IsBigEndian = SystemIsBigEndian();
\r
170 BamReader::BamReaderPrivate::~BamReaderPrivate(void) {
\r
174 // calculate bins that overlap region ( left to reference end for now )
\r
175 int BamReader::BamReaderPrivate::BinsFromRegion(int refID, int left, uint16_t list[MAX_BIN]) {
\r
177 // get region boundaries
\r
178 uint32_t begin = (unsigned int)left;
\r
179 uint32_t end = (unsigned int)References.at(refID).RefLength - 1;
\r
181 // initialize list, bin '0' always a valid bin
\r
185 // get rest of bins that contain this region
\r
187 for (k = 1 + (begin>>26); k <= 1 + (end>>26); ++k) { list[i++] = k; }
\r
188 for (k = 9 + (begin>>23); k <= 9 + (end>>23); ++k) { list[i++] = k; }
\r
189 for (k = 73 + (begin>>20); k <= 73 + (end>>20); ++k) { list[i++] = k; }
\r
190 for (k = 585 + (begin>>17); k <= 585 + (end>>17); ++k) { list[i++] = k; }
\r
191 for (k = 4681 + (begin>>14); k <= 4681 + (end>>14); ++k) { list[i++] = k; }
\r
193 // return number of bins stored
\r
197 // populates BAM index data structure from BAM file data
\r
198 bool BamReader::BamReaderPrivate::BuildIndex(void) {
\r
200 // check to be sure file is open
\r
201 if (!mBGZF.IsOpen) { return false; }
\r
203 // move file pointer to beginning of alignments
\r
206 // get reference count, reserve index space
\r
207 int numReferences = References.size();
\r
208 for ( int i = 0; i < numReferences; ++i ) {
\r
209 Index.push_back(ReferenceIndex());
\r
212 // sets default constant for bin, ID, offset, coordinate variables
\r
213 const uint32_t defaultValue = 0xffffffffu;
\r
216 uint32_t saveBin(defaultValue);
\r
217 uint32_t lastBin(defaultValue);
\r
219 // reference ID data
\r
220 int32_t saveRefID(defaultValue);
\r
221 int32_t lastRefID(defaultValue);
\r
224 uint64_t saveOffset = mBGZF.Tell();
\r
225 uint64_t lastOffset = saveOffset;
\r
228 int32_t lastCoordinate = defaultValue;
\r
230 BamAlignment bAlignment;
\r
231 while( GetNextAlignment(bAlignment) ) {
\r
233 // change of chromosome, save ID, reset bin
\r
234 if ( lastRefID != bAlignment.RefID ) {
\r
235 lastRefID = bAlignment.RefID;
\r
236 lastBin = defaultValue;
\r
239 // if lastCoordinate greater than BAM position - file not sorted properly
\r
240 else if ( lastCoordinate > bAlignment.Position ) {
\r
241 printf("BAM file not properly sorted:\n");
\r
242 printf("Alignment %s : %d > %d on reference (id = %d)", bAlignment.Name.c_str(), lastCoordinate, bAlignment.Position, bAlignment.RefID);
\r
246 // if valid reference && BAM bin spans some minimum cutoff (smaller bin ids span larger regions)
\r
247 if ( (bAlignment.RefID >= 0) && (bAlignment.Bin < 4681) ) {
\r
249 // save linear offset entry (matched to BAM entry refID)
\r
250 ReferenceIndex& refIndex = Index.at(bAlignment.RefID);
\r
251 LinearOffsetVector& offsets = refIndex.Offsets;
\r
252 InsertLinearOffset(offsets, bAlignment, lastOffset);
\r
255 // if current BamAlignment bin != lastBin, "then possibly write the binning index"
\r
256 if ( bAlignment.Bin != lastBin ) {
\r
258 // if not first time through
\r
259 if ( saveBin != defaultValue ) {
\r
261 // save Bam bin entry
\r
262 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
263 BamBinMap& binMap = refIndex.Bins;
\r
264 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
267 // update saveOffset
\r
268 saveOffset = lastOffset;
\r
270 // update bin values
\r
271 saveBin = bAlignment.Bin;
\r
272 lastBin = bAlignment.Bin;
\r
274 // update saveRefID
\r
275 saveRefID = bAlignment.RefID;
\r
277 // if invalid RefID, break out (why?)
\r
278 if ( saveRefID < 0 ) { break; }
\r
281 // make sure that current file pointer is beyond lastOffset
\r
282 if ( mBGZF.Tell() <= (int64_t)lastOffset ) {
\r
283 printf("Error in BGZF offsets.\n");
\r
287 // update lastOffset
\r
288 lastOffset = mBGZF.Tell();
\r
290 // update lastCoordinate
\r
291 lastCoordinate = bAlignment.Position;
\r
294 // save any leftover BAM data (as long as refID is valid)
\r
295 if ( saveRefID >= 0 ) {
\r
296 // save Bam bin entry
\r
297 ReferenceIndex& refIndex = Index.at(saveRefID);
\r
298 BamBinMap& binMap = refIndex.Bins;
\r
299 InsertBinEntry(binMap, saveBin, saveOffset, lastOffset);
\r
302 // simplify index by merging chunks
\r
305 // iterate over references
\r
306 BamIndex::iterator indexIter = Index.begin();
\r
307 BamIndex::iterator indexEnd = Index.end();
\r
308 for ( int i = 0; indexIter != indexEnd; ++indexIter, ++i ) {
\r
310 // get reference index data
\r
311 ReferenceIndex& refIndex = (*indexIter);
\r
312 BamBinMap& binMap = refIndex.Bins;
\r
313 LinearOffsetVector& offsets = refIndex.Offsets;
\r
315 // store whether reference has alignments or no
\r
316 References[i].RefHasAlignments = ( binMap.size() > 0 );
\r
318 // sort linear offsets
\r
319 sort(offsets.begin(), offsets.end());
\r
323 // rewind file pointer to beginning of alignments, return success/fail
\r
327 // calculates alignment end position based on starting position and provided CIGAR operations
\r
328 int BamReader::BamReaderPrivate::CalculateAlignmentEnd(const int& position, const vector<CigarOp>& cigarData) {
\r
330 // initialize alignment end to starting position
\r
331 int alignEnd = position;
\r
333 // iterate over cigar operations
\r
334 vector<CigarOp>::const_iterator cigarIter = cigarData.begin();
\r
335 vector<CigarOp>::const_iterator cigarEnd = cigarData.end();
\r
336 for ( ; cigarIter != cigarEnd; ++cigarIter) {
\r
337 char cigarType = (*cigarIter).Type;
\r
338 if ( cigarType == 'M' || cigarType == 'D' || cigarType == 'N' ) {
\r
339 alignEnd += (*cigarIter).Length;
\r
346 // clear index data structure
\r
347 void BamReader::BamReaderPrivate::ClearIndex(void) {
\r
348 Index.clear(); // sufficient ??
\r
351 // closes the BAM file
\r
352 void BamReader::BamReaderPrivate::Close(void) {
\r
355 HeaderText.clear();
\r
356 IsRegionSpecified = false;
\r
359 // create BAM index from BAM file (keep structure in memory) and write to default index output file
\r
360 bool BamReader::BamReaderPrivate::CreateIndex(void) {
\r
365 // build (& save) index from BAM file
\r
367 ok &= BuildIndex();
\r
368 ok &= WriteIndex();
\r
370 // return success/fail
\r
374 // returns RefID for given RefName (returns References.size() if not found)
\r
375 int BamReader::BamReaderPrivate::GetReferenceID(const string& refName) const {
\r
377 // retrieve names from reference data
\r
378 vector<string> refNames;
\r
379 RefVector::const_iterator refIter = References.begin();
\r
380 RefVector::const_iterator refEnd = References.end();
\r
381 for ( ; refIter != refEnd; ++refIter) {
\r
382 refNames.push_back( (*refIter).RefName );
\r
385 // return 'index-of' refName ( if not found, returns refNames.size() )
\r
386 return distance(refNames.begin(), find(refNames.begin(), refNames.end(), refName));
\r
389 // get next alignment (from specified region, if given)
\r
390 bool BamReader::BamReaderPrivate::GetNextAlignment(BamAlignment& bAlignment) {
\r
392 // if valid alignment available
\r
393 if ( LoadNextAlignment(bAlignment) ) {
\r
395 // if region not specified, return success
\r
396 if ( !IsRegionSpecified ) { return true; }
\r
398 // load next alignment until region overlap is found
\r
399 while ( !IsOverlap(bAlignment) ) {
\r
400 // if no valid alignment available (likely EOF) return failure
\r
401 if ( !LoadNextAlignment(bAlignment) ) { return false; }
\r
404 // return success (alignment found that overlaps region)
\r
408 // no valid alignment
\r
409 else { return false; }
\r
412 // calculate closest indexed file offset for region specified
\r
413 int64_t BamReader::BamReaderPrivate::GetOffset(int refID, int left) {
\r
415 // calculate which bins overlap this region
\r
416 uint16_t* bins = (uint16_t*)calloc(MAX_BIN, 2);
\r
417 int numBins = BinsFromRegion(refID, left, bins);
\r
419 // get bins for this reference
\r
420 const ReferenceIndex& refIndex = Index.at(refID);
\r
421 const BamBinMap& binMap = refIndex.Bins;
\r
423 // get minimum offset to consider
\r
424 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
425 uint64_t minOffset = ( (unsigned int)(left>>BAM_LIDX_SHIFT) >= offsets.size() ) ? 0 : offsets.at(left>>BAM_LIDX_SHIFT);
\r
427 // store offsets to beginning of alignment 'chunks'
\r
428 std::vector<int64_t> chunkStarts;
\r
430 // store all alignment 'chunk' starts for bins in this region
\r
431 for (int i = 0; i < numBins; ++i ) {
\r
432 uint16_t binKey = bins[i];
\r
434 map<uint32_t, ChunkVector>::const_iterator binIter = binMap.find(binKey);
\r
435 if ( (binIter != binMap.end()) && ((*binIter).first == binKey) ) {
\r
437 const ChunkVector& chunks = (*binIter).second;
\r
438 std::vector<Chunk>::const_iterator chunksIter = chunks.begin();
\r
439 std::vector<Chunk>::const_iterator chunksEnd = chunks.end();
\r
440 for ( ; chunksIter != chunksEnd; ++chunksIter) {
\r
441 const Chunk& chunk = (*chunksIter);
\r
442 if ( chunk.Stop > minOffset ) {
\r
443 chunkStarts.push_back( chunk.Start );
\r
452 // if no alignments found, else return smallest offset for alignment starts
\r
453 if ( chunkStarts.size() == 0 ) { return -1; }
\r
454 else { return *min_element(chunkStarts.begin(), chunkStarts.end()); }
\r
457 // saves BAM bin entry for index
\r
458 void BamReader::BamReaderPrivate::InsertBinEntry(BamBinMap& binMap,
\r
459 const uint32_t& saveBin,
\r
460 const uint64_t& saveOffset,
\r
461 const uint64_t& lastOffset)
\r
464 BamBinMap::iterator binIter = binMap.find(saveBin);
\r
466 // create new chunk
\r
467 Chunk newChunk(saveOffset, lastOffset);
\r
469 // if entry doesn't exist
\r
470 if ( binIter == binMap.end() ) {
\r
471 ChunkVector newChunks;
\r
472 newChunks.push_back(newChunk);
\r
473 binMap.insert( pair<uint32_t, ChunkVector>(saveBin, newChunks));
\r
478 ChunkVector& binChunks = (*binIter).second;
\r
479 binChunks.push_back( newChunk );
\r
483 // saves linear offset entry for index
\r
484 void BamReader::BamReaderPrivate::InsertLinearOffset(LinearOffsetVector& offsets,
\r
485 const BamAlignment& bAlignment,
\r
486 const uint64_t& lastOffset)
\r
488 // get converted offsets
\r
489 int beginOffset = bAlignment.Position >> BAM_LIDX_SHIFT;
\r
490 int endOffset = ( CalculateAlignmentEnd(bAlignment.Position, bAlignment.CigarData) - 1) >> BAM_LIDX_SHIFT;
\r
492 // resize vector if necessary
\r
493 int oldSize = offsets.size();
\r
494 int newSize = endOffset + 1;
\r
495 if ( oldSize < newSize ) {
\r
496 Roundup32(newSize);
\r
497 offsets.resize(newSize, 0);
\r
501 for(int i = beginOffset + 1; i <= endOffset ; ++i) {
\r
502 if ( offsets[i] == 0) {
\r
503 offsets[i] = lastOffset;
\r
508 // returns whether alignment overlaps currently specified region (refID, leftBound)
\r
509 bool BamReader::BamReaderPrivate::IsOverlap(BamAlignment& bAlignment) {
\r
511 // if on different reference sequence, quit
\r
512 if ( bAlignment.RefID != CurrentRefID ) { return false; }
\r
514 // read starts after left boundary
\r
515 if ( bAlignment.Position >= CurrentLeft) { return true; }
\r
517 // return whether alignment end overlaps left boundary
\r
518 return ( CalculateAlignmentEnd(bAlignment.Position, bAlignment.CigarData) >= CurrentLeft );
\r
521 // jumps to specified region(refID, leftBound) in BAM file, returns success/fail
\r
522 bool BamReader::BamReaderPrivate::Jump(int refID, int position) {
\r
524 // if data exists for this reference and position is valid
\r
525 if ( References.at(refID).RefHasAlignments && (position <= References.at(refID).RefLength) ) {
\r
527 // set current region
\r
528 CurrentRefID = refID;
\r
529 CurrentLeft = position;
\r
530 IsRegionSpecified = true;
\r
532 // calculate offset
\r
533 int64_t offset = GetOffset(CurrentRefID, CurrentLeft);
\r
535 // if in valid offset, return failure
\r
536 if ( offset == -1 ) { return false; }
\r
538 // otherwise return success of seek operation
\r
539 else { return mBGZF.Seek(offset); }
\r
542 // invalid jump request parameters, return failure
\r
546 // load BAM header data
\r
547 void BamReader::BamReaderPrivate::LoadHeaderData(void) {
\r
549 // check to see if proper BAM header
\r
551 if (mBGZF.Read(buffer, 4) != 4) {
\r
552 printf("Could not read header type\n");
\r
556 if (strncmp(buffer, "BAM\001", 4)) {
\r
557 printf("wrong header type!\n");
\r
561 // get BAM header text length
\r
562 mBGZF.Read(buffer, 4);
\r
563 unsigned int headerTextLength = BgzfData::UnpackUnsignedInt(buffer);
\r
564 if ( IsBigEndian ) { SwapEndian_32(headerTextLength); }
\r
566 // get BAM header text
\r
567 char* headerText = (char*)calloc(headerTextLength + 1, 1);
\r
568 mBGZF.Read(headerText, headerTextLength);
\r
569 HeaderText = (string)((const char*)headerText);
\r
571 // clean up calloc-ed temp variable
\r
575 // load existing index data from BAM index file (".bai"), return success/fail
\r
576 bool BamReader::BamReaderPrivate::LoadIndex(void) {
\r
578 // clear out index data
\r
581 // skip if index file empty
\r
582 if ( IndexFilename.empty() ) { return false; }
\r
584 // open index file, abort on error
\r
585 FILE* indexStream = fopen(IndexFilename.c_str(), "rb");
\r
587 printf("ERROR: Unable to open the BAM index file %s for reading.\n", IndexFilename.c_str() );
\r
591 size_t elementsRead = 0;
\r
593 // see if index is valid BAM index
\r
595 elementsRead = fread(magic, 1, 4, indexStream);
\r
596 if (strncmp(magic, "BAI\1", 4)) {
\r
597 printf("Problem with index file - invalid format.\n");
\r
598 fclose(indexStream);
\r
602 // get number of reference sequences
\r
603 uint32_t numRefSeqs;
\r
604 elementsRead = fread(&numRefSeqs, 4, 1, indexStream);
\r
605 if ( IsBigEndian ) { SwapEndian_32(numRefSeqs); }
\r
607 // intialize space for BamIndex data structure
\r
608 Index.reserve(numRefSeqs);
\r
610 // iterate over reference sequences
\r
611 for (unsigned int i = 0; i < numRefSeqs; ++i) {
\r
613 // get number of bins for this reference sequence
\r
615 elementsRead = fread(&numBins, 4, 1, indexStream);
\r
616 if ( IsBigEndian ) { SwapEndian_32(numBins); }
\r
619 RefData& refEntry = References[i];
\r
620 refEntry.RefHasAlignments = true;
\r
623 // intialize BinVector
\r
626 // iterate over bins for that reference sequence
\r
627 for (int j = 0; j < numBins; ++j) {
\r
631 elementsRead = fread(&binID, 4, 1, indexStream);
\r
633 // get number of regionChunks in this bin
\r
634 uint32_t numChunks;
\r
635 elementsRead = fread(&numChunks, 4, 1, indexStream);
\r
637 if ( IsBigEndian ) {
\r
638 SwapEndian_32(binID);
\r
639 SwapEndian_32(numChunks);
\r
642 // intialize ChunkVector
\r
643 ChunkVector regionChunks;
\r
644 regionChunks.reserve(numChunks);
\r
646 // iterate over regionChunks in this bin
\r
647 for (unsigned int k = 0; k < numChunks; ++k) {
\r
649 // get chunk boundaries (left, right)
\r
652 elementsRead = fread(&left, 8, 1, indexStream);
\r
653 elementsRead = fread(&right, 8, 1, indexStream);
\r
655 if ( IsBigEndian ) {
\r
656 SwapEndian_64(left);
\r
657 SwapEndian_64(right);
\r
661 regionChunks.push_back( Chunk(left, right) );
\r
664 // sort chunks for this bin
\r
665 sort( regionChunks.begin(), regionChunks.end(), ChunkLessThan );
\r
667 // save binID, chunkVector for this bin
\r
668 binMap.insert( pair<uint32_t, ChunkVector>(binID, regionChunks) );
\r
671 // load linear index for this reference sequence
\r
673 // get number of linear offsets
\r
674 int32_t numLinearOffsets;
\r
675 elementsRead = fread(&numLinearOffsets, 4, 1, indexStream);
\r
676 if ( IsBigEndian ) { SwapEndian_32(numLinearOffsets); }
\r
678 // intialize LinearOffsetVector
\r
679 LinearOffsetVector offsets;
\r
680 offsets.reserve(numLinearOffsets);
\r
682 // iterate over linear offsets for this reference sequeence
\r
683 uint64_t linearOffset;
\r
684 for (int j = 0; j < numLinearOffsets; ++j) {
\r
685 // read a linear offset & store
\r
686 elementsRead = fread(&linearOffset, 8, 1, indexStream);
\r
687 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
688 offsets.push_back(linearOffset);
\r
691 // sort linear offsets
\r
692 sort( offsets.begin(), offsets.end() );
\r
694 // store index data for that reference sequence
\r
695 Index.push_back( ReferenceIndex(binMap, offsets) );
\r
698 // close index file (.bai) and return
\r
699 fclose(indexStream);
\r
703 // populates BamAlignment with alignment data under file pointer, returns success/fail
\r
704 bool BamReader::BamReaderPrivate::LoadNextAlignment(BamAlignment& bAlignment) {
\r
706 // read in the 'block length' value, make sure it's not zero
\r
708 mBGZF.Read(buffer, 4);
\r
709 unsigned int blockLength = BgzfData::UnpackUnsignedInt(buffer);
\r
710 if ( IsBigEndian ) { SwapEndian_32(blockLength); }
\r
711 if ( blockLength == 0 ) { return false; }
\r
713 // keep track of bytes read as method progresses
\r
716 // read in core alignment data, make sure the right size of data was read
\r
717 char x[BAM_CORE_SIZE];
\r
718 if ( mBGZF.Read(x, BAM_CORE_SIZE) != BAM_CORE_SIZE ) { return false; }
\r
719 bytesRead += BAM_CORE_SIZE;
\r
721 if ( IsBigEndian ) {
\r
722 for ( int i = 0; i < BAM_CORE_SIZE; i+=sizeof(uint32_t) ) {
\r
723 SwapEndian_32p(&x[i]);
\r
727 // set BamAlignment 'core' data and character data lengths
\r
728 unsigned int tempValue;
\r
729 unsigned int queryNameLength;
\r
730 unsigned int numCigarOperations;
\r
731 unsigned int querySequenceLength;
\r
733 bAlignment.RefID = BgzfData::UnpackSignedInt(&x[0]);
\r
734 bAlignment.Position = BgzfData::UnpackSignedInt(&x[4]);
\r
736 tempValue = BgzfData::UnpackUnsignedInt(&x[8]);
\r
737 bAlignment.Bin = tempValue >> 16;
\r
738 bAlignment.MapQuality = tempValue >> 8 & 0xff;
\r
739 queryNameLength = tempValue & 0xff;
\r
741 tempValue = BgzfData::UnpackUnsignedInt(&x[12]);
\r
742 bAlignment.AlignmentFlag = tempValue >> 16;
\r
743 numCigarOperations = tempValue & 0xffff;
\r
745 querySequenceLength = BgzfData::UnpackUnsignedInt(&x[16]);
\r
746 bAlignment.MateRefID = BgzfData::UnpackSignedInt(&x[20]);
\r
747 bAlignment.MatePosition = BgzfData::UnpackSignedInt(&x[24]);
\r
748 bAlignment.InsertSize = BgzfData::UnpackSignedInt(&x[28]);
\r
750 // calculate lengths/offsets
\r
751 const unsigned int dataLength = blockLength - BAM_CORE_SIZE;
\r
752 const unsigned int cigarDataOffset = queryNameLength;
\r
753 const unsigned int seqDataOffset = cigarDataOffset + (numCigarOperations * 4);
\r
754 const unsigned int qualDataOffset = seqDataOffset + (querySequenceLength+1)/2;
\r
755 const unsigned int tagDataOffset = qualDataOffset + querySequenceLength;
\r
756 const unsigned int tagDataLen = dataLength - tagDataOffset;
\r
758 // set up destination buffers for character data
\r
759 char* allCharData = (char*)calloc(sizeof(char), dataLength);
\r
760 uint32_t* cigarData = (uint32_t*)(allCharData + cigarDataOffset);
\r
761 char* seqData = ((char*)allCharData) + seqDataOffset;
\r
762 char* qualData = ((char*)allCharData) + qualDataOffset;
\r
763 char* tagData = ((char*)allCharData) + tagDataOffset;
\r
765 // get character data - make sure proper data size was read
\r
766 if ( mBGZF.Read(allCharData, dataLength) != (signed int)dataLength) { return false; }
\r
769 bytesRead += dataLength;
\r
771 // clear out any previous string data
\r
772 bAlignment.Name.clear();
\r
773 bAlignment.QueryBases.clear();
\r
774 bAlignment.Qualities.clear();
\r
775 bAlignment.AlignedBases.clear();
\r
776 bAlignment.CigarData.clear();
\r
777 bAlignment.TagData.clear();
\r
780 bAlignment.Name = (string)((const char*)(allCharData));
\r
782 // save query sequence
\r
783 // -----------------------
\r
784 // Added: 3-25-2010 DWB
\r
785 // Improved: reduced repeated memory allocations as string grows
\r
786 bAlignment.QueryBases.reserve(querySequenceLength);
\r
787 // -----------------------
\r
789 for (unsigned int i = 0; i < querySequenceLength; ++i) {
\r
790 char singleBase = DNA_LOOKUP[ ( ( seqData[(i/2)] >> (4*(1-(i%2)))) & 0xf ) ];
\r
791 bAlignment.QueryBases.append( 1, singleBase );
\r
794 // save sequence length
\r
795 bAlignment.Length = bAlignment.QueryBases.length();
\r
797 // save qualities, convert from numeric QV to FASTQ character
\r
798 // -----------------------
\r
799 // Added: 3-25-2010 DWB
\r
800 // Improved: reduced repeated memory allocations as string grows
\r
801 bAlignment.Qualities.reserve(querySequenceLength);
\r
802 // -----------------------
\r
804 for (unsigned int i = 0; i < querySequenceLength; ++i) {
\r
805 char singleQuality = (char)(qualData[i]+33);
\r
806 bAlignment.Qualities.append( 1, singleQuality );
\r
809 // save CIGAR-related data;
\r
810 // -----------------------
\r
811 // Added: 3-25-2010 DWB
\r
812 // Improved: reduced repeated memory allocations as string grows
\r
813 bAlignment.AlignedBases.reserve(querySequenceLength);
\r
814 // -----------------------
\r
817 for (unsigned int i = 0; i < numCigarOperations; ++i) {
\r
819 if ( IsBigEndian ) { SwapEndian_32(cigarData[i]); }
\r
821 // build CigarOp struct
\r
823 op.Length = (cigarData[i] >> BAM_CIGAR_SHIFT);
\r
824 op.Type = CIGAR_LOOKUP[ (cigarData[i] & BAM_CIGAR_MASK) ];
\r
827 bAlignment.CigarData.push_back(op);
\r
829 // build AlignedBases string
\r
834 bAlignment.AlignedBases.append( bAlignment.QueryBases.substr(k, op.Length) ); // for 'M', 'I' - write bases
\r
838 k += op.Length; // for 'S' - skip over query bases
\r
842 bAlignment.AlignedBases.append( op.Length, '-' ); // for 'D' - write gap character
\r
846 bAlignment.AlignedBases.append( op.Length, '*' ); // for 'P' - write padding character;
\r
850 bAlignment.AlignedBases.append( op.Length, 'N' ); // for 'N' - write N's, skip bases in query sequence
\r
851 // -----------------------
\r
852 // Removed: 3-25-2010 DWB
\r
853 // Contributed: ARQ
\r
854 // Fixed: compliance with actual 'N' definition in BAM spec
\r
856 // -----------------------
\r
860 break; // for 'H' - do nothing, move to next op
\r
863 printf("ERROR: Invalid Cigar op type\n"); // shouldn't get here
\r
869 // -----------------------
\r
870 // Added: 3-25-2010 DWB
\r
871 // Fixed: endian-correctness for tag data
\r
872 // -----------------------
\r
873 if ( IsBigEndian ) {
\r
875 while ( (unsigned int)i < tagDataLen ) {
\r
877 i += 2; // skip tag type (e.g. "RG", "NM", etc)
\r
878 uint8_t type = toupper(tagData[i]); // lower & upper case letters have same meaning
\r
879 ++i; // skip value type
\r
889 SwapEndian_16p(&tagData[i]);
\r
890 i+=2; // sizeof(uint16_t)
\r
895 SwapEndian_32p(&tagData[i]);
\r
896 i+=4; // sizeof(uint32_t)
\r
900 SwapEndian_64p(&tagData[i]);
\r
901 i+=8; // sizeof(uint64_t)
\r
906 while (tagData[i]) { ++i; }
\r
907 ++i; // increment one more for null terminator
\r
911 printf("ERROR: Invalid tag value type\n"); // shouldn't get here
\r
919 bAlignment.TagData.resize(tagDataLen);
\r
920 memcpy((char*)bAlignment.TagData.data(), tagData, tagDataLen);
\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 // rounds value up to next power-of-2 (used in index building)
\r
1061 void BamReader::BamReaderPrivate::Roundup32(int& value) {
\r
1063 value |= value >> 1;
\r
1064 value |= value >> 2;
\r
1065 value |= value >> 4;
\r
1066 value |= value >> 8;
\r
1067 value |= value >> 16;
\r
1071 // saves index data to BAM index file (".bai"), returns success/fail
\r
1072 bool BamReader::BamReaderPrivate::WriteIndex(void) {
\r
1074 IndexFilename = Filename + ".bai";
\r
1075 FILE* indexStream = fopen(IndexFilename.c_str(), "wb");
\r
1076 if ( indexStream == 0 ) {
\r
1077 printf("ERROR: Could not open file to save index\n");
\r
1081 // write BAM index header
\r
1082 fwrite("BAI\1", 1, 4, indexStream);
\r
1084 // write number of reference sequences
\r
1085 int32_t numReferenceSeqs = Index.size();
\r
1086 if ( IsBigEndian ) { SwapEndian_32(numReferenceSeqs); }
\r
1087 fwrite(&numReferenceSeqs, 4, 1, indexStream);
\r
1089 // iterate over reference sequences
\r
1090 BamIndex::const_iterator indexIter = Index.begin();
\r
1091 BamIndex::const_iterator indexEnd = Index.end();
\r
1092 for ( ; indexIter != indexEnd; ++ indexIter ) {
\r
1094 // get reference index data
\r
1095 const ReferenceIndex& refIndex = (*indexIter);
\r
1096 const BamBinMap& binMap = refIndex.Bins;
\r
1097 const LinearOffsetVector& offsets = refIndex.Offsets;
\r
1099 // write number of bins
\r
1100 int32_t binCount = binMap.size();
\r
1101 if ( IsBigEndian ) { SwapEndian_32(binCount); }
\r
1102 fwrite(&binCount, 4, 1, indexStream);
\r
1104 // iterate over bins
\r
1105 BamBinMap::const_iterator binIter = binMap.begin();
\r
1106 BamBinMap::const_iterator binEnd = binMap.end();
\r
1107 for ( ; binIter != binEnd; ++binIter ) {
\r
1109 // get bin data (key and chunk vector)
\r
1110 uint32_t binKey = (*binIter).first;
\r
1111 const ChunkVector& binChunks = (*binIter).second;
\r
1113 // save BAM bin key
\r
1114 if ( IsBigEndian ) { SwapEndian_32(binKey); }
\r
1115 fwrite(&binKey, 4, 1, indexStream);
\r
1117 // save chunk count
\r
1118 int32_t chunkCount = binChunks.size();
\r
1119 if ( IsBigEndian ) { SwapEndian_32(chunkCount); }
\r
1120 fwrite(&chunkCount, 4, 1, indexStream);
\r
1122 // iterate over chunks
\r
1123 ChunkVector::const_iterator chunkIter = binChunks.begin();
\r
1124 ChunkVector::const_iterator chunkEnd = binChunks.end();
\r
1125 for ( ; chunkIter != chunkEnd; ++chunkIter ) {
\r
1127 // get current chunk data
\r
1128 const Chunk& chunk = (*chunkIter);
\r
1129 uint64_t start = chunk.Start;
\r
1130 uint64_t stop = chunk.Stop;
\r
1132 if ( IsBigEndian ) {
\r
1133 SwapEndian_64(start);
\r
1134 SwapEndian_64(stop);
\r
1137 // save chunk offsets
\r
1138 fwrite(&start, 8, 1, indexStream);
\r
1139 fwrite(&stop, 8, 1, indexStream);
\r
1143 // write linear offsets size
\r
1144 int32_t offsetSize = offsets.size();
\r
1145 if ( IsBigEndian ) { SwapEndian_32(offsetSize); }
\r
1146 fwrite(&offsetSize, 4, 1, indexStream);
\r
1148 // iterate over linear offsets
\r
1149 LinearOffsetVector::const_iterator offsetIter = offsets.begin();
\r
1150 LinearOffsetVector::const_iterator offsetEnd = offsets.end();
\r
1151 for ( ; offsetIter != offsetEnd; ++offsetIter ) {
\r
1153 // write linear offset value
\r
1154 uint64_t linearOffset = (*offsetIter);
\r
1155 if ( IsBigEndian ) { SwapEndian_64(linearOffset); }
\r
1156 fwrite(&linearOffset, 8, 1, indexStream);
\r
1160 // flush buffer, close file, and return success
\r
1161 fflush(indexStream);
\r
1162 fclose(indexStream);
\r