// ***************************************************************************
// BamMultiReader_p.cpp (c) 2010 Derek Barnett, Erik Garrison
// Marth Lab, Department of Biology, Boston College
-// All rights reserved.
// ---------------------------------------------------------------------------
-// Last modified: 17 January 2011 (DB)
+// Last modified: 3 October 2011 (DB)
// ---------------------------------------------------------------------------
// Functionality for simultaneously reading multiple BAM files
// *************************************************************************
#include <api/BamAlignment.h>
#include <api/BamMultiReader.h>
-#include <api/internal/BamMultiMerger_p.h>
+#include <api/SamConstants.h>
+#include <api/algorithms/Sort.h>
#include <api/internal/BamMultiReader_p.h>
using namespace BamTools;
using namespace BamTools::Internal;
// ctor
BamMultiReaderPrivate::BamMultiReaderPrivate(void)
- : m_alignments(0)
- , m_isCoreMode(false)
- , m_sortOrder(BamMultiReader::SortedByPosition)
+ : m_alignmentCache(0)
{ }
// dtor
BamMultiReaderPrivate::~BamMultiReaderPrivate(void) {
- // close all open BAM readers
+ // close all open BAM readers (& clean up cache)
Close();
-
- // clean up alignment cache
- delete m_alignments;
- m_alignments = 0;
}
-// close the BAM files
+// close all BAM files
void BamMultiReaderPrivate::Close(void) {
+ CloseFiles( Filenames() );
+}
- // clear out alignment cache
- m_alignments->Clear();
-
- // iterate over readers
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
- for ( ; readerIter != readerEnd; ++readerIter ) {
+// close requested BAM file
+void BamMultiReaderPrivate::CloseFile(const string& filename) {
+ vector<string> filenames(1, filename);
+ CloseFiles(filenames);
+}
- // close reader
- BamReader* reader = (*readerIter).first;
- BamAlignment* alignment = (*readerIter).second;
- if ( reader ) reader->Close();
+// close requested BAM files
+void BamMultiReaderPrivate::CloseFiles(const vector<string>& filenames) {
- // delete pointers
- delete reader;
- reader = 0;
- delete alignment;
- alignment = 0;
+ // iterate over filenames
+ vector<string>::const_iterator filesIter = filenames.begin();
+ vector<string>::const_iterator filesEnd = filenames.end();
+ for ( ; filesIter != filesEnd; ++filesIter ) {
+ const string& filename = (*filesIter);
+ if ( filename.empty() ) continue;
+
+ // iterate over readers
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
+ for ( ; readerIter != readerEnd; ++readerIter ) {
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
+
+ // if reader matches requested filename
+ if ( reader->GetFilename() == filename ) {
+
+ // remove reader's entry from alignment cache
+ m_alignmentCache->Remove(reader);
+
+ // clean up reader & its alignment
+ reader->Close();
+ delete reader;
+ reader = 0;
+
+ // delete reader's alignment entry
+ BamAlignment* alignment = item.Alignment;
+ delete alignment;
+ alignment = 0;
+
+ // remove reader from reader list
+ m_readers.erase(readerIter);
+
+ // on match, just go on to next filename
+ // (no need to keep looking and item iterator is invalid now anyway)
+ break;
+ }
+ }
}
- // clear out readers
- m_readers.clear();
-
- // reset default flags
- m_isCoreMode = false;
- m_sortOrder = BamMultiReader::SortedByPosition;
+ // make sure alignment cache is cleaned up if all readers closed
+ if ( m_readers.empty() && m_alignmentCache ) {
+ m_alignmentCache->Clear();
+ delete m_alignmentCache;
+ m_alignmentCache = 0;
+ }
}
-// saves index data to BAM index files (".bai"/".bti") where necessary, returns success/fail
-bool BamMultiReaderPrivate::CreateIndexes(bool useStandardIndex) {
+// creates index files for BAM files that don't have them
+bool BamMultiReaderPrivate::CreateIndexes(const BamIndex::IndexType& type) {
bool result = true;
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
- for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
- result &= reader->CreateIndex(useStandardIndex);
+
+ // iterate over readers
+ vector<MergeItem>::iterator itemIter = m_readers.begin();
+ vector<MergeItem>::iterator itemEnd = m_readers.end();
+ for ( ; itemIter != itemEnd; ++itemIter ) {
+ MergeItem& item = (*itemIter);
+ BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
+
+ // if reader doesn't have an index, create one
+ if ( !reader->HasIndex() )
+ result &= reader->CreateIndex(type);
}
+
return result;
}
-IBamMultiMerger* BamMultiReaderPrivate::CreateMergerForCurrentSortOrder(void) const {
- switch ( m_sortOrder ) {
- case ( BamMultiReader::SortedByPosition ) : return new PositionMultiMerger;
- case ( BamMultiReader::SortedByReadName ) : return new ReadNameMultiMerger;
- case ( BamMultiReader::Unsorted ) : return new UnsortedMultiMerger;
- default : //print error
- return 0;
- }
-}
+IMultiMerger* BamMultiReaderPrivate::CreateAlignmentCache(void) const {
-const string BamMultiReaderPrivate::ExtractReadGroup(const string& headerLine) const {
+ // fetch SamHeader
+ SamHeader header = GetHeader();
- string readGroup("");
- stringstream headerLineSs(headerLine);
- string part;
+ // if BAM files are sorted by position
+ if ( header.SortOrder == Constants::SAM_HD_SORTORDER_COORDINATE )
+ return new MultiMerger<Algorithms::Sort::ByPosition>();
- // parse @RG header line, looking for the ID: tag
- while( getline(headerLineSs, part, '\t') ) {
- stringstream partSs(part);
- string subtag;
- getline(partSs, subtag, ':');
- if ( subtag == "ID" ) {
- getline(partSs, readGroup, ':');
- break;
- }
- }
- return readGroup;
+ // if BAM files are sorted by read name
+ if ( header.SortOrder == Constants::SAM_HD_SORTORDER_QUERYNAME )
+ return new MultiMerger<Algorithms::Sort::ByName>();
+
+ // otherwise "unknown" or "unsorted", use unsorted merger and just read in
+ return new MultiMerger<Algorithms::Sort::Unsorted>();
}
-// makes a virtual, unified header for all the bam files in the multireader
-const string BamMultiReaderPrivate::GetHeaderText(void) const {
+const vector<string> BamMultiReaderPrivate::Filenames(void) const {
- // just spit single header out if only have one reader open
- if ( m_readers.size() == 1 ) {
+ // init filename container
+ vector<string> filenames;
+ filenames.reserve( m_readers.size() );
+ // iterate over readers
+ vector<MergeItem>::const_iterator itemIter = m_readers.begin();
+ vector<MergeItem>::const_iterator itemEnd = m_readers.end();
+ for ( ; itemIter != itemEnd; ++itemIter ) {
+ const MergeItem& item = (*itemIter);
+ const BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
- vector<ReaderAlignment>::const_iterator readerBegin = m_readers.begin();
- const ReaderAlignment& entry = (*readerBegin);
- const BamReader* reader = entry.first;
- if ( reader == 0 ) return "";
- return reader->GetHeaderText();
+ // store filename if not empty
+ const string& filename = reader->GetFilename();
+ if ( !filename.empty() )
+ filenames.push_back(filename);
}
- string mergedHeader = "";
- map<string, bool> readGroups;
+ // return result
+ return filenames;
+}
- // foreach extraction entry (each BAM file)
- vector<ReaderAlignment>::const_iterator readerBegin = m_readers.begin();
- vector<ReaderAlignment>::const_iterator readerIter = readerBegin;
- vector<ReaderAlignment>::const_iterator readerEnd = m_readers.end();
- for ( ; readerIter != readerEnd; ++readerIter ) {
+SamHeader BamMultiReaderPrivate::GetHeader(void) const {
+ const string& text = GetHeaderText();
+ return SamHeader(text);
+}
- // get header from reader
- const BamReader* reader = (*readerIter).first;
+// makes a virtual, unified header for all the bam files in the multireader
+string BamMultiReaderPrivate::GetHeaderText(void) const {
+
+ // N.B. - right now, simply copies all header data from first BAM,
+ // and then appends RG's from other BAM files
+ // TODO: make this more intelligent wrt other header lines/fields
+
+ // if no readers open
+ const size_t numReaders = m_readers.size();
+ if ( numReaders == 0 ) return string();
+
+ // retrieve first reader's header
+ const MergeItem& firstItem = m_readers.front();
+ const BamReader* reader = firstItem.Reader;
+ if ( reader == 0 ) return string();
+ SamHeader mergedHeader = reader->GetHeader();
+
+ // iterate over any remaining readers (skipping the first)
+ for ( size_t i = 1; i < numReaders; ++i ) {
+ const MergeItem& item = m_readers.at(i);
+ const BamReader* reader = item.Reader;
if ( reader == 0 ) continue;
- string headerText = reader->GetHeaderText();
- if ( headerText.empty() ) continue;
-
- // store header text in lines
- map<string, bool> currentFileReadGroups;
- const vector<string> lines = SplitHeaderText(headerText);
-
- // iterate over header lines
- vector<string>::const_iterator linesIter = lines.begin();
- vector<string>::const_iterator linesEnd = lines.end();
- for ( ; linesIter != linesEnd; ++linesIter ) {
-
- // get next line from header, skip if empty
- const string headerLine = (*linesIter);
- if ( headerLine.empty() ) { continue; }
-
- // if first file, save HD & SQ entries
- if ( readerIter == readerBegin ) {
- if ( headerLine.find("@HD") == 0 || headerLine.find("@SQ") == 0) {
- mergedHeader.append(headerLine.c_str());
- mergedHeader.append(1, '\n');
- }
- }
- // (for all files) append RG entries if they are unique
- if ( headerLine.find("@RG") == 0 ) {
-
- // extract read group name from line
- const string readGroup = ExtractReadGroup(headerLine);
-
- // make sure not to duplicate @RG entries
- if ( readGroups.find(readGroup) == readGroups.end() ) {
- mergedHeader.append(headerLine.c_str() );
- mergedHeader.append(1, '\n');
- readGroups[readGroup] = true;
- currentFileReadGroups[readGroup] = true;
- } else {
- // warn iff we are reading one file and discover duplicated @RG tags in the header
- // otherwise, we emit no warning, as we might be merging multiple BAM files with identical @RG tags
- if ( currentFileReadGroups.find(readGroup) != currentFileReadGroups.end() ) {
- cerr << "WARNING: duplicate @RG tag " << readGroup
- << " entry in header of " << reader->GetFilename() << endl;
- }
- }
- }
- }
+ // retrieve current reader's header
+ const SamHeader currentHeader = reader->GetHeader();
+
+ // append current reader's RG entries to merged header
+ // N.B. - SamReadGroupDictionary handles duplicate-checking
+ mergedHeader.ReadGroups.Add(currentHeader.ReadGroups);
+
+ // TODO: merge anything else??
}
- // return merged header text
- return mergedHeader;
+ // return stringified header
+ return mergedHeader.ToString();
}
// get next alignment among all files
bool BamMultiReaderPrivate::GetNextAlignment(BamAlignment& al) {
- return LoadNextAlignment(al, false);
+ return PopNextCachedAlignment(al, true);
}
// get next alignment among all files without parsing character data from alignments
bool BamMultiReaderPrivate::GetNextAlignmentCore(BamAlignment& al) {
- return LoadNextAlignment(al, true);
+ return PopNextCachedAlignment(al, false);
}
// ---------------------------------------------------------------------------------------
//
// NB: The following GetReferenceX() functions assume that we have identical
-// references for all BAM files. We enforce this by invoking the above
-// validation function (ValidateReaders) to verify that our reference data
-// is the same across all files on Open, so we will not encounter a situation
-// in which there is a mismatch and we are still live.
+// references for all BAM files. We enforce this by invoking the
+// ValidateReaders() method to verify that our reference data is the same
+// across all files on Open - so we will not encounter a situation in which
+// there is a mismatch and we are still live.
//
// ---------------------------------------------------------------------------------------
// returns the number of reference sequences
-const int BamMultiReaderPrivate::GetReferenceCount(void) const {
- const ReaderAlignment& firstReader = m_readers.front();
- const BamReader* reader = firstReader.first;
+int BamMultiReaderPrivate::GetReferenceCount(void) const {
+
+ // handle empty multireader
+ if ( m_readers.empty() )
+ return 0;
+
+ // return reference count from first reader
+ const MergeItem& item = m_readers.front();
+ const BamReader* reader = item.Reader;
if ( reader ) return reader->GetReferenceCount();
- else return 0;
+
+ // invalid reader
+ return 0;
}
// returns vector of reference objects
const RefVector BamMultiReaderPrivate::GetReferenceData(void) const {
- const ReaderAlignment& firstReader = m_readers.front();
- const BamReader* reader = firstReader.first;
+
+ // handle empty multireader
+ if ( m_readers.empty() )
+ return RefVector();
+
+ // return reference data from first BamReader
+ const MergeItem& item = m_readers.front();
+ const BamReader* reader = item.Reader;
if ( reader ) return reader->GetReferenceData();
- else return RefVector();
+
+ // invalid reader
+ return RefVector();
}
// returns refID from reference name
-const int BamMultiReaderPrivate::GetReferenceID(const string& refName) const {
- const ReaderAlignment& firstReader = m_readers.front();
- const BamReader* reader = firstReader.first;
+int BamMultiReaderPrivate::GetReferenceID(const string& refName) const {
+
+ // handle empty multireader
+ if ( m_readers.empty() )
+ return -1;
+
+ // return reference ID from first BamReader
+ const MergeItem& item = m_readers.front();
+ const BamReader* reader = item.Reader;
if ( reader ) return reader->GetReferenceID(refName);
- else return -1; // ERROR case - how to report
-}
+ // invalid reader
+ return -1;
+}
// ---------------------------------------------------------------------------------------
-// checks if any readers still have alignments
-bool BamMultiReaderPrivate::HasOpenReaders(void) {
- return ( m_alignments->Size() > 0 );
+// returns true if all readers have index data available
+// this is useful to indicate whether Jump() or SetRegion() are possible
+bool BamMultiReaderPrivate::HasIndexes(void) const {
+
+ // handle empty multireader
+ if ( m_readers.empty() )
+ return false;
+
+ bool result = true;
+
+ // iterate over readers
+ vector<MergeItem>::const_iterator readerIter = m_readers.begin();
+ vector<MergeItem>::const_iterator readerEnd = m_readers.end();
+ for ( ; readerIter != readerEnd; ++readerIter ) {
+ const MergeItem& item = (*readerIter);
+ const BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
+
+ // see if current reader has index data
+ result &= reader->HasIndex();
+ }
+
+ return result;
}
-// returns whether underlying BAM readers ALL have an index loaded
-// this is useful to indicate whether Jump() or SetRegion() are possible
-bool BamMultiReaderPrivate::IsIndexLoaded(void) const {
- bool ok = true;
- vector<ReaderAlignment>::const_iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::const_iterator readerEnd = m_readers.end();
+// returns true if multireader has open readers
+bool BamMultiReaderPrivate::HasOpenReaders(void) {
+
+ // iterate over readers
+ vector<MergeItem>::const_iterator readerIter = m_readers.begin();
+ vector<MergeItem>::const_iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- const BamReader* reader = (*readerIter).first;
- if ( reader ) ok &= reader->IsIndexLoaded();
+ const MergeItem& item = (*readerIter);
+ const BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
+
+ // return true whenever an open reader is found
+ if ( reader->IsOpen() ) return true;
}
- return ok;
+
+ // no readers open
+ return false;
}
-// jumps to specified region(refID, leftBound) in BAM files, returns success/fail
+// performs random-access jump using (refID, position) as a left-bound
bool BamMultiReaderPrivate::Jump(int refID, int position) {
- bool ok = true;
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
+ // NB: While it may make sense to track readers in which we can
+ // successfully Jump, in practice a failure of Jump means "no
+ // alignments here." It makes sense to simply accept the failure,
+ // UpdateAlignments(), and continue.
+
+ // iterate over readers
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
if ( reader == 0 ) continue;
- ok &= reader->Jump(refID, position);
- if ( !ok ) {
- cerr << "ERROR: could not jump " << reader->GetFilename()
+ // attempt jump() on each
+ if ( !reader->Jump(refID, position) ) {
+ cerr << "BamMultiReader ERROR: could not jump " << reader->GetFilename()
<< " to " << refID << ":" << position << endl;
- exit(1);
}
}
- if (ok) UpdateAlignments();
- return ok;
+ // returns status of cache update
+ return UpdateAlignmentCache();
}
-bool BamMultiReaderPrivate::LoadNextAlignment(BamAlignment& al, bool coreMode) {
+// locate (& load) index files for BAM readers that don't already have one loaded
+bool BamMultiReaderPrivate::LocateIndexes(const BamIndex::IndexType& preferredType) {
- // bail out if no more data to process
- if ( !HasOpenReaders() ) return false;
-
- // "pop" next alignment and reader
- ReaderAlignment nextReaderAlignment = m_alignments->TakeFirst();
- BamReader* reader = nextReaderAlignment.first;
- BamAlignment* alignment = nextReaderAlignment.second;
+ bool result = true;
- // save it by copy to our argument
- al = BamAlignment(*alignment);
+ // iterate over readers
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
+ for ( ; readerIter != readerEnd; ++readerIter ) {
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
- // peek to next alignment & store in cache
- m_isCoreMode = coreMode;
- SaveNextAlignment(reader,alignment);
+ // if reader has no index, try to locate one
+ if ( !reader->HasIndex() )
+ result &= reader->LocateIndex(preferredType);
+ }
- // return success
- return true;
+ return result;
}
// opens BAM files
-bool BamMultiReaderPrivate::Open(const vector<string>& filenames,
- bool openIndexes,
- bool coreMode,
- bool preferStandardIndex)
-{
- // store core mode flag
- m_isCoreMode = coreMode;
-
- // first clear out any prior alignment cache prior data
- if ( m_alignments ) {
- m_alignments->Clear();
- delete m_alignments;
- m_alignments = 0;
- }
+bool BamMultiReaderPrivate::Open(const vector<string>& filenames) {
+
+ bool openedOk = true;
- // create alignment cache based on sorting mode
- m_alignments = CreateMergerForCurrentSortOrder();
- if ( m_alignments == 0 ) return false;
+ // put all current readers back at beginning
+ openedOk &= Rewind();
// iterate over filenames
vector<string>::const_iterator filenameIter = filenames.begin();
vector<string>::const_iterator filenameEnd = filenames.end();
for ( ; filenameIter != filenameEnd; ++filenameIter ) {
- const string filename = (*filenameIter);
+ const string& filename = (*filenameIter);
+ if ( filename.empty() ) continue;
- bool openedOk = true;
+ // attempt to open BamReader
BamReader* reader = new BamReader;
- openedOk = reader->Open(filename, "", openIndexes, preferStandardIndex);
+ const bool readerOpened = reader->Open(filename);
- // if file opened ok
- if ( openedOk ) {
+ // if opened OK, store it
+ if ( readerOpened )
+ m_readers.push_back( MergeItem(reader, new BamAlignment) );
- // try to read first alignment
- bool fileOk = true;
- BamAlignment* alignment = new BamAlignment;
- fileOk &= ( coreMode ? reader->GetNextAlignmentCore(*alignment)
- : reader->GetNextAlignment(*alignment) );
+ // otherwise clean up invalid reader
+ else delete reader;
- if ( fileOk ) {
+ // update method return status
+ openedOk &= readerOpened;
+ }
- m_readers.push_back( make_pair(reader, alignment) );
- m_alignments->Add( make_pair(reader, alignment) );
+ // if more than one reader open, check for consistency
+ if ( m_readers.size() > 1 )
+ openedOk &= ValidateReaders();
- } else {
- cerr << "WARNING: could not read first alignment in "
- << filename << ", ignoring file" << endl;
+ // update alignment cache
+ openedOk &= UpdateAlignmentCache();
- // if only file available & could not be read, return failure
- if ( filenames.size() == 1 )
- return false;
- }
+ // return success
+ return openedOk;
+}
+
+bool BamMultiReaderPrivate::OpenFile(const std::string& filename) {
+ vector<string> filenames(1, filename);
+ return Open(filenames);
+}
+
+bool BamMultiReaderPrivate::OpenIndexes(const vector<string>& indexFilenames) {
+
+ // TODO: This needs to be cleaner - should not assume same order.
+ // And either way, shouldn't start at first reader. Should start at
+ // first reader without an index?
+
+ // make sure same number of index filenames as readers
+ if ( m_readers.size() != indexFilenames.size() )
+ return false;
+
+ // init result flag
+ bool result = true;
+
+ // iterate over BamReaders
+ vector<string>::const_iterator indexFilenameIter = indexFilenames.begin();
+ vector<string>::const_iterator indexFilenameEnd = indexFilenames.end();
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
+ for ( ; readerIter != readerEnd; ++readerIter ) {
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
+
+ // open index filename on reader
+ if ( reader ) {
+ const string& indexFilename = (*indexFilenameIter);
+ result &= reader->OpenIndex(indexFilename);
}
- // TODO; any further error handling when openedOK is false ??
- else return false;
+ // increment filename iterator, skip if no more index files to open
+ if ( ++indexFilenameIter == indexFilenameEnd )
+ break;
}
- // files opened ok, at least one alignment could be read,
- // now need to check that all files use same reference data
- ValidateReaders();
+ // TODO: any validation needed here??
+
+ // return success/fail
+ return result;
+}
+
+bool BamMultiReaderPrivate::PopNextCachedAlignment(BamAlignment& al, const bool needCharData) {
+
+ // skip if no alignments available
+ if ( m_alignmentCache == 0 || m_alignmentCache->IsEmpty() )
+ return false;
+
+ // pop next merge item entry from cache
+ MergeItem item = m_alignmentCache->TakeFirst();
+ BamReader* reader = item.Reader;
+ BamAlignment* alignment = item.Alignment;
+ if ( reader == 0 || alignment == 0 )
+ return false;
+
+ // set char data if requested
+ if ( needCharData ) {
+ alignment->BuildCharData();
+ alignment->Filename = reader->GetFilename();
+ }
+
+ // store cached alignment into destination parameter (by copy)
+ al = *alignment;
+
+ // load next alignment from reader & store in cache
+ SaveNextAlignment(reader, alignment);
+
+ // return success
return true;
}
-// print associated filenames to stdout
-void BamMultiReaderPrivate::PrintFilenames(void) const {
+// returns BAM file pointers to beginning of alignment data & resets alignment cache
+bool BamMultiReaderPrivate::Rewind(void) {
- vector<ReaderAlignment>::const_iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::const_iterator readerEnd = m_readers.end();
- for ( ; readerIter != readerEnd; ++readerIter ) {
- const BamReader* reader = (*readerIter).first;
- if ( reader == 0 ) continue;
- cout << reader->GetFilename() << endl;
+ // attempt to rewind files
+ if ( !RewindReaders() ) {
+ cerr << "BamMultiReader ERROR: could not rewind file(s) successfully";
+ return false;
}
+
+ // return status of cache update
+ return UpdateAlignmentCache();
}
// returns BAM file pointers to beginning of alignment data
-bool BamMultiReaderPrivate::Rewind(void) {
+bool BamMultiReaderPrivate::RewindReaders(void) {
bool result = true;
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
+
+ // iterate over readers
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
if ( reader == 0 ) continue;
+
+ // attempt rewind on BamReader
result &= reader->Rewind();
}
+
return result;
}
void BamMultiReaderPrivate::SaveNextAlignment(BamReader* reader, BamAlignment* alignment) {
- // must be in core mode && NOT sorting by read name to call GNACore()
- if ( m_isCoreMode && m_sortOrder != BamMultiReader::SortedByReadName ) {
- if ( reader->GetNextAlignmentCore(*alignment) )
- m_alignments->Add( make_pair(reader, alignment) );
- }
-
- // not in core mode and/or sorting by readname, must call GNA()
- else {
- if ( reader->GetNextAlignment(*alignment) )
- m_alignments->Add( make_pair(reader, alignment) );
- }
+ // if can read alignment from reader, store in cache
+ // N.B. - lazy building of alignment's char data,
+ // only populated on demand by sorting merger or client call to GetNextAlignment()
+ if ( reader->GetNextAlignmentCore(*alignment) )
+ m_alignmentCache->Add(MergeItem(reader, alignment));
}
// sets the index caching mode on the readers
-void BamMultiReaderPrivate::SetIndexCacheMode(const BamIndex::BamIndexCacheMode mode) {
+void BamMultiReaderPrivate::SetIndexCacheMode(const BamIndex::IndexCacheMode mode) {
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
+ // iterate over readers
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
if ( reader == 0 ) continue;
+
+ // set reader's index cache mode
reader->SetIndexCacheMode(mode);
}
}
// alignments here." It makes sense to simply accept the failure,
// UpdateAlignments(), and continue.
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
+ // iterate over alignments
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
if ( reader == 0 ) continue;
+
+ // attempt to set BamReader's region of interest
if ( !reader->SetRegion(region) ) {
- cerr << "ERROR: could not jump " << reader->GetFilename() << " to "
+ cerr << "BamMultiReader WARNING: could not jump " << reader->GetFilename() << " to "
<< region.LeftRefID << ":" << region.LeftPosition << ".."
<< region.RightRefID << ":" << region.RightPosition << endl;
}
}
- UpdateAlignments();
- return true;
-}
-
-void BamMultiReaderPrivate::SetSortOrder(const BamMultiReader::SortOrder& order) {
-
- // skip if no change needed
- if ( m_sortOrder == order ) return;
-
- // set new sort order
- m_sortOrder = order;
-
- // create new alignment cache based on sort order
- IBamMultiMerger* newAlignmentCache = CreateMergerForCurrentSortOrder();
- if ( newAlignmentCache == 0 ) return; // print error?
-
- // copy old cache contents to new cache
- while ( m_alignments->Size() > 0 ) {
- ReaderAlignment value = m_alignments->TakeFirst();
- newAlignmentCache->Add(value);
- }
-
- // remove old cache structure & point to new cache
- delete m_alignments;
- m_alignments = newAlignmentCache;
+ // return status of cache update
+ return UpdateAlignmentCache();
}
// updates our alignment cache
-void BamMultiReaderPrivate::UpdateAlignments(void) {
+bool BamMultiReaderPrivate::UpdateAlignmentCache(void) {
+
+ // create alignment cache if not created yet
+ if ( m_alignmentCache == 0 ) {
+ m_alignmentCache = CreateAlignmentCache();
+ if ( m_alignmentCache == 0 ) {
+ // set error string
+ return false;
+ }
+ }
- // clear the cache
- m_alignments->Clear();
+ // clear any prior cache data
+ m_alignmentCache->Clear();
// iterate over readers
- vector<ReaderAlignment>::iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::iterator readerEnd = m_readers.end();
+ vector<MergeItem>::iterator readerIter = m_readers.begin();
+ vector<MergeItem>::iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
- BamReader* reader = (*readerIter).first;
- BamAlignment* alignment = (*readerIter).second;
- if ( reader == 0 ) continue;
+ MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
+ BamAlignment* alignment = item.Alignment;
+ if ( reader == 0 || alignment == 0 ) continue;
+
+ // save next alignment from each reader in cache
SaveNextAlignment(reader, alignment);
}
-}
-// splits the entire header into a list of strings
-const vector<string> BamMultiReaderPrivate::SplitHeaderText(const string& headerText) const {
- stringstream header(headerText);
- vector<string> lines;
- string item;
- while ( getline(header, item) )
- lines.push_back(item);
- return lines;
+ // if we get here, ok
+ return true;
}
// ValidateReaders checks that all the readers point to BAM files representing
// alignments against the same set of reference sequences, and that the
// sequences are identically ordered. If these checks fail the operation of
// the multireader is undefined, so we force program exit.
-void BamMultiReaderPrivate::ValidateReaders(void) const {
+bool BamMultiReaderPrivate::ValidateReaders(void) const {
+
+ // skip if no readers opened
+ if ( m_readers.empty() )
+ return true;
+
+ // retrieve first reader
+ const MergeItem& firstItem = m_readers.front();
+ const BamReader* firstReader = firstItem.Reader;
+ if ( firstReader == 0 ) return false;
- // retrieve first reader data
- const BamReader* firstReader = m_readers.front().first;
- if ( firstReader == 0 ) return; // signal error?
- const RefVector firstReaderRefData = firstReader->GetReferenceData();
+ // retrieve first reader's header data
+ const SamHeader& firstReaderHeader = firstReader->GetHeader();
+ const string& firstReaderSortOrder = firstReaderHeader.SortOrder;
+
+ // retrieve first reader's reference data
+ const RefVector& firstReaderRefData = firstReader->GetReferenceData();
const int firstReaderRefCount = firstReader->GetReferenceCount();
const int firstReaderRefSize = firstReaderRefData.size();
// iterate over all readers
- vector<ReaderAlignment>::const_iterator readerIter = m_readers.begin();
- vector<ReaderAlignment>::const_iterator readerEnd = m_readers.end();
+ vector<MergeItem>::const_iterator readerIter = m_readers.begin();
+ vector<MergeItem>::const_iterator readerEnd = m_readers.end();
for ( ; readerIter != readerEnd; ++readerIter ) {
+ const MergeItem& item = (*readerIter);
+ BamReader* reader = item.Reader;
+ if ( reader == 0 ) continue;
- // get current reader data
- BamReader* reader = (*readerIter).first;
- if ( reader == 0 ) continue; // error?
+ // get current reader's header data
+ const SamHeader& currentReaderHeader = reader->GetHeader();
+ const string& currentReaderSortOrder = currentReaderHeader.SortOrder;
+
+ // check compatible sort order
+ if ( currentReaderSortOrder != firstReaderSortOrder ) {
+ // error string
+ cerr << "BamMultiReader ERROR: mismatched sort order in " << reader->GetFilename()
+ << ", expected " << firstReaderSortOrder
+ << ", but found " << currentReaderSortOrder << endl;
+ return false;
+ }
+
+ // get current reader's reference data
const RefVector currentReaderRefData = reader->GetReferenceData();
const int currentReaderRefCount = reader->GetReferenceCount();
const int currentReaderRefSize = currentReaderRefData.size();
- // init container iterators
+ // init reference data iterators
RefVector::const_iterator firstRefIter = firstReaderRefData.begin();
RefVector::const_iterator firstRefEnd = firstReaderRefData.end();
RefVector::const_iterator currentRefIter = currentReaderRefData.begin();
if ( (currentReaderRefCount != firstReaderRefCount) ||
(firstReaderRefSize != currentReaderRefSize) )
{
- cerr << "ERROR: mismatched number of references in " << reader->GetFilename()
+ cerr << "BamMultiReader ERROR: mismatched number of references in " << reader->GetFilename()
<< " expected " << firstReaderRefCount
<< " reference sequences but only found " << currentReaderRefCount << endl;
- exit(1);
+ return false;
}
// this will be ok; we just checked above that we have identically-sized sets of references
// here we simply check if they are all, in fact, equal in content
while ( firstRefIter != firstRefEnd ) {
-
const RefData& firstRef = (*firstRefIter);
const RefData& currentRef = (*currentRefIter);
if ( (firstRef.RefName != currentRef.RefName) ||
(firstRef.RefLength != currentRef.RefLength) )
{
- cerr << "ERROR: mismatched references found in " << reader->GetFilename()
+ cerr << "BamMultiReader ERROR: mismatched references found in " << reader->GetFilename()
<< " expected: " << endl;
+
+ // print first reader's reference data
RefVector::const_iterator refIter = firstReaderRefData.begin();
RefVector::const_iterator refEnd = firstReaderRefData.end();
for ( ; refIter != refEnd; ++refIter ) {
}
cerr << "but found: " << endl;
+
+ // print current reader's reference data
refIter = currentReaderRefData.begin();
refEnd = currentReaderRefData.end();
for ( ; refIter != refEnd; ++refIter ) {
cerr << entry.RefName << " " << entry.RefLength << endl;
}
- exit(1);
+ return false;
}
// update iterators
++currentRefIter;
}
}
+
+ // if we get here, everything checks out
+ return true;
}